Washington Academy of Sciences

Founded in 1898

EXECUTIVE COMMITTEE

President

Robert M. McCracken

President-Elect

Armand B. Weiss

Secretary

Donald O. Buttermore

Treasurer

Grover C. Sherlin

Past President

James E. Spates

Vice President, Membership Affairs

M. Sue Bogner

Vice President, Administrative Affairs

William H. Benesch

Vice President, Junior Academy Affairs

Marylin F. Krupsaw

Vice President, Affiliate Affairs

Edith L. R. Corliss

Board of Managers

R. Clifton Bailey

Jean K. Boek

Betty Jane Long

Lawson M. McKenzie

‘Thomas N. Pyke

Marcia S. Smith

REPRESENTATIVES FROM

AFFILIATED SOCIETIES |

Delegates are listed on inside rear

cover of each Journal.

ACADEMY OFFICE

1101 N. Highland Street

Arlington, VA 22201

Phone: (703) 527-4800

EDITORIAL BOARD

Editor:

John J. O’Hare, CAE-Link Corpora-

tion

Associate Editors:

Albert G. Gluckman, University of

Maryland

Joseph H. Neale, Georgetown Uni-

versity

Marc Rothenberg, Smithsonian Insti-

tution

Marc M. Sebrechts, Catholic Univer-

sity of America

Edward J. Wegman, George Mason

University

The Journal

This journal, the official organ of the Wash-

ington Academy of Sciences, publishes orig-

inal scientific research, critical reviews,

historical articles, proceedings of scholarly

meetings of its affiliated societies, reports of

the Academy, and other items of interest to

Academy members. The Journal appears

four times a year (March, June, September,

and December). The December issue con-

tains a directory of the current membership

of the Academy.

Subscription Rates

Members, fellows, and life members in good

standing receive the Journal without charge.

Subscriptions are available on a calendar year

basis, payable in advance. Payment must be

made in U.S. currency at the following rates:

US. and Canada... >...) ).33eeeaee $19.00

Other countries... .. .:....2..3eeeeeee 22.00

Single copies, when available.........

Claims for Missing Issues

Claims will not be allowed if received more

than 60 days after the day of ae plus time

normally required for postal delivery and

claim. No claims will be allowed because of

failure to notify the Academy of a change of

address.

Notification of Change of Address

Address changes should be sent promptly to

the Academy office. Such notifications

should show both old and new addresses and

zip-code numbers, where applicable.

Published quarterly in March, June, September, and December of each year by the

Washington Academy of Sciences, 1101 N. Highland Street, Arlington, VA 22201.

Second-class postage paid at Arlington, VA, and additional mailing offices.

Journal of the Washington Academy of Sciences,

Volume 80, Number 1, Pages 1-15, March 1990

Antecedents of Pregnancy

Among Women Marines

Meg Gerrard and Teddy D. Warner

Iowa State University

ABSTRACT

Nine hundred fifty-six women Marines participated in a prospective study of the an-

tecedents of pregnancy. The results indicated that attitudes toward pregnancy assessed

during recruit training and contraceptive behavior prior to entry into the Marine Corps

predict pregnancy. The study also revealed that the pregnancy rate among first term

women Marines is significantly higher than that for other women the same age.

Antecedents of Pregnancy Among Women Marines

Becoming an effective contraceptor requires that a woman negotiate a com-

plex sequence of psychological and behavioral events including: (a) becoming

aware of the high risk of pregnancy when intercourse is unprotected, (b)

obtaining adequate information about contraception, (c) acquiring the con-

traceptive devices and/or knowledge necessary to prevent pregnancy, and (d)

using those devices and/or information consistently and effectively.'* Many

young women fail to negotiate these steps successfully. In fact, it is estimated

that 24 percent of 18 year old women, and 44 percent of 20 year old women

have experienced at least one pregnancy.’ Numerous studies have indicated

that most of these pregnancies are unintended.**>

Antecedents of Unplanned Pregnancy

In spite of the fact that birth control information and effective contraceptive

methods are available to most young women in the United States, substantial

numbers of sexually active women use unreliable methods of contraception

(e.g., rhythm, withdrawal) or use no method at all. Research on the ante-

cedents of unplanned pregnancy indicates that stress, anxiety and major life

transitions are associated with irregular or ineffective contraception. More

specifically, Miller® suggests that women are particularly susceptible to de-

2 MEG GERRARD AND TEDDY D. WARNER

creased contraceptive vigilance when they move away from their nuclear fam-

ilies and/or enter new environments. He hypothesizes that these kinds of

situations may deplete a woman’s psychological energy and detract from her

ability to be continuously vigilant in avoiding conception. It is not surprising

then, that confusion about career goals is also associated with ineffective use

of contraceptives which lead to unplanned pregnancy.’

One of the most reliable findings regarding psychological antecedents of

unplanned pregnancy is that ineffective contraception is inhibited by a negative

orientation toward sex. This predisposition results in conservative or conflicted

attitudes towards sex and discomfort with the decision to have sexual inter-

course.'® Women with these negative attitudes (called erotophobia or sex guilt)

tend to have difficulty engaging in rational decision making about contracep-

tion. For example, sex guilt and erotophobia have been associated with: (a)

lack of knowledge about sex and contraception;’ (b) avoiding information

about birth control;? (c) inconsistent and irrational attitudes and beliefs about

birth control;'° (d) choosing ineffective methods of contraception;'!™ and (e)

inconsistent use of chosen birth control methods.'*"* These relationships are

very reliable, but they have been demonstrated primarily on college student

samples.

One purpose of the current study was to examine the antecedents of un-

planned pregnancy in a non-student sample of women who are at relatively

high risk of unplanned pregnancy because of their frequency of sexual inter-

course and use of ineffective contraceptives. More specifically, the current

study was designed to examine the antecedents of unplanned pregnancy in

first term women Marines.

Women Marines

The preceding research on the situational and psychological antecedents of

unplanned pregnancy suggests that first term women Marines may be partic-

ularly at risk—they are sexually active, have just completed a very stressful

course in recruit training, are isolated from family and friends, and are likely

to be experiencing uncertainty about their career goals and their relationships

with men.

A series of studies conducted at the Naval Personnel Research and Devel-

opment Center in San Diego in the early 1980s investigated attitudes, char-

acteristics, and behaviors relevant to pregnancy and pregnancy attrition among

first term women Marines. In the first of these studies, Royle’ examined the

relation between background variables, experience in the Marine Corps, and

attrition among 1,271 recruits who entered the Marine Corps between 1976

ANTECEDENTS OF PREGNANCY 3

and 1980. Her data revealed that women who attrite (both for pregnancy and

for other reasons) have fewer “‘masculine”’ interests (e.g., sports) than preg-

nant women who remain in the Corps.

The second in this series of studies compared the backgrounds and Marine

Corps experiences of first term women Marines who attrite and those who do

not attrite.'° This survey of 142 women revealed that women Marines in general

have a relatively traditional orientation regarding having a family, and they

plan to combine motherhood with their careers. However, those with the most

traditional family orientations adapted least well to Marine Corps life—they

were less satisfied and less well-adjusted than those with a less traditional

orientation toward motherhood, suggesting that they would be most likely to

attrite because of pregnancy.

In an extension of this research, Gerrard and Royle’’ examined traditional

sex role orientation, feelings of isolation, and dissatisfaction with the Marine

Corps as possible predictors of both pregnancy and pregnancy attrition in 610

first term enlisted women Marines. We found that both pregnancy and preg-

nancy attrition were predicted by the traditional orientation identified in the

earlier two studies. Whether a woman left the service once she was pregnant,

was also determined in large part by her commitment to family vs. career. In

addition, we found that dissatisfaction with the Marine Corps did not discrim-

inate between non-pregnant women, pregnant women who remained in the

Corps, and women who had a pregnancy-related attrition. This latter finding

suggests that dissatisfaction with the Marine Corps was not related to becoming

pregnant.

A fourth study in this series!® assessed first term Marines’ sexual and con-

traceptive behaviors relevant to planned pregnancy. It revealed that many

women (and men) had not adequately protected themselves from planned

pregnancy prior to recruit training—10 percent of the women reported using

no method of birth control the last time they had intercourse, and another 11

percent reported using relatively ineffective methods (withdrawal and rhythm).

In addition, 16 percent of the women reported that they had experienced at

least one pregnancy prior to recruit training.

It is important to note that with the exception of 196 of the 1,271 women

in Royle! study, the women in all four of these studies were surveyed after

recruit training. Thus, these studies did not investigate prediction of pregnancy

from the women’s attitudes in recruit training or sexual and contraceptive

behaviors prior to recruit training.

Estimating pregnancy rates. The two previous studies that were designed to

predict pregnancy and pregnancy attrition?!’ both employed a limited clas-

sification system for pregnancy. In Royle,’ a woman was classified as pregnant

4 MEG GERRARD AND TEDDY D. WARNER

if Marine records indicated that she left the Corps because she was pregnant,

or indicated that she added an infant dependent between completion of recruit

training and completion of her enlistment. The second study'’ added self-

reported pregnancy, but only if the woman was pregnant at the time of the

survey. Thus, both studies failed to identify any women who became pregnant

after recruit training but did not deliver, either because of miscarriage or

abortion (unless in the latter study the women were pregnant at the time of

the survey). Likewise, women who got pregnant but left the Marine Corps

for reasons other than pregnancy were classified as non-pregnant in both

studies.

Two facts suggest that these methods of identifying pregnancies result in an

underestimation of the pregnancy rate among first term women Marines. The

first is pilot data on the sexual activity and contraceptive use of women Marines

reported in Gerrard and Royle.'’ Estimates based on these data suggest that

the actual pregnancy rate is significantly higher than the 6 percent to 16

pregnancy rates derived from the official records.!’ The second is data sug-

gesting that abortions are common among women in the military. Hoiberg”’

reports that approximately 10 percent of Navy women received abortions in

Navy hospitals each year between 1973 and 1975, and that for every 100 women

in the Navy who got pregnant and left the service, there were an additional

60 to 80 women who got pregnant, but chose to end the pregnancy by abortion.

Hoiberg’s data were collected between 15 and 18 years ago at a time when

abortions were being performed in Navy hospitals. Therefore, it is not ap-

propriate to extrapolate from these data to estimate current abortion or preg-

nancy rates. It is, however, reasonable to assume that a significant proportion

of women in the military still have abortions rather than carry their pregnancies

to term, and aborted pregnancies have not been included in previous studies

of pregnancy among women Marines.

Overview

The purpose of the current study was twofold. First it was designed to

examine the relationships between erotophobia, sexual behavior, contracep-

tive effectiveness and pregnancy in a non-student sample. And second, it was

designed to extend the findings of the previous series of studies of women

Marines by examining the attitudinal and behavioral antecedents of pregnancy

among first term women Marines. Specifically, this study was designed to

determine whether women Marines’ attitudes during recruit training and their

sexual and contraceptive behavior prior to entering the Marine Corps predict

pregnancy during the first term of enlistment.

ANTECEDENTS OF PREGNANCY 5

Method

Subjects

The participants were 956 women Marines who completed recruit training

between November 1986 and September 1987. All of the women were high

school graduates, and 23 percent had some college. Six hundred and twenty-

eight (66%) of the women were white, 208 (22%) were black, and 65 (7%)

were Hispanic. The mean age of the women was 19.5, and 91 percent were

single, 6 percent married and 3 percent divorced or separated.

Procedure

The first author asked for volunteers from 10 randomly selected recruit

training classes at the Women’s Battalion at the Parris Island Marine Corps

Recruit Training Depot between November 1986 and August 1987. Because

of the possibility that women in Marine Corps recruit training would feel

coerced even if they were told that their participation was voluntary, potential

participants were assured that there would be no adverse consequences as-

sociated with either skipping questions that they considered too personal, or

not completing the questionnaire. No military personnel other than the recruits

were present during the data collection sessions, and no military personnel

were aware of which potential participants completed either the initial ques-

tionnaire or the follow-up questionnaires. Ninety-eight percent of the-potential

participants agreed to participate in the study and completed the initial ques-

tionnaire.

The initial survey was administered in a classroom setting in groups ranging

in size from 66 to 112 within 2 weeks prior to graduation from recruit training.

Follow-up surveys were mailed to each woman at her duty station 6, 12, and

18 months after their graduation.

Measures

Sexual Opinion Survey. This 21 item instrument (SOS) was developed by

White, Fisher, Byrne and Kingma”’ to measure emotional reactions to sex-

uality (erotophobia/erotophilia). It assesses reactions to a variety of sexual

activities on a seven point scale (e.g., ““Masturbation can be exciting,”’ “I do

not enjoy day dreaming about sexual matters.’’) The SOS is related to affective

responses to erotica and to approach/avoidance reactions to a variety of

sex-related topics (for a review”!). It has also been shown to be related to

contraceptive knowledge, and use of effective contraceptive methods.”!' 4?"

4 The scale is unrelated to social desirability.”

6 MEG GERRARD AND TEDDY D. WARNER

Sexual and contraceptive history. This instrument is an adaptation of ques-

tionnaires used by Royle, Molof, Winchell, and Gerrard,'® and Geis and

Gerrard" to collect contraceptive and sexual histories from college students

and women Marines. It asks the woman to describe her sexual history and

contraceptive behavior in detail:

Starting with your first sexual partner, indicate all the periods of time you

were sexually active. For each [period] indicate the frequency of intercourse,

and the method of birth control you and your partner used. . . . Please start

with the first man you had sexual intercourse with and work forward... .

Be sure to include all periods of sexual activity even if you did not use any

method of birth control.

Attitudes about pregnancy. Because the women’s attitudes about getting

pregnant were considered to be very important, we assessed these attitudes

in a variety of ways. We inquired about whether the women planned to ever

have children, and if so, how many they would like to have and at what age

they would like to start their families. In addition, we had the women estimate

the likelihood that they would experience a pregnancy in the 12 months fol-

lowing recruit training, indicate how inconvenient they thought it would be if

they were to get pregnant in the 12 months following recruit training, and

indicate how unhappy they would be if they were to become pregnant during

that 12 months.

Birth control opinion questionnaire. This instrument was designed specifi-

cally for this study to measure attitudes toward, and biases against specific

birth control methods. It assesses the women’s perceptions of the effectiveness

of specific methods (on a 7 point scale from 1 = “extremely effective”, to

7 = “not at all effective’), and the woman’s intentions to use the methods

in the future (also on a 7 point scale).

Birth control knowledge. Birth control knowledge was assessed using a 23

item multiple choice instrument adapted from the knowledge test used by

Royle et al.'** This test is designed to assess information useful in avoiding

unplanned pregnancy rather than biological or technical information about

conception and contraception. The internal consistency of this instrument was

acceptable (alpha coefficient = .77).

Follow-up Surveys

The follow-up surveys (at 6, 12, and 18 months) were mailed to participants

at their duty stations. A second questionnaire was sent to women who failed

to return a follow-up questionnaire within four weeks after it was mailed. If

the questionnaire was returned “addressee unknown,” or “moved, left no

ANTECEDENTS OF PREGNANCY 7

forwarding address,”’ the address was confirmed with Marine Corps Head-

quarters, and a second copy was mailed. If the second mailing also resulted

in the return of the questionnaire by the post office or mail service at the

woman’s last duty station, the woman was counted as “‘unreachable”’ for that

follow-up. This procedure resulted in three possible sources of attrition from

the study: (1) failure to locate women for follow-up; (2) participant failure to

respond; and (3) attrition from the Marine Corps.

Response rate. The response rate (responders/ (initial participants - unreach-

able participants - attrites) for the six month follow-up was 46 percent, at the

12 month follow-up was 38 percent, and at the 18 month follow-up was 30

percent. Responders and nonresponders were not significantly different in

terms of age, education, IQ, ethnic background, sexual and contraceptive

attitudes or behaviors prior to recruit training, or initial attitudes toward

motherhood.

Indicators of Pregnancy

A woman was classified as pregnant in the current study in three ways: The

official record of pregnancy attrition provided by Headquarters U.S. Marine

Corps, indication on Marine Corps records that the woman had added an

infant dependent between completion of recruit training and the end of the

study, and self-report of pregnancy since recruit training on follow-up ques-

tionnaires.

Results

Sexual and Contraceptive Experience Prior to Recruit Training

As a group the women entered recruit training with a significant amount

of sexual experience (see Table 1). Eighty-five percent had engaged in sexual

intercourse prior to joining the Marine Corps, with the nonvirgins reporting

an average of 5.7 sexual partners. The sexually experienced women reported

having intercourse an average of 8.9 times per month in the 3 months im-

mediately prior to recruit training.**

At the initial survey administration, the women were asked to indicate which

method(s) of birth control they usually used prior of recruit training and which

method(s) they used the Jast time they had intercourse. Fifty-five percent of

the nonvirgins reported using relatively effective methods (i.e., oral contra-

ceptive or condoms) the last time that they had intercourse. Fifteen percent

reported using less effective methods (i.e., rhythm or withdrawal), and 19

8 MEG GERRARD AND TEDDY D. WARNER

Table 1.—Sexual Activity and Contraceptive Use Prior to Recruit Training

Had prior sexual experience 85.0

Frequency of intercourse* 8.9

Number of partners° Dall

Method Used at Method

Last Intercourse Usually Used

Oral contraceptives S990 46%

Condom 16 15

Rhythm 6 5

Withdrawal 10 6

None 19 14

“Average frequency of intercourse per month over the 3 months prior to recruit training.

"Total number of partners prior to recruit training.

percent reported using no method of birth control the last time that they had

intercourse (see Table 1).

Assuming that the birth control method a woman reports using the last time

she engaged in intercourse is a good predictor of her future contraceptive use,

and that her previous frequency of intercourse is a good predictor of her future

frequency, it is possible to compute a projected pregnancy rate for the sample.

The formula for this computation is

5! [P, FR Fq CE)

where P; = proportion of women using method 1

FR; = typical failure rate for method i

fq; = adjusted frequency of intercourse for women using method i

C = correction factor for missing data

P = proportion of women who are sexually active

Using this formula, we projected that between 21 and 25 percent of the

women in this sample would get pregnant in the first year of their enlistment. ***

Twenty-one percent is a conservative estimate based on the assumption that

women who were virgins during recruit training would not get pregnant in the

first year after recruit training. Twenty-five percent is the estimate based on

the assumption that these women would become sexually active, and that their

frequency of intercourse and contraceptive use would be comparable to that

of the women who were sexually experienced prior to recruit training (i.e.,

deleting P from the computation).

ANTECEDENTS OF PREGNANCY 9

Relationship Between Attitudes Toward Sex and Sexual and

Contraceptive Behavior

Attitudes toward sex do predict sexual activity in the women Marines—

erotophobic women Marines (the top one-half of the SOS distribution) were

almost twice as likely as erotophilic women Marines (the botton one-half of

the SOS distribution) to be virgins entering recruit training (19.2% vs. 10.0%;

z = 4.53, p < .01). The sexually active erotophobic women also reported

fewer sexual partners than did the sexually active erotophilic women (4.3 vs.

7.3; t(688) = 5.70, p < .001), and less frequent intercourse (7.9 times per

month vs. 10.2 times per month; t(706) = 3.44, p < .01).**** Comparison

of the erotophobic and erotophilic women’s contraceptive use, however, re-

vealed no significant differences in these groups’ use of contraceptives (all

ps > .90).

Attitudes toward Pregnancy During Recruit Training

During recruit training, 92 percent of the women reported that they planned

to have children at some time in the future, with the average number of

children planned being 2.4. The average age that they planned to have their

first child was 24.9. The mean response to the question ‘How inconvenient

would it be for you to get pregnant in the next year?” was 6.2 (on a scale

where 7 = extremely inconvenient). Their answers to the question ‘““How

unhappy would you be if you were to become pregnant in the next year?”’,

clearly demonstrated that a sizable proportion of the women were ambivalent

about the possibility of a pregnancy (m = 4.7 on a7 point scale where 7 =

“extremely unhappy’’).

Actual Pregnancy Rate

A simple additive computation of the pregnancy rate indicates that 25 per-

cent of the women would get pregnant during the first 12 months following

recruit training. This computation assumes that the rate at 12 months is the

sum of the number of self-reported pregnancies, plus the number of pregnancy

related attritions between 0 and 6 months, plus the number of self-reported

pregnancies, pregnancy related attritions, and new infant dependents between

6 and 12 months. A more conservative calculation (the number of women

reporting pregnancies)/(the total number of women who returned question-

naires at the 6 month and 12 month follow-up) results in an estimate of 18.2

percent.

Sixty percent of the women who conceived in the first 18 months after

recruit training reported that they intended to carry their pregnancies to term

10 MEG GERRARD AND TEDDY D. WARNER

and keep their babies. Twelve percent reported miscarriages and 19 percent

reported induced abortion. The remaining 9 percent were still pregnant and

undecided about their plans at the time of the survey.

Antecedents of Pregnancy

A series of exploratory multivariate analyses of variance was conducted to

identify attitudes and behaviors that could be used to predict which women

Marines became pregnant during the first 18 months after recruit training.

These analyses revealed that women who became pregnant were significantly

different at recruit training from those who avoided pregnancy on a number

of variables. These variables can be characterized along two dimensions: (1)

the attitudes about pregnancy the women held during recruit training, and (2)

their attitudes about contraception and their contraceptive behavior prior to

enlistment. Measures of these attitudes and behaviors were entered into a

discriminant function analysis which confirmed that they reliably predict which

women became pregnant (X? (df = 9) = 33.10, p < .01; see Table 2).

Altitudes toward pregnancy. Women Marines who became pregnant had

more positive attitudes toward pregnancy during recruit training than did

women Marines who did not become pregnant. More specifically, during

Table 2.—Differences Between Pregnant and Non-Pregnant Women Marines

Discriminant Function Analysis X? (df = 9) = 33.10, p < .01

Univariate Tests

Wilks’ Lambda

Variable Pregnant Nonpregnant |

Perceived convenience of pregnancy (in Sjozil 6.14 .908*

next 12 months)?

Plans to get pregnant in next 3 years? 4.61 5.42 .928*

Estimated likelihood of pregnancy (in 20.65 10573 .960*

next 12 months)°

Typical failure rate of method of birth 30.91 JHE) .946*

control used at last intercourse*

Knowledge of birth control‘ 18.66 19512 936"

Opinion of rhythm and withdrawal‘ 10.77 10.85 OTe

Intention to use rhythm and withdrawal 10.67 Pip39 .922*

Previous pregnancy" 1.63 1.75 529-

‘Rating scale ranges from 1 “‘not at all inconvenient” to 7 = “extremely inconvenient.”

’Rating scale ranges from 1 = ‘“‘definitely plan to get pregnant” to 7 = ‘‘definitely do not plan to get

pregnant.”

‘Rating scale ranges from 0 to 100.

‘Scale ranges from 0 to 100 with numbers indicating the typical likelihood of pregnancy over 12 months.

‘Scale ranges from 0-40; high scores indicate more knowledge.

‘Scale ranges from 2 = “‘extremely effective” to 14 = “extremely ineffective.”

*Scale ranges from 2 = “would definitely use” to 14 = “definitely would not use.”’

"Scale ranges from 1 to 2 with higher numbers indicating greater proportions have had a previous

pregnancy.

ap = 101

ANTECEDENTS OF PREGNANCY 11

recruit training, women who later became pregnant were significantly more

likely to report that they thought that a pregnancy within the next 12 months

would not be inconvenient (5.91 vs. 6.14 on a 7 point scale ranging from 1 =

‘not at all inconvenient” to 7 = ‘“‘extremely inconvenient’; Wilks’ Lambda

F = .91, p < .01). Women who became pregnant were also more likely to

report that they were planning to get pregnant in the next 3 years (4.61 vs.

5.42 on a 7 point scale where 1 = “definitely plan to get pregnant” and 7 =

“definitely do not plan to get pregnant”; Wilks’ Lambda F = .93, p < .01),

and were more likely than other recruits to report that they were likely to get

pregnant during the coming year (20.7 vs. 10.8 on a 100 point scale, p < .01).

Women who conceived during the first year after recruit training were less

likely to have previously experienced a pregnancy than those who did not

(Wilks’ Lambda = .929, p < .01).

Contraceptive behavior prior to entering recruit training. Pregnant women

Marines were more likely to have a history of unprotected intercourse and/

or inadequate contraceptive protection prior to recruit training than were the

women who did not get pregnant. The typical failure rate of the method birth

control the pregnant women used prior to recruit training was 30.9 percent

failure, as compared to 22.5 for the women who did not get pregnant ( Wilks’

Lambda = .95, p < .01). Consistent with this history, the women who became

pregnant were less knowledgeable about contraception, more likely to rate

rhythm and withdrawal as effective methods of birth control, and more likely

to report intentions to use rhythm, and withdrawal (both Wilks’ Lambdas >

.90, both ps < .01).

Discussion

Perhaps the most striking result from the current study is that the pregnancy

rate for first term women Marines is significantly higher than the rate for

other women of the same age—18 to 25 percent of this sample got pregnant

in the first year after recruit training versus a 10 to 11 percent pregnancy rate

per year for the general population.** In addition, pregnant women in the

current sample were significantly more likely to carry their pregnancies to

term and keep their babies than are other pregnant women their age (the

abortion rate for pregnant women age 18-24 is typically about 40%°” com-

pared to the 20% reported in this sample). These differences between women

Marines and the general population are consistent with the fact that women

Marines are more family oriented and are likely to be planning to have children

at an earlier age and than are other women their age.*°

12 MEG GERRARD AND TEDDY D. WARNER

There are a number of possible explanations for these differences between

women Marines and other women. One is that the Marine Corps attracts more

traditional (family oriented) women. These women then, are more likely to

have children. Another possibility is that the first term of service in the Marine

Corps presents women with opportunities conducive to sexual activity, (e.g.,

they are outnumbered by men approximately 20 to 1),*° or that women in the

masculine environment of the Marine Corps feel pressure to prove their fem-

ininity, and this pressure leads them to become involved in risky sexual re-

lationships that lead to pregnancy. Yet another possibility is that the stress

and major life change involved in entering the Marine Corps leads to either

less contraceptive vigilance or decreased motivation to avoid pregnancy.°’

Predicting Pregnancy

The variables that discriminate between those women Marines who get

pregnant and those who did not suggest that, in large part, differences in

knowledge, attitudes and behavior patterns that the women bring with them

into the Marine Corps are responsible for the relatively high pregnancy rate

among these women. That is, the best predictors of pregnancy in the first 18

months of service are lack of knowledge about birth control, positive attitudes

toward pregnancy reported during recruit training, and ineffective contracep-

tive behaviors practiced prior to recruit training. This does not rule out the

possibility that women Marines are less vigilant with their contraception after

undergoing the stress of recruit training, or that these women engage in riskier

sexual behaviors because of situational pressures. It is clear however, that

women Marines have more traditional attitudes and are more family oriented

than most women their age even before they leave recruit training to work in

a male dominated environment, and that these attitudes are associated with

pregnancy.

Planned vs. Unplanned Pregnancy

Any discussion of pregnancy in the Marine Corps would be incomplete

without mention of the possibility that women Marines who become pregnant

do so intentionally. The current data do indicate that some women Marine’s

pregnancies are planned, some are accidental, and some are the result of

ambivalence about pregnancy or emotional conflict about their sexual behav-

ior. It is impossible, however, to determine what percent of the women fall

into each of these three categories. It is clear though, that the majority of

pregnancies among first term women Marines fit the definition of unplanned

pregnancy—they are pregnancies which were not intended at the time of

ANTECEDENTS OF PREGNANCY 13

conception. In other words, ambivalence about pregnancy may inhibit women

from taking steps 3 and 4 outlined in the introduction, and thus result in some

pregnancies that were not intended, but were not entirely unwanted.

Erotophobia and Sexual and Contraceptive Behavior

Although the SOS scores of the women in this study predicted their sexual

activity, the current data fail to replicate the previously reported association

between negative attitudes toward sex and use of ineffective methods of con-

traception. More specifically, the erotophobic women Marines were not less

effective contraceptors than were the erotophilic women Marines. Two dif-

ferences between this sample of women Marines and the college student sam-

ples used in previous studies could be responsible for this difference. One is

that this sample is significantly more sexually experienced than college stu-

dents, both in terms of their total number of partners and in terms of their

frequency of intercourse.*® The other difference is that this sample of Marine

women had more experience with a variety of contraceptive methods than

have college women.” These differences between samples raise the possibility

that attitudes toward sex are predictive of contraceptive use only in less sex-

ually and contraceptively experienced samples, like college students. Regard-

less of the reason for the failure to replicate, however, these data suggest

caution in generalizing from research on college students’ contraceptive be-

havior to other samples.

Actual Pregnancy Rate

In spite of the relatively high pregnancy rate in this sample, one must

entertain the possibility that the current study has underestimated the actual

pregnancy rate among women Marines. Although the current study represents

an improvement in identifying pregnancies, our classification system was not

perfect. That is, there are three ways in which the current study could have

misclassified a pregnant women as non-pregnant. First, a woman in the current

study would have been classified as non-pregnant if she was pregnant but left

from the Marine Corps for reasons other than pregnancy. Second, a woman

who delivered a baby would not have been classified as pregnant if the baby

were not listed as her dependent but rather was listed as her husband’s de-

pendent. And third, we must consider the possibility that discomfort or un-

happiness about a pregnancy could have led some women to fail to respond

to the follow-up questionnaires. All of these factors raise the possibility that

the actual pregnancy rate among women Marines is higher than the 18 to 25

percent estimate the current data suggest.

14 MEG GERRARD AND TEDDY D. WARNER

Summary

In general, first term women Marines enter the Marine Corps with relatively

high levels of sexual experience and relatively ineffective contraceptive habits.

These attitudes and behaviors combine with the women’s positive attitudes

toward pregnancy and motherhood to result in approximately one-fourth of

the women getting pregnant during the first year following recruit training.

Notes

*An unplanned pregnancy is defined as a pregnancy that was unintended at the time of conception.

**The women Marines in this study were significantly more sexually active than a comparison sample

of over 300 freshmen women from 3 large universities.** More specifically, only 66% of the college women

were nonvirgins, and the sexually active college women reported an average of 3.9 sexual partners and

engaging in intercourse an average of 6.0 times per month.

***Typical failure rates for specific methods of birth control are the percent of women that would be

expected to get pregnant over the course of one year using the method.”’ These failure rates include both

pregnancies due to method failure and pregnancies due to failure to use the method correctly and consis-

tently.

****Different ns for different analyses are the result of incomplete data.

This research was supported by Office of Naval Research Contract #K85-K-0695 awarded to the first

author. Correspondence regarding this manuscript should be sent to Meg Gerrard, Department of Psy-

chology, Iowa State University, Ames, Iowa 50010.

References

1. Byrne, D. 1983. Sex without contraception. In D. Byrne and W. A. Fisher (Eds.), Adolescents, Sex,

and Contraception (pp. 3-31). New York: Lawrence Erlbaum.

2. Gerrard, M. 1987. Emotional and cognitive barriers of effective contraception: Are males and females

really different? In K. Kelley (Ed.), Females, Males, and Sexuality. New York: McGraw-Hill.

3. Hayes, C. D. (Ed.). 1987. Risking the Future: Adolescent Sexuality, Pregnancy, and Childbearing (Vol.

1). Washington, DC: National Academy Press.

4. Jones, E. F. and Forrest, J. D. 1989. Contraceptive failure in the United States: Revised estimates

from 1982 National Survey of Family Growth. Fam Plann Perspect, 21:103-109.

5. Zelnick, M. and Kantner, J. F. 1980. Sexual activity, contraceptive use and pregnancy among met-

ropolitan-area teenagers: 1971-1979. Fam Plann Perspect, 12:230-237.

6. Miller, W. B. 1973. Psychological vulnerability to unwanted pregnancy. Fam Plann Perspect, 5:199-

201.

7. Adler, N. E. 1981. Sex roles and unwanted pregnancy in adolescent and adult women. Professional

Psychol, 12:56-66.

8. Mosher, D. L. and Cross, H. J. 1979. Sex guilt and premarital sexual experiences of college students.

J Consult Clin Psychol, 36:27-32.

9. Goldfarb, L., Gerrard, M., Gibbons, F. X., and Plante, T. 1988. Attitudes toward sex, arousal, and

the retention of contraceptive information. J Pers Soc Psychol, 55:634-641.

10. Gerrard, M., McCann, L., and Fortini, M. 1983. Prevention of unwanted pregnancy. Am J Community

Psychol, 11:153-168.

11. Geis, B. D. and Gerrard, M. 1984. Predicting male and female contraceptive behavior: A discriminant

analysis of high, moderate, and low contraceptive effectiveness groups. J Pers Soc Psychol, 46:669—

680.

12. Gerrard, M. 1977. Sex guilt in abortion patients. J Consult Clin Psychol, 45:708.

13. Gerrard, M. 1982. Sex, sex guilt, and contraceptive use. J Pers Soc Psychol, 42:153-158.

14. Gerrard, M. 1987. Sex, sex guilt, and contraceptive use revisited: Trends in the 1980s. J Pers Soc

Psychol, 52:975—980.

15. Royle, M. H. 1983. First-term attrition among Marine Corps women: Some associated factors (Technical

Report NPRDC TR 83-22). San Diego, CA: Navy Personnel Research and Development Center.

16. Kerce, E. W. and Royle, M. H. 1984. First-term enlisted Marine Corps women: Their background and

experiences (Technical Report NPRDC TR 84-57). San Diego, CA: Navy Personnel Research and

Development Center.

if.

18.

19.

PAN

2A.

Jef

ZS:

24.

2S:

26.

28.

29.

30.

ANTECEDENTS OF PREGNANCY 15

Gerrard, M. and Royle, M. H. 1985. Predicting pregnancy and pregnancy attrition in first-term Marine

Corps women (Technical Report NPRDC 85-32). San Diego, CA: Navy Personnel Research and

Development Center.

Royle, M. H., Molof, M. J., Winchell, J. D., and Gerrard, M. 1986. Development of a pilot sex

education program for enlisted Marines (Technical Report NPRDC TR 86-9). San Diego, CA: Navy

Personnel Research and Development Center.

Hoiberg, A. 1983. Women in the Navy: Performance, Health, and Motherhood. In F. O. Margiotta,

J. Brown, and M. J. Collins (Eds.), Changing U.S. Military Manpower Realities. Boulder, CO: Westview

Press.

White, L. A., Fisher, W. A., Byrne, D., and Kingma, R. 1977. Development and validation of a

measure of affective orientation to erotic stimuli: The Sexual Opinion Survey. Paper presented at the

meeting of the Midwestern Psychological Association, Chicago, IL.

Fisher, W. A., Byrne, D., White, L. A., and Kelley, K. 1988. Erotophobia-erotophilia as a dimension

of personality. J of Sex Research, 25:123-151.

Fisher, W. A. 1980. Erotophobia-erotophilia and performance in a human sexuality course. Unpub-

lished manuscript, University of Western Ontario.

Fisher, W. A., Bryne, D., Edmunds, M., Miller, C. T., Kelly, K., and White, L. A. 1979. Psychological

and situation-specific correlates of contraceptive behavior among university women. J Sex Research,

15-58-59).

Gerrard, M. and Reis, T. 1989. Retention of contraceptive information in the classroom. J of Sex

Research, 26:315-323.

Gerrard, M., Kurylo, M., and Reis, T. (In press.) Self-esteem, erotophobia & retention of contraceptive

information in the classroom. J of Applied Soc Psychol.

Gerrard, M. 1989. Antecedents of pregnancy and pregnancy attrition in first term women Marines

(Technical Report ONR-89-1). Ames, IA: Iowa State University.

Hatcher, R. A., Guest, F., Stewart, F., Stewart, G. K., Trussell, J., Cerel, S., and Cates, W. 1986.

Contraceptive Technology: 1986-1987. New York: Irvington.

Blum, R. 1987. Contemporary threats to adolescent health in the United States. JAMA, 24:3390-

3395,

Tanfer, K. and Horn, M. C. 1985. Contraceptive use, pregnancy and fertility patterns among single

American women in their 20s. Fam Plann Perspect, 17:10-19.

Navy Times, October 16, 1989, Profile Supplement.

Journal of the Washington Academy of Sciences,

Volume 80, Number 1, Pages 16-25, March 1990

The Discovery of Oscillatory

Electric Current

Albert Gerard Gluckman

Institute for Physical Science and Technology, University of

Maryland, College Park, Maryland 20742

ABSTRACT

The history of the discovery of electric oscillation is reviewed.

I. Introduction

After Oersted’s discovery in 1820, of a magnetic influence from a galvanic

circuit, the two separate phenomena of electricity and of magnetism were tied

together to form the new science of electromagnetism. The next evolutionary

advance taken, was the discovery in 1826 by Felix Savary of Paris, of oscillatory

currents produced by a discharge. His discovery was reasoned from experi-

mental observation of the magnetic periodicity of fine steel needles, when

they became magnetized as a consequence of the discharge current of the

Leyden jar.

During the period of years from 1835 until 1842, Joseph Henry of the College

of New Jersey at Princeton (later to become Princeton University) developed

his theories on the nature of the discharge current. For the purposes of this

paper, only those of his discharge studies that were concerned with his con-

ceptualization of the phenomenon of electric oscillation are considered. Prof.

Henry’s 1842 experiments concerning the oscillatory nature of the discharge

current, were completed with his knowledge of the previous work of Savary,

whom he mentions. But Henry’s experiments were masterpieces of applica-

tions of self-made equipment, not only to prove the existence of oscillatory

currents discharging from the Leyden jar, but also to show the existence of

oscillatory currents for various induction at a distance setups, as well as for

what he called “‘dynamic induction”’ at a distance from lightning.

Karl Wilhelm Knochenhauer postulated a theory in 1842, for the existence

THE DISCOVERY OF OSCILLATORY ELECTRIC CURRENT 17

of electric oscillations in an electrically stressed aether impregnating what we

now call the dielectric medium of the condenser (1.e., capacitor). This electrical

stressing of the aether in the glass, was a consequence, he thought, of the

charging of the Leyden jar. In 1847, Hermann Ludwig Ferdinand Helmholtz

derived the oscillation of electric discharge currents from the principle of the

conservation of energy.

2. Studies by Felix Savary on Magnetic Periodicity, and His Subsequent Deduction

of the Existence of Oscillatory Current

‘Discharge currents produced in the early 19th century, were those from the

Leyden jar, or the magic tableau (tablet) which is also known as the Franklin

square. This is a device consisting of a sheet of glass dielectric that is sand-

wiched in between two sheets of foil, and this may also have been called a

battery. Another source for a discharge current was the cascade or succession

of sparks that were thrown from the prime conductor of a hand cranked friction

machine. The 18th century version of this machine uses a belt to obtain a high

rate of rotation, generally of a glass disk or glass sphere that is held against

a leather “‘rubber’’. This type of machine was invented by Francis Hawksbee

in 1709. Today, such a machine would be called an electrostatic generator or

an electron pump. Later 19th century machines of this type include the Holtz

and the Wimshurst generators. Other than this, the magneto-electric generator

was in use, such as the Saxton machine invented by the American Joseph

Saxton of Philadelphia and London, or the Pixii machine.

I begin with the discharge studies of Felix Savary (born in 1797 and died

in 1841) who was Professor of Astronomy and Geodesy at the Ecole Poly-

technique of Paris. What then was the intellectual source for Prof. Savary’s

investigation of the magnetism that is developed from the discharge circuit,

and how was he led to the distinction between a discharge current and a voltaic

current whose source was the pile of Volta? He and de Monferrand (called

Demonferrand by James Cumming! of the same period, who was President

of the Cambridge Philosophical Society and professor of chemistry at Cam-

bridge University; see ref. 2) had published a study to demonstrate Coulomb’s

laws of force for electric currents in closed voltaic circuits, and a new appli-

cation of the formula of Ampére to represent the mutual action of two infinitely

small portions of the electric current. Self-induction and mutual induction

were but an evolutionary step removed from the study of the mutual actions

of the magnetism of one or two circuits. Because this study by Savary and de

Monferrand is a side issue in this history, it is not placed in the reference

section, and for those readers who find such things of interest, it suffices to

18 ALBERT GERARD GLUCKMAN

say that their study was read at the Academy of Sciences meeting of 3 February

1823, and published in vol. 22 (1823) pp. 259-264, of the Bibliothéque uni-

verselle (LC: Q2.A77).

But these studies by Savary and de Monferrand answer only a part of the

question regarding the intellectual source of Savary’s inquiry into the nature

of the discharge current. The other part of the question seems to be answerable

from Savary’s comments about a new series of experiments that were carried

out by Leopoldo Nobili (born in 1784 and died in Florence in August of 1835).

According to the remarks translated from French’® in Savary’s paper:

[pages 8-9]

‘Since the researches of Mr. Arago, Mr. Nobili has published on some

interesting magnetization experiments. One of these consists in making an

examination, of either the electric discharge, or the current of the pile (Mr.

Nobili has never separated these two means of magnetization), with a plane

spiral of copper wire. If between the spires [which are] insulated one from

the other[,] one fixes, perpendicularly to its plane, some needles of steel,

one finds that the needles situated towards the center and adjacent to the

circumference are magnetized in contrary senses; [and] that by consequence

at a certain distance from the center the magnetization is null.”

We can recognize a number of ideas from this paragraph. Firstly, note that

Nobili at that time failed to distinguish between the current of the discharge

(which today we recognize as a.c.) and the current of the pile (which today

we recognize as d.c.). Secondly, and this is most important, Savary saw in

Nobili’s experiment, the contrary senses of magnetism imparted to the steel

needle depending upon its position with regard to the center of the spiral and

the circumference of the spiral; implying that the magnetism of a needle can

change polarity depending upon where it is situated in distance at various

positions on the wire spiral. This idea can be regarded as the most important

source of these investigations by Savary which led him to his discovery of the

oscillatory nature of the discharge current.

Consider now, his experiments. They are summarized in Table I.

Nobili’s above-mentioned study of the influence of the position of the needle

fixed perpendicularly to the flat coil, on the polarity of magnetism after the

discharge current has passed, I believe, is the germinal ideal in Savary’s re-

search on oscillatory current. The following question and reply were made by

Savary.'° The reply is a statement of his theory which came from his experi-

ments.

[pages 54-56]

“is the movement of electricity during the discharge, composed, to the

contrary, of a suite of oscillations, transmitted from the wire to the sur-

rounding medium, and soon amortisized by the resistances which increase

rapidly with the absolute velocity of the agitated particles?”

THE DISCOVERY OF OSCILLATORY ELECTRIC CURRENT 19

Table 1.—Experiments reported by F. Savary in his 1827 paper

1. Rectilinear wire experiments (using different lengths of platinum wire) to detect periodicity of

magnetism due to the discharge current. (He used needles 5mm, 10mm, and 15mm in length.)

2. Experiments to study the mutual influence of the different parts of the discharge circuit (using a

brass wire 1 meter in length).

3. Experiment to determine if the force of magnetization of a discharge can be modified by resistance.

(The concept of impedance was unknown.)

(a) Savary discussed the production of high temperatures in the platinum wire during the discharge.

(b) He studied whether tempering of the steel needles affected the outcome of the magnetization.

. Experiment to study the influence of the hardness of the needles on their magnetization.

. Experiment to study the influence of the diameter of the needles on their magnetization.

. Experiments to detect periodicity of magnetism on steel needles using a brass wire helix wound on a

hollow wooden cylinder 9cm long, about 6.5mm in diameter. He discussed:

(a) the quantity of electric fluid in a Leyden bottle;

(b) the experiment of Arago in which two helices were used, wound in the same sense and placed

one within the other;

(d) the effects from systems consisting of 3 or 4 helices enclosed one in the other and turning

alternatively in opposite senses.

7. Experiment to magnetize 3 needles placed together in the same helix using the same discharge. The

needles were respectively 5mm, 10mm, and 15mm in length.

8. Experiments in reduction or augmentation of magnetism by the use of copper, silver, and of tin

sheathing, of needles placed in the helix.

Nn &

NOTE. Savary mentioned that in order to eliminate the effect of terrestrial magnetism, one always

places the needles during the discharge in a direction perpendicular to the magnetic meridian.

‘‘All the phenomena conduce this hypothesis, which in fact depends, not

only on intensity, but the sense of magnetism following the laws by which

small movements are amortisized in the wire, in the medium which surrounds

it, [and] in the substance which receives and conserves the magnetization.”

.... An oscillating pendulum in an atmosphere . . . is an example of this

genre of movement.”

So, therefore, Savary’s findings can be summarized for their comparison

with the similar experiments conducted by J. Henry. Thus:

(a) the needles are made to oscillate in time (dynamical phenomenon);

(b) each needle possesses a sense of magnetization whatever is the distance of the

needle to the nearby wire;

of opposite magnetisms that are observed at diverse distances from a conductor;

(d) “‘the electric movement during the discharge is composed . . . of a [train] of

oscillations transmitted from the wire . . . to the surrounding environs [which]

soon dies 4 2f.7 and,

(e) the oscillations have a finite amplitude.

The 1826 studies by Savary on the oscillatory nature of the discharge current

were also analyzed a year later by Gerrit Moll of the Netherlands, and were

mentioned elsewhere at that time, as science news items.!”!8

20 ALBERT GERARD GLUCKMAN

3. The Researches of Joseph Henry Concerning the Oscillatory Nature of the

Leyden Jar Discharge

In 1835, Joseph Henry® propounded a theory concerning his observations

of changes in the sense of magnetization (i.e., changes in polarity) and changes

in the amplitudes of magnetization of steel needles that were exposed to the

action of the discharge current. I think that this theory is but an evolutionary

step to what I call his mature theory of 1842 on the subject of oscillation. A

summary of his findings is made in Table II.

Table I1.—Henry’s 1835 experimental discoveries published in ‘Contributions, No. IT’’

(a) the direction of the magnetic polarity of the needles varies with their distance from the wire

(b) this action of inducing magnetic polarity is periodical

tertiary at a yet greater distance, oppositely directed to the secondary current, etc.

(a) and (b) are the same statements as given by Savary in 1826-7. Hypothesis

of higher order but weaker currents at greater distances from the discharge

wire. This could be considered as a rough equivalence for an alternating

electromagnetic field.

In 1838, Henry published his ‘‘Contributions, No. III’’.’ A synopsis of his

findings about discharge current appears in Table III.

Table III.—Henry’s 1838 experiments results on the magnetism of discharge currents and currents

induced from sparks from the prime conductor of a generator

(a’') the direction of the magnetic polarity due to the secondary current, varies with its distance from

the primary circuit

(b’) the action of inducing magnetic polarity from the secondary circuit is periodical

(c’) the intensity of the induction decreases with increasing distance from the wire

Other than studying magnetic periodicity due to the secondary current, this

1838 study adds nothing essentially new to the 1835 study. In Henry’s 1842

publication”’ ‘Contributions, No. V’’, he explicitly mentions that an electric

discharge is alternating (oscillating), and he proposes the mechanical mech-

anism of a Franklin fluid. Helmholtz on the other hand, in his independent

discovery of the oscillatory character of a discharge (in 1847) which he derived

from the principle of the conservation of energy applied to electricity, did not

attempt to provide any such mechanism, but he did provide a mathematical

foundation for his theory. At that time, the Weber-Fechner theory had just

been proposed in 1845-6, which exerted influence on Helmholtz’s researches

into his application of the concept of the conservation of energy to electricity.

THE DISCOVERY OF OSCILLATORY ELECTRIC CURRENT 21

It was left to William Thomson (later Lord Kelvin) in England, to provide a

mathematical theory to describe electric oscillation in 1853.

In articles 113-134 of his 1838 publication,’ Henry described experiments

demonstrating oscillatory characteristics of the discharge current. From these

experiments he showed 1, the discovery of a difference in the direction of

galvanic (d.c.) currents and ordinary (a.c.) currents of the different orders;

and 2, he conceived the idea that the direction of the currents might depend

on the distance of the conductors. This latter idea is theoretically the same

as that which was proposed by Henry in 1835. Henry noted in article 116:’

“When a discharge from the half gallon jar was passed through one of these

[narrow strips of tinfoil], an induced current in the same direction was ob-

tained from the other. The ribands were then sep[a]rated, by plates of glass,

to the distance of 1/20th of an inch; the current was still in the same direction,

or plus. When the distance was increased to about 1/8th of an inch, no

induced current could be obtained; and when they were still further sep[a]rated

the current again appeared, but was now found to have a different direction,

or to be minus. No other change was observed in the direction of the current;

the intensity of the induction decreased as the ribands were sep[a]rated. The

existence and direction of the current, in this experiment, were determined

by the polarity of the needle in the spiral attached to the ends of the ribands.”’

Art, 134. “. . . the facts here presented . . . appear to be intimately con-

nected with various phenomena, which have been known for some years,

but which have not been referred to any general law of action. Of this class

are the discoveries of Savary, on the alternate magnetism of steel needles,

placed at different distances from the line of a discharge of ordinary elec-

tercrty ce.”

Compare the above statement from 1838 with his statement of Feb. 1835,

found in Henry’s publication® “Contributions ... No. II... .”

“When a current is transmitted through a wire, and a number of small needles

are placed transverse to it, but at different distances, the direction of the

magnetic polarity of the needles varies with their distance from the con-

ducting wire. The action is also periodical; diminishing as the distance in-

creases, until it becomes zero; the polarity of the needles is then inverted,

acquires a maximum, decreases to zero again, and then resumes the first

polarity; several alterations of this kind being observed. Now this is precisely

what would take place if we suppose that the principle current induces a

secondary one in an opposite direction in the air surrounding the conductor,

and this again another in an opposite direction at a great distance, and so

on. The needles at different distances would be acted on by the different

currents, and thus the phenomena described would be produced.”’

Henry seems to have applied to the notion of electric oscillation, a mech-

anistic fluid conception. The notion of electric fluids with hydrodynamical

properties had already been conceived of at an earlier date. Ben Franklin

22 ALBERT GERARD GLUCKMAN

proposed a single electric fluid in contradistinction to a two-fluid theory.

Objections to Franklin’s theory were overcome by an ad hoc hypothesis pro-

vided by Franz Ulrich Theodor (Theodosius) AEpinus in his book Tentamen

theoria electricitatis et magnetismi |‘“‘An Essay on the theory of electricity and

magnetism’’|”’. His book was published in the year 1759 by the Imperial Acad-

emy of St. Petersburg.

It is important to remember that Henry developed this hypothesis in relation

to the dielectric wall of the condenser, and had probably reasoned that the

alternating current flow could explain the open circuit of the conduction cur-

rent interrupted by the non-conducting wall of the condenser. Maxwell had

not yet developed his own notion of the displacement current 0,D which

allowed the closure of the conduction current through the wall of the con-

denser.

Henry’s laboratory notes*’? of June Ist and 2nd of 1842, document the

progression of his magnetization experiments which led to his 1842 publica-

tion.'° The principal new finding reported in this 1842 paper, with regard to

the oscillation of the discharge current, stemmed from ‘“‘a new examination

of the phenomena of the change in direction of the induced currents, with a

change in distance, etc.”’ This went a step beyond Savary’s discoveries, in that

Henry introduced the concept that oscillation can be induced by the process

of induction, to occur in currents of higher orders, however feeble they might

be. This is the basic notion of the working principle of the transformer device.

4. The Researches of Helmholtz and of Knochenhauer on Electric Oscillation

In the year 1847, there appeared in print an extensive treatise by Hermann

Helmholtz,’ a physician and physicist, in which he described his theory of the

conservation of energy. In one section, he described the energy equivalent of

the electrical processes, and one can see how heat energy and electrical energy

have an intimate connection, both being but modes of energy. Thus:

[page 33]

“The energy equivalent of the electrical processes”

‘Riess *) has shown through experiments, that”. . . [in the circuit of the

discharge current, he] ‘‘developed heat proportional to the value Q/S. With

S he designates only the surface area of the coating of the [Leyden] flask

Out of his experiments has Vorsselmann de Heer **) furthermore

followed, as like Knochenhauer***) on his own, that the development of

heat from the same charge of the same battery remains the same .. .”

THE DISCOVERY OF OSCILLATORY ELECTRIC CURRENT 23

“Tt is easy to explain this law, as soon as we imagine to ourselves the

discharging of a battery not as a one way movement of the electricity in one

direction, but as a back and forth fluctuation of itself between the both

coatings, in oscillation, which becomes ever smaller, until the entire kinetic

energy itself is annihilated through the sum of the resistances.”’

*) ““Poged. Ann. XLIII 47.”

**) “Poged. Ann. XLVIII 292. See there the observation of Riess. Especially p.

S20) 7

***) “Ann. LXII 364. LXIV 64.”

On page 32 of his 1842-3 paper,"’ K. W. Knochenhauer mentioned his

conception of electric oscillations in a stressed aether. He was in agreement

with the views of Michael Faraday on the nature of the dielectric, except that

Knochenhauer developed a view in which an electric aether impregnated the

dielectric material of the condenser, as well as the space surrounding it. He

stated that (English translation from German):

“. . . IT will call this the electric stress of the aether. This will arise through

the continued charging, and comes finally to such a degree, that the non-

conductor can not further resist its congestion, and the electric oscillation

of the stressed aether follows. For namely nothing other than as a singular

oscillation, whose kind and manner is yet to be found, do I consider to be

the spark.”

He placed a great emphasis on the study of the electric spark. From the

above remarks, it is evident that he believed an oscillation was transmitted

by means of a stressed electric aether. His concept preceded the researches

of J. C. Maxwell on the electromagnetic theory of light, by 21 years. These

remarks by Knochenhauer on oscillation were made in 1842, the same year

in which Prof. Henry published his mature theory on the oscillation of the

discharge current, on the other side of the Atlantic Ocean. But Prof. Henry

considered these oscillating currents as oscillations of a hydrodynamical elec-

tric fluid. He did not consider the properties of such a fluid, as to whether it

was viscous or ideal, compressible or incompressible; but merely assumed its

existence.

In Europe, a number of researchers began to come to grips with the theory

of electric oscillation in the decade of the 1840s. Amongst the European

researchers, H. W. Dove? mentioned the researches of both Savary and Henry.

And W. G. Hankel’ discussed Savary’s 1826-7 researches with the magneti-

zation of needles. Hankel discussed Ampére’s theory of magnetism, whose

hypothesis concerning magnetism he supported, and he discussed Dove’s 1841

paper in opposition to it. And he also mentioned the work of Marianini,

Henry, and Riess.

ALBERT GERARD GLUCKMAN

Peter Theophil Riess, was familiar with the researches of Henry.'*!> How-

ever, in 1840, he did not accept the theory of electric oscillation, in this, the

early part of his career. This can be seen from his remarks, which I have

translated from his 19th latinic German.

‘““How impermissible is the conclusion of an anomalous magnetization being

dependent on the change in the current direction, relative to Savary’s ex-

periments, that it depends on the mass and the hardness of the magnetized

needle, one must therefore declare about the secondary current, that it is at

this or every distance [from the primary] changing its direction, according

as the one or the other needle itself is employed in the proof.”

This shows that Savary’s theory of electric oscillation was not universally

accepted by 1840.

10.

References

. Cumming, James. ‘““On the Magnetising of Needles by Currents and Electric Sparks”, Edinburgh

Journal of Science, vol. 5 (April-October, 1826) page 369 only. This journal was later incorporated

into the London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science.

. Demonferrand, Jean Baptiste F. A manual of electro dynamics, chiefly translated from the Manuel

d’électricité dynamique or Treatise on the mutual action of electric conductors and magnets, of J. F-

Demonferrand with notes and additions, comprehending the latest discoveries and improvements (trans-

lated by James Cumming), J. & J. J. Deighton, Cambridge, and C. & J. Rivington, London, 1827

(see page 247 for his discussion of Savary’s discovery of electric oscillation).

. Dove, Heinrich Wilhelm. ‘Recherches sur les courants d’induction dus a l’aimantation du fer par

l’électricité ordinaire”’ [‘“‘Researches on currents of induction due to the magnetization of iron by

ordinary [a.c.] electricity”], Annales de chimie et de physique, 3rd series, vol. 4 (1842), pp. 336-358.

. Hankel, Wilhelm Gottleib. ““Ueber die Magnetisirung von StahInadeln durch den elektrischen Funken

und den Nebenstrom desselben” [‘‘On the magnetization of steel needles through electric sparks and

the same from the secondary current’’], Pogg. Ann., vol. 65 (1845), pp. 537-568.

. Helmholtz, Hermann Ludwig Ferdinand. ‘Ueber die Erhaltung der Kraft” [‘‘On the conservation of

energy, a physical dissertation lectured on in the meeting of the physical society of Berlin on the 23rd

July 1847.”]; refer to ““Ostwald’s Klassiker der Exakten Wissenschaften. Nr. 1.”, Printing house of

Wilhelm Engelmann, Leipzig 1902 (Berlin Press and printing house of G. Reimer, 1847.). See also

the translation by J. Tyndall, and published in his ‘Scientific Memoirs’’, London (1853) vol. 1, page

143.

. Henry, Joseph. “Contributions to Electricity and Magnetism, No. II, ‘On the Influence of a Spiral

Conductor in Increasing the Intensity of Electricity from a Galvanic Arrangement of a Single Pair,

etc.’ ’’, Transactions of the American Philosophical Society, new series, Vol. 5 (read February 6, 1835)

pp. 223-231. Also found in: (a) Scientific Writings, Vol. I, page 92; and (b) “‘The discovery of induced

electric currents’’, vol. 1, Memoirs by Joseph Henry, edited by J. S. Ames, American Book Company,

New York, 1900.

. Henry, Joseph. “Contributions to Electricity and Magnetism. No. III. ‘On Electro-dynamic Induc-

tion.’ ’, Transactions of the American Philosophical Society, vol. 6 (read Nov. 2, 1838), pp. 303-338.

See also: (a) Silliman’s American Journal of Science, vol. 38 (Jan 1840), pp. 209-243; (b) Sturgeon’s

Annals of Electricity, etc., vol. 4, pages 281-310; (c) London, Edinburgh, and Dublin Philosophical

Magazine, vol. 16 (March 1840), pp. 200-210; pp. 254-265; pp. 551-562; (d) Becquerel’s Traité

expérimental de I’ Electricité, etc., 1837, vol. v, pp. 87-107; (e) Annales de Chimie et de Physique, 3rd

series, vol. 3 (Dec. 1841), pp. 394-407; and (f) Poggendorff’s Annalen der Physik und Chemie,

Supplemental vol. 1, Following Volume 51 (1842), pp. 282-312.

. Henry, Joseph. Laboratory Notebook (1 June 1842), p. 272. Smithsonian Archives, Washington,

Henry, Joseph. Laboratory Notebook (2 June 1842), p. 275. Smithsonian Archives, Washington,

DC.

Henry, Joseph. ‘Contributions to electricity and magnetism. No. V. On Induction from Ordinary

Electricity; and on the Oscillatory Discharge’, Proceedings of the American Philosophical Society,

ft.

12.

13.

14.

iS.

16.

18.

vol. 2 (June 17, 1842), pp. 193-196. Also found in: (a) Scientific Writings, Vol. I, page 200; and (b)

Memoirs by Joseph Henry, edited by J. S. Ames, American Book Company, New York, 1900.

Knochenhauer, Karl Wilhelm. ““Versuche tiber gebundene Elektricitat’’ [‘“‘Experiments concerning

bound electricity’’], Pogg. Ann., vol. 58 (1843), pp. 31-49 (Completed 1 Sept. 1842); pp. 211-231

(Completed 1 Sept. 1842); and pp. 391-409 (Completed February 1843).

Moll, Gerrit. ‘““Eenige electro-magnetische Proeven’’, Bijdragen tot de natuurkindige wetenschappen,

Amsterdam, vol. 2 (1827), pp. 372-375. About Savary’s oscillatory discharge experiments.

Riess, Peter Theophil. ““Fortgesetzte Untersuchungen tiber den Nebenstrom der elektrischen Batterie”

(‘Continued investigations concerning the secondary current of the electric battery’’], Pogg. Ann.,

vol. 50 (1840), pp. 1-24.

Riess, Peter Theophil. Repertorium der Physik, volumes 5 & 6, Berlin (1842); ‘““Zweiter Abschnitt”’,

“Die Lehre von der Elektricitat” [“‘Lessons in Electricity’’]; see especially page 164.

Riess, Peter Theophil. “Ueber die Influenzelektricitat und die Theorie des Condensators’’, Poggen-

dorff’s Annalen, vol. 73 (1848), pp. 367-405.

Savary, Felix. “Mémoire sur |’Aimantation” [Memoir on Magnetization’’], Annales de chimie et de

physique, vol. 34 (1827), pp. 5-57, “Read at the Academy of Sciences the Sth of July 1826’’; and pp.

220-221 entitled: “ADDITION au Memoire de M[onsieur]. Savary sur l’Aimantation”’.

‘“‘New Facts in Electro-Magnetism”’, The Quarterly Journal of Science, Literature, and Art, vol. 22

(1826), pp. 383-384. Reviewer’s comments on F. Savary’s experiments on oscillatory discharge current.

‘Nouveaux faits d’électro-magnétisme’’, Le Globe (August 2, 1826), pp. 511-512. Copy in Smithsonian

Archives; original in the National Library of France. This is a synopsis of Savary’s results of experiments

on oscillatory discharge current.

Journal of the Washington Academy of Sciences,

Volume 80, Number 1, Pages 25-36, March 1990

Identification of a Superforce in

the Einstein Field Equations

Dr. Robert J. Heaston

IIT Research Institute, Chicago, IL 60616

ABSTRACT

The Einstein field equations provide the underlying principles to theories of gravi-

tation, the big bang, black holes, and cosmology in general. Many variants of these

equations have been developed by Einstein and subsequent investigators. These variants

include differences in mathematical form, components, arithmetic signs, and the presence

of particular constants, and corresponding mathematical solutions. The objective of this

paper is to examine the variants of the Einstein field equations where the combination

of fundamental constants c*/G occurs. This combination of the speed of light, c, and the

universal gravitational constant, G, has the units of force. Significant relationships of this

force to the Planck mass, Planck length, cosmic numbers, color force between quarks,

and the Einstein field equations are derived and discussed. The characteristics of c’/G

fulfill predictions for the superforce.

26 ROBERT J. HEASTON

Introduction

The Einstein field equations are the starting point for theories of the big

bang, black holes, superstrings, and cosmology in general. A brief description

will be given of the Einstein field equations. Then, it will be shown how an

extremely strong force, that fulfills the predictions of a superforce, is contained

in the Einstein field equations. This force has been overlooked since nothing

is added to or taken away from the Einstein field equations. The properties

of the superforce will be explained. Important quantities which are contained

in various cosmological theories will be related to the superforce. The resulting

relationships will be used to suggest a different approach to the big bang.

Two points will be emphasized over and over again. One point is that the

Einstein field equations will be taken just as Einstein proposed them. The

other point is that the results presented here fulfill predictions made by several

physicists. Starting with the same equations analyzed by others, the outcome

is predictable, but not exactly in the way that would be normally expected.

The Einstein Field Equations and Their Interpretation

Albert Einstein published! an extensive description of his general theory of

relativity in 1916. Over the next few years, he published other papers that

presented variants on this theory. Subsequent to Einstein, others also pub-

lished additional variations on the general theory of relativity, all based upon

the same formulations by Einstein. There have been so many different vari-

ations that C. Misner’ and others list 37 accepted forms of the Einstein field

equations that define the foundation to the general theory of relativity that

represents Our most advanced understanding of the theory of gravitation.

Einstein? proposed that the formulation of one of his field equations of

gravitation be stated as follows:

TE eis ual on RY) (1)

He defined the terms in this equation in another publication.* In equation (1),

denotes the contracted Riemann tensor of curvature,

R_ represents the scalar of curvature formed by repeated contraction,

is the energy-tensor of “‘matter’’,

g., 1S the fundamental tensor, and

K 1S a constant.

Einstein loosely used « as a constant that had several different magnitudes

and units. It is defined in inverted form on page 160 of the 1917 publication

IDENTIFICATION OF A SUPERFORCE IN THE EINSTEIN FIELD EQUATIONS 27

ask! = 8 m G/c’, where G is Newton’s universal constant of gravitation and

c is the speed of light.

At the time that Einstein proposed equation (1), physical observations

indicated that the universe was spatially finite. On the other hand, equation

(1) predicts an expanding universe. To correct this problem, he decided that

‘‘on the left-hand side of [the] field equation . . . we may add the fundamental

tensor g,,, multiplied by a universal constant, —),”

so that

tia Nei ei KC Regyee hy Ze) (2)

The constant, \, has since come to be called the “‘cosmological constant.”

The insertion of the cosmological constant assures that equation (2) is a model

of a spatially finite universe.

Let us rearrange equation (2), to put it in one of the more currently ex-

pressed forms

Rigiee Wi 2sER bo/\p =e kil 3 (3)

In this case /\ = —)/k and k = —1/k. Equation (3), with and without the

cosmological constant, will be the focus of the remainder of this paper.

It is important to note the dimensional units used in equation (3). Each of

the groups of components are in units of cm~*. The reason for these units is

that the Einstein field equations are only considered valid for a small volume

of space. The actual units, in cgs form, of each group of components is cm/

cm*, or length per unit volume. Specifically, R,,, R, and /\ are in these units.

The fundamental tensor is dimensionless. The right side of equation (3) has

a more complicated structure. In fact, Einstein’ is reputed to have said that

the components of the right hand side were a blemish in his theory because

they are essentially non-geometrical entities.

Different dimensional units may be used to define T,,, and k. When T,,, is

expressed as energy per unit volume, k is |/force. When T,,, is a mass per unit

volume, or density, k has units of cm/gr. Should T,,, be a momentum per unit

volume, k assumes sec/gr units.

The units of k in equation (3) are particularly important for the case when

T,,, IS expressed as an energy per unit volume. Under these conditions, k =

8 m7 G/c*. It would appear from the various solutions of equation (3) that the

8 m and the G/c* components are from different sources. The combination of

fundamental constants G/c*, or rather the inverse, c*/G, is of particular in-

terest. The speed of light to the fourth power divided by Newton’s universal

gravitational constant, c‘/G, has the units of force. The magnitude of this

28 ROBERT J. HEASTON

force, referred to as Fs, is enormous: Fs = 1.2 x 10” dynes. If F; existed in

nature, it would be superstrong. Compelling reasons dictate that Fs be called

the superforce. Attention will be focused on this superforce and its unusual

characteristics for the remainder of this paper.

Superforce

P. Davies published® a book in 1984 called “‘Superforce: The Search for a

Grand Unified Theory of Nature.’’ He makes the following statements in this

book:

. . Investigations point toward a compelling idea, that all nature is ulti-

mately controlled by the activities of a single superforce. . . .The search for

a superforce can be traced to the early work of Einstein and others, .. .”

The major premise of this paper is that c*/G is this superforce. It is contained

in the Einstein field equations and was overlooked by him and everyone else.

The probable reasons for this oversight will be explained. Several derivations

will be performed to justify calling c*/G the superforce.

Combinations of fundamental constants’ which repeatedly turn up in cos-

mological theories are the:

Planck mass, m, = (hc/27G)!”” (4)

Planck length, r, = (hG/2mc*)'”? (5)

Planck time, t, = (hG/2mc°)!” (6)

The Planck mass is considered to be the largest possible particle mass, m, =

2.2 x 10~° gr. The Planck length, r, = 1.6 x 10~* cm, is a measure of the

failure of the vacuum and may be the minimum size black hole. The Planck

time, t, = 5.4 x 10°-“'sec, is the time it takes light to travel the Planck length.

Many theories of the origin of the universe go back to the Planck time. The

Planck length is involved in more theories than the other Planck terms. Ac-

cording to C. Isham, it is a

‘fundamental length in nature. .. , and it is at this scale that we might

expect to see some effects of quantum gravity. . . . it has long been assumed

that something rather odd will happen as . .. matter passes through the

Planck length... .”

The Planck length, mass, and time are all associated through mathematical

approximations’ to several measurements of the universe. These measure-

IDENTIFICATION OF A SUPERFORCE IN THE EINSTEIN FIELD EQUATIONS — 29

ments include: the solar mass; main-sequence lifetime of stars; the Hawking

evaporation rate for black holes; maximum allowable size for a planet; and,

age, mass, and density of the universe.

A straightforward explanation may be given as to why the Planck functions

turn up so frequently in astrophysical and cosmological theories. If two Planck

masses are assumed to interact over a range equal to the Planck length, the

resulting Newtonian gravitational force is equal to the superforce.

m, °G/r, = c*/G (7)

It may be inferred from this interaction between two boundary limit relation-

ships such as the Planck mass and the Planck length that F; represents a limit

of some sort. The superforce should also be related to the same phenomena

that involves the Planck mass and the Planck length. Thus, according to

Wagoner and Goldsmith,’

“Therefore, for now we must regard the Planck barrier as another effective

limit to our universe.”’

We cannot make calculations earlier than the Planck era. The reason why the

Planck mass, Planck length and Planck time occur so frequently in cosmo-

logical theories is that they are indirectly contained in the Einstein field equa-

tions. Based upon equation (7), k in equation (3) is equal to 8 7 r}/m, °G in

energy units and 8 7 t;/m, °G in mass units. The Planck conditions are easily

introduced in this manner into the Einstein field equations.

There is a further link between the superforce and the Einstein field equa-

tions. A weak-field solution of the Einstein field equations leads to the pre-

diction of gravitational waves. A rotating mass has a time-varying quadrapole

moment that generates gravitational radiation with a gravitational luminosity,

L. According to Douglas and Braginsky,"°

. an upper limit on the luminosity of an astronomical source can be

estimated...”

be GlG =-4 x,10?).eras/sec (8)

The limit to gravitational luminosity is the superforce times the speed of light,

Fy - c, which is equivalent to a maximum energy flow per unit time.

The relationships indicated by equations (7) and (8) have associated the

superforce with limiting conditions for physical phenomena. The inference is

that the superforce may also represent a limit. It will be assumed, for the sake

of discussion, that F, is the maximum possible force in the universe. If Fs

30 ROBERT J. HEASTON

were the maximum possible force and m, were the maximum possible particle

mass, there should be a maximum acceleration, called the Planck acceleration,

ap, to go with the Planck mass, which 1s

a, = F;/m, = (27c’/hG)!”” (9)

The magnitude of a,, which is composed of fundamental constants, is 5.56 x

10°° cm/sec’. The Planck acceleration is defined here for the first time.

One of the most often cited solutions to the Einstein field equations is the

Schwartzschild limit, r, = 2mG/c’. This limit prescribes the event horizon of

a non-rotating black hole where the gravitational strength of a collapsing body

of matter is so strong that not even light can escape. There may be some

exception” due to quantum effects which leads to Hawking radiation or Hawk-

ing evaporation. Nevertheless, there is a probable boundary to black holes

defined by the Schwartzschild limit. There is a similar Newtonian gravitational

collapse limit as a result of the superforce,

m’G/ry = c*/G (10)

ty= mG/c? (11)

By comparison, 2ry = 1,. The gravitational collapse limit based upon New-

tonian gravitation is half that of Einsteinian gravitation. This difference could

be due to different mechanisms, the geometrical shape of the universe used

in the Einstein field equations, or to some unknown reason. The similarities

between ry and r, demand an explanation.

Before rejecting ry in favor of r,, it should be noted that the following

relationship,

n = MG/Re? (12)

where M and R are the mass and radius of astrophysical bodies, is referred

to by C. Will’ as a

‘‘characteristic measure of the size of relativistic effects in bodies.”

The value of n for the sun is 10~°; for a white dwarf, it is 10°°; and, it

approaches 0.3 for a neutron star. There are no observations for n > | so

that ry may indeed have a physical significance.

If the Newtonian gravitational collapse limit defined by equation (11) is

multiplied by c? on both sides and rearranged, the result is

ES) aR (13)

The relationship given by equation (13) would imply that ry is indeed the

gravitational collapse limit of matter. In other words, equations (11) and (13)

IDENTIFICATION OF A SUPERFORCE IN THE EINSTEIN FIELD EQUATIONS _ 31

indicate that in a plot of the Newtonian gravitational force versus distance,

the area under the curve at ry and F; is equal to the rest mass energy of the

interacting entities. The inference from this observation is that the superforce,

F;, confines matter/energy in black holes.

Since the superforce is composed of two fundamental constants, its mag-

nitude would be effected by any changes in the fundamental constants. It

would require relatively large changes in the speed of light or the gravitational

constant to produce a significant change in the superforce. The evolution of

the universe would not have proceeded the way it has if there had been any

major variations in the fundamental constants.'’ Consequently, it may be

assumed that the magnitude of the superforce has remained constant over the

age of the universe.

Another interesting set of numbers can be generated with the help of the

superforce. The Coulomb force may be used to derive a series of functions

similar to the Planck functions. By definition, the Coulomb force is equal to

the superforce at a distance called the Coulomb length,

Rete Gre ye Ae | x 1078 em (14)

The time it takes for light to travel this distance is the Coulomb time,

tk = (e?G/c5) = 4.6 x 10-“ sec (15)

The Coulomb mass is calculated when the gravitational collapse limit is equal

to the Coulomb length,

Meher 9 x 1s ® or (16)

The Coulomb acceleration is calculated from Fs = mxax, or

eel EC) Chan) 1-6. 5 1X1 OP? emi sec (17)

Some of the Coulomb functions have been derived before, but not from the

superforce. It is possible to derive each of these same Coulomb functions by

multiplying the corresponding Planck function by the square root of the fine

structure constant.

Cosmic Numbers

The role of the superforce in explaining various cosmic numbers further

justifies the assertion that F, is the largest possible force. Cosmic numbers

are very large numbers that recur in cosmological theories for no known

32 ROBERT J. HEASTON

physical reason. One of these cosmic numbers is the Eddington number",

Ne = hc?/4a?m‘G? (18)

Sir Arthur Eddington was an outstanding astronomer who was a contemporary

of Einstein. Eddington was one of the earliest proponents of the special and

general theories of relativity. In the year 1919, he led the expedition” to

Principe Isle in the Atlantic Ocean south of Nigeria to measure the Einstein

prediction of the gravitational bending of light rays. As a result of a derivation

based upon the Einstein field equations, Eddington derived N; and noted that

Ne ~ 10 for the mass of one of the mesons. Eddington thought that this

number, in itself, was significant and even assumed that there were 10* protons

and 10* electrons in the universe. The resulting mass was close to physical

observations of the day. Eddington later changed his mind’® about equation

(18) when he said it was equal to 2 x 136 x 2”° and called this result the

‘“‘cosmical number’’.

Recognition of the superforce, however, finally does allow a physical inter-

pretation for the Eddington number as defined by equation (18). Consider

the relative magnitude of the superforce, F;, to that of the Newtonian grav-

itational force, F,, at the Compton wavelength,

F./Fg = (he/2am?G)? = Ny (19)

In essence, the Eddington number is the relative strength of the largest force

in the universe to the weakest force.

Another of the cosmic numbers, N,, is given by

N, = (N,)!? = he/2am’G (20)

The inverse of this cosmic number also corresponds to a representation of the

gravitational coupling constant. In this case, the color force between quarks,

F., could be defined in the following way relative to the superforce,

F. = F,/N, = 2nm°’c3/h (21)

The color force would have a theoretical magnitude of 7.1 x 10° dynes for

the mass of a proton. It just so happens that this is precisely the measured”

magnitude of the color force at the Compton wavelength of a proton. The

color force between quarks is bracketed between the superforce and the grav-

itational force. The superforce is 1.7 x 10°* times the color force between

quarks. This should give validity to calling c*/G the superforce.

Why did Einstein, and others, overlook the superforce in the Einstein field

equations? There are three probable reasons. One of the reasons is the em-

IDENTIFICATION OF A SUPERFORCE IN THE EINSTEIN FIELD EQUATIONS — 33

phasis on spacetime and the curvature of space in equations (1), (2), and (3).

Another reason has to do with the various forms of k and k. The most likely

reason is because of the assumption made by Einstein and almost every sub-

sequent researcher using the Einstein field equations. Einstein assumed that

the fundamental constants c and G were equal to one; c = G = 1. This

assumption equates 1.2 x 10” to 1 and wipes out the visible presence of the

superforce.

Cosmological Model

If there were a superforce as predicted, and this force had a finite magnitude

which could very well be the maximum possible force in the universe, then,

the superforce could significantly impact theories about the origin of the uni-

verse. A particular cosmological model is suggested from the characteristics

of the superforce.

There are two predominant theories of the origin of the universe, both

based on the Einstein field equations. One approach is called the “standard

big bang.” In the beginning, all of the matter-energy of the universe was

collapsed into a dimensionless point at infinite density and infinite force. This

point at zero time is called a singularity because all of the laws of nature are

unique at this point since the laws must be applicable to infinite conditions

rather than the finite characteristics of the observable universe. Once the

expansion of the big bang began, the normal laws of physics prevailed. Pre-

dictions for the singularity have been made back to the Planck time, or so-

called Planck era. Many of the characteristics of the current universe can be

predicted but there are difficulties with the smoothness and flatness problems.

To correct for the shortcomings of the standard big bang model, A. Guth’®

and others proposed an “‘inflationary big bang.”’ In this model, the expansion

of the universe takes place in two stages. The first stage is an extremely rapid

expansion from a singularity until the Planck era. Then, the expansion pro-

ceeds the same as the big bang. This approach resolves many of the problems

encountered with the standard big bang, but the inflationary big bang has its

own problems. The primary problem is trying to explain events during the

inflationary period up to the Planck time. It is necessary to define a false

vacuum with negative pressure. There is also a vacuum energy density asso-

ciated with the false vacuum. It is suggested that the reason that it is necessary

to define a false vacuum is to hurdle the threshold posed by the superforce.

In fact, the superforce enters into the definition of the vacuum energy density,

Uyac, and the pressure of the vacuum, Pyac:

C= Ua = hel 8 WO (22)

34 ROBERT J. HEASTON

Equation (22) essentially nulls out the inflationary period at the superforce.

A third alternative to the standard big bang and the inflationary big bang

is possible with identification of the superforce. The model is called the “‘finite

big bang.” All of the original matter-energy of the universe contracted to a

finite gravitational collapse limit confined by the superforce. A valid solution

for the Einstein field equations at the beginning of time is

Roo x CVG = 4tT (23)

In this case, all of the mass-energy of the universe, 47T\y, was contained in

a region, Ro, and held together by the superforce, c*/G. There is no need to

resort to a singularity and all of the accompanying infinities. The universe

began from a finite ball of mass-energy. At about 10~* sec, the family of

fermions was created with decay of the Coulomb masses in the original struc-

ture. At about 10~** sec, the family of bosons was generated from decay of

the pre-existent Planck masses. After 107! sec, the superforce decayed and

the fundamental forces took control. The smoothness and flatness problems

are particularly resolved by this approach.

Conclusions

The superforce was first identified by the author in 1976. Subsequent to the

original derivation, several presentations were made and papers prepared”

which included the superforce but did not emphasize it. This is the first such

paper that addresses the superforce by itself. Based upon experience, it has

been observed that there is a general reticence in endowing c*/G with any

significance. This reaction was understandable initially, but more and more

evidence has been accumulated on the hidden role of c*/G and its impact on

the theoretical understanding of the universe.

Observations about c*/G may be grouped into two categories: those which

are mathematically and observationally correct and those that are more spec-

ulative. From a mathematical and dimensional perspective, there is no doubt

that c’/G has the units of force. This force has an enormous magnitude that

is 1.7 x 10°8 times stronger than the measured color force between quarks

and the strong interaction between nucleons. Moreover, there is no doubt

that c’/G appears in inverse form in versions of the Einstein field equations.

When Newton’s universal law of gravitation is expressed in terms of the Planck

mass and the Planck length, it is equal to c*/G. Thus, both Newtonian and

Planck conditions are present as an identity in the Einstein field equations.

IDENTIFICATION OF A SUPERFORCE IN THE EINSTEIN FIELD EQUATIONS = 35

A Planck acceleration may be defined using Newton’s force law, the Planck

mass, and c*/G. Various combinations of fundamental constants may be de-

fined using the Coulomb force and c*/G that are related through the square

root of the fine structure constant and the Planck mass, length, time, and

acceleration, respectively. A physical explanation of the Eddington number,

as well as other cosmic numbers, are possible with c*/G. The gravitational

coupling constant can be derived from this understanding of the Eddington

number. The maximum possible gravitational luminosity is defined in terms

of c*/G. The false vacuum pressure and energy density are also based upon

c’/G. All of the statements about c*/G in this paragraph may be backed up

by mathematical expressions. Moreover, these comments are all valid without

introducing the concept that c*/G may be the superforce. The intent has been

to identify those relationships where c*/G plays a role either by its presence

or as a bridge between other functions. It may be concluded that c*/G is more

than a fortuitous combination of the fundamental constants because of its

ubiquitous roles in theoretical and observational physics.

From a speculative and judgmental viewpoint, it would appear that c*/G

deserves to be called the superforce. It is far stronger than any of the four

fundamental forces and could satisfy the conditions for convergence of the

four fundamental forces. Almost all solutions to the Einstein field equations

could be interpreted differently based upon the presence of the superforce.

The meaning of the cosmological constant, the conditions of the big bang, the

structure of black holes, the unification of the four fundamental forces, and

the concepts of strings and superstrings must all account for the role of the

Superforce.

The superforce is a combination of fundamental constants that surely must

represent some boundary condition in nature. Apparently, there is more than

One maximum boundary condition: maximum velocity, c; maximum particle

mass, m,; maximum gravitational luminosity, L; maximum force, Fs; and,

maximum acceleration, a,. Because of these maximum conditions, there may

also be a minimum gravitational collapse limit. These boundary conditions

suggest that there is no singularity as currently defined. The superforce pro-

vides an additional justification for what R. Matzner and his co-authors” state

many physicists already believe,

) true singularities do mot exist, . .7-7

The big bang started from finite dimensions. Black holes terminate in finite

dimensions. Nature is finite. Only the supernatural is infinite. All of these

conditions are fulfilled by the superforce, c*/G, which has always been present

in the Einstein field equations.

36 ROBERT J. HEASTON

Correspondence may be addressed to the author at the above address. This paper was presented on

October 4, 1989, as a colloquium talk to the Physics Department of the Illinois Institute of Technology,

Chicago, Illinois.

References

1. Einstein, A. 1923. The Foundation of the General Theory of Relativity. Published in 1916 and reprinted

in The Principle of Relativity. New York: Dover Publications, Inc., pp. 109-164.

2. Misner, C. W., Thorne, K. S., and Wheeler, J. A. 1973. Gravitation. San Francisco: W. H. Freeman

and Company, Cover overleaf.

3. Einstein, A. 1923. Cosmological Considerations on the General Theory of Relativity. Published in

1917 and reprinted in The Principle of Relativity. New York: Dover Publications, Inc., p. 186. The

equations are cited as expressed but current terminology has been substituted.

4. Einstein, A. 1923. Do Gravitational Fields Play an Essential Part in the Structure of the Elementary

Particles of Matter? Published in 1919 and reprinted in The Principle of Relativity. New York: Dover

Publications, Inc., p. 192.

5. Gibbons, G. W. 1979. Quantum Field Theory in Curved Spacetime. General Relativity: An Einstein

Centenary Survey, Edited by S. W. Hawking and W. Israel, Cambridge: Cambridge University Press,

p. 639.

6. Davies, P. 1984. Superforce: The Search for a Grand Unified Theory of Nature. New York: Simon and

Schuster, pp. 5-6.

7. Wagoner, R. V. and Goldsmith, D. W. 1983. Cosmic Horizons: Understanding the Universe. San

Francisco: W. H. Freeman and Company, pp. 93, 147, and 159.

8. Isham, C. 1989. Quantum Gravity. The New Physics, Edited by Paul Davies. New York: Cambridge

University Press, pp. 70-71.

9. Carr, B. and Rothman, T. 1985. Coincidences in Nature and the Hunt for the Anthropic Principle.

Frontiers of Modern Physics, Tony Rothman, Senior Author. New York: Dover Publications, Inc. pp.

107-130.

10. Douglas, D. H. and Braginsky, V. B. 1979. Gravitational-Radiation Experiments. General Relativity:

An Einstein Centenary Survey, Edited by S. W. Hawking and W. Israel. Cambridge: Cambridge

University Press, p. 92.

11. Hawking, S. W. 1985. A Brief History of Time: From the Big Bang to Black Holes. New York: Bantam

Books, pp. 104-110.

12. Will, C. 1989. The Renaissance of General Relativity. The New Physics, Edited by Paul Davies. New

York: Cambridge University Press, pp. 23-24.

13. Rozental, I. L. 1987. Big Bang, Big Bounce: How Particles and Fields Drive Cosmic Evolution. New

York: Springer-Verlag, p. v.

14. Eddington, A. 1936. Relativity Theory of Protons and Electrons, New York: Macmillan, p. 272.

15. Bernstein, J. 1980. Einstein. New York: Penguin Books, pp. 141-146.

16. Newman, J. R., Editor. 1956. Commentary on Sir Arthur Stanley Eddington. The World of Mathe-

matics, Vol. 2. New York: Simon and Schuster, p. 1069.

17. Bloom, E. D. and Feldman, G. J. May 1982. Quarkonium, Scientific American, p. 42.

18. Guth, A. and Steinhardt, P. 1989. The Inflationary Universe. The New Physics, Edited by Paul Davies.

New York: Cambridge University Press, p. 58.

19. Heaston, R. J. (a) February 1977, Speculations on a Unified Model of the Four Fundamental Forces,

AAAS Annual Meeting; (b) 1978, Unified Interaction Model of the Four Fundamental Forces, Spec-

ulations in Science and Technology, Vol. 1, No. 1, 71-75; (c) June 1980, Redefinition of the Four

Fundamental Forces, Proceedings of Army Science Conference, 203-217; (d) January 1982, Generic

Field Theory, AAAS Poster Session; (e) April 1982, A New Look at the Concept of Force, Physics

Seminar, University of Mississippi; (f) 1983, A New Look at the Concept of Force, Speculations in

Science and Technology, Vol. 6. No. 5, 485-497; (g) April 1985, Pathways to a Unified Field Theory,

Seminar, Naval Weapons Center; (h) February 1986, Super Unification, Seminar, University of Mis-

sissippi; (i) October 1989, Identification of a Superforce in the Einstein Field Equations, Colloquium,

Illinois Institute of Technology.

20. Matzner, R., Piran, T. S., and Rothman, T. 1985. Demythologizing the Black Hole. Frontiers of

Modern Physics, Tony Rothman, Senior Author, New York: Dover Publications, Inc. p. 46.

DELEGATES TO THE WASHINGTON ACADEMY OF SCIENCES,

REPRESENTING THE LOCAL AFFILIATED SOCIETIES

PaimsOpiical SOCiety Of WaSMINGtON..... 0.6.05... .c0.. ene ede ee gee dae e nes ene eel James F. Goff

ParanopoOlerical Socicty Of Washington. ..0......3..000. 5.56 eeadcteeesese- Edward J. Lehman

EI IC AL SOCICLY (OL WaASMEMPTON 6.) ccs ce beetle babe ch aaeeac evans tole ee Austin B. Williams

eEMed SOCICLY Of WaSliNPGON:. 62... jcc. ee ea ee os aeeecse dene be's Jo-Anne A. Jackson

Paromological Society of Washington ...............2. esc ee eee eee eeneees Manya B. Stoetzel

ene eC OPTADING SOCICLY e.g Yate ee clle load Waa dds cae aceeauees Stanley G. Leftwich

Bemried) SOCcty OF WashINStOM 2.65. ee ee eae wd enee eee James V. O’Connor

wee socicty of ine District-of Columbia. ... 2.2... 00... cbc ccs eet eneedneeniis John P. Utz

PRM MEte EN IS(ONICAWSOCICLY) 4. 22 se eek. oe ek obs Salen Gace ede Ube andes Paul H. Oehser

Seeiemecah SOCIctY OF WaSHINGION . 2.0... 6... ae oh ene debe whe dens Conrad B. Link

_ Society of American Foresters, Washington Section ................... Forrest Fenstermaker

SeEMemeRNT SOCICLY (OU EMPINICELS 255 sos je ie cadens aces veweesdesbvcweaneeees Alvin Reiner

Institute of Electrical and Electronics Engineers, Washington Section...... George Abraham

American Society of Mechanical Engineers, Washington Section................ Michael Chi

Piomuamiological Society of Washington...............0. 2.0.0.0 c cece ewes Kendall G. Powers

American Society for Microbiology, Washington Branch................... To be determined

Society of American Military Engineers, Washington Post............. Charles A. Burroughs

American Society of Civil Engineers, National Capital Section.......... Herbert A. Pennock

Society for Experimental Biology and Medicine, DC Section ............ Cyrus R. Creveling

American Society for Metals, Washington Chapter .....................08. Pamela S. Patrick

American Association of Dental Research, Washington Section............ J. Terrell Hoffeld

American Institute of Aeronautics and Astronautics, National Capital

SDD oo eosin agit cee obra Sele Siepio lias cide ety ete se nr ae Reginald C. Smith

ramie mean Wietcorological Society, DC Chapter...............0.0.5..0-204. A. James Wagner

Poericnes society Of Washington, ... 05066. cee eee cee ee ede eee Ralph Webb

meoustical Society of America, Washington Chapter....................... K. Richard Cook

Pidenead Nucicar Society, Washington Section.................2..50..ceneeecees Kamal Araj

ineite ot Food Technologists, Washington Section....:........2....-..2.+05- Elvira L. Paz

American Ceramic Society, Baltimore-Washington Section .............. Joseph H. Simmons

PeeeerMMNCMIMC AID SOCICUN 2 Sheers 1A. anan cs dav accede bode taessseevencnaee Alayne W. Adams

Pam ston Iaistory.or Science Chub ys). c2 ey... coe ce eee ee eee eee Albert G. Gluckman

American Association of Physics Teachers, Chesapeake Section ............ Peggy A. Dixon

Wptical Saciety of America, National Capital Section...:................... William C. Graver

American Society of Plant Physiologists, Washington Area Section ... Walter Shropshire, Jr.

Washington Operations Research/Management Science Council .............. John G. Honig

hasten Society ef America, Washington Section.........6....0. 00.0 ee eee es Carl Zeller

American Institute of Mining, Metallurgical and Petroleum Engineers,

OED TUG CIT STS SOTO) CAs Sie tc le ones ne Ronald Munson

suomi oe ait ASUTOMOMIETS,, © 4.04424 bah ee ded nae sve ee sees escent Robert H. McCracken

Mathematics Association of America, MD-DC-VA Section ................ Alfred B. Willcox

Mistncto Columbia Institute of Chemists.................00.6000eeees Miloslav Rechcigl, Jr.

Pistnce ot Columbia Psychological Association ........ 2.25. .....0.00 eee sees ecneee Jane Flinn

Poasnmeton amt Mechineal (Group... 0. belo. ee ck ek hele ee de eed Robert F. Brady, Jr.

American Phytopathological Society, Potomac Division................... Deborah R. Fravel

Society for General Systems Research, Metropolitan Washington

Ee ee Ue ie tess ee dea Ven dann bod balew es Ronald W. Manderscheid

Diviial Factors Society, Potomac Chapter... 6) 6. .i coe gees ote cee ee Thomas B. Malone

Parmctican bishenes Society, Potomac Chapter ....................04.0e0.8s5- Robert J. Sousa

Assouiation for Science, Technology and Innovation................0.0.2.0.05: Ralph I. Cole

Pastel SORIOMMBICAl (SORICEY bic ee dn hee eek cee eels odes eee es Ronald W. Manderscheid

Institute of Electrical and Electronics Engineers, Northern Virginia Section.... Ralph I. Cole

Association for Computing Machinery, Washington Chapter............. Charles E. Youman

Py eM Rath TASCA SOCICLY .. 6 555 lenis ce ee blew weet blew e ewanienas bh Robert Jernigan

Society of Manufacturing Engineers, Washington, DC Chapter ............. James E. Spates

Sovicty wir tndusinal Enemecrs, Chapter 14... 0.2... ccd ced ewe ees John Larry Baer

Delegates continue to represent their societies until new appointments are made.

Washington Academy of Sciences

1101 N. Highland St.

Arlington, Va. 22201

Return Requested with Form 3579

2nd Class Postage Paid

at Arlington, Va.

and additional mailing offices.

" f/ VOLUME 80

Number 2

Journal of the Tee 100

WASHINGTON

ACADEMY ..SCIENCES

ISSN 0043-0439

Issued Quarterly

at Washington, D.C.

CONTENTS

Introduction:

HENRY L. TAYLOR, “Human-Computer Interaction: Psychological Per-

spectives”’

©) iu je le faye jee (oo) te) [s) 'o @ (6 (ele. «) @ © 6/)\@ erie © \e) @ 6 eo 6 © © @ % 0) ©) 0 © © © ve, wes je @) 0) ee oie ec ete

Conference on Human-Computer interaction, jointly sponsored by the Human

Factors Society, Potomac Chapter, and the American Psychological Association,

Division of Experimental and Engineering Psychologists, in Arlington, VA, March

1-2, 1990.

PART I

Articles:

DAVID E. KIERAS, “An Overview of Human-Computer Interaction” .... 39

DANIEL R. SEWELL and WILLIAM B. JOHNSON, “The Effects of

Rapid Prototyping on User Behavior in System Design” .................

Washington Academy of Sciences

Founded in 1898

EXECUTIVE COMMITTEE

President

Armand B. Weiss

President-Elect

Walter E. Boek

Secretary

F. K. Mostofi

Treasurer

Norman Doctor

Past President

Robert H. McCracken

Vice President, Membership Affairs

Marie Bourgeois

Vice President, Administrative Affairs

Grover C. Sherlin

Vice President, Junior Academy Affairs

Marylin F. Krupsaw

Vice President, Affiliate Affairs

Edith L. R. Corliss

Board of Managers

R. Clifton Bailey

Jean K. Boek

James W. Harr

Betty Jane Long

Thomas N. Pyke

T. Dale Stewart

REPRESENTATIVES FROM

AFFILIATED SOCIETIES

Delegates are listed on inside

rear cover of each Journal.

ACADEMY OFFICE

1101 N. Highland Street

Arlington, VA 22201

Phone: (703) 527-4800

EDITORIAL BOARD

Editor:

John J. O’Hare, CAE-Link Corpora-

tion

Associate Editors:

Albert G. Gluckman, University of

Maryland

Joseph H. Neale, Georgetown Uni-

versity

Marc Rothenberg, Smithsonian Insti-

tution

Marc M. Sebrechts, Catholic Univer-

sity of America

Edward J. Wegman, George Mason

University

The Journal

This journal, the official organ of the Wash-

ington Academy of Sciences, publishes orig-

inal scientific research, critical reviews,

historical articles, proceedings of scholarly

meetings of its affiliated societies, reports of

the Academy, and other items of interest to

Academy members. The Journal appears

four times a year a June, September,

and December). The December issue con-

tains a directory of the current membership

of the Academy.

Subscription Rates

Members, fellows, and life members in good

standing receive the Journal without charge.

Subscriptions are available on a calendar year

basis, payable in advance. Payment must be

made in U.S. currency at the following rates:

U.S. and Canada... ....0;233eeeee $19.00

Other countries... 2. .)..). 22 22.00

Single copies, when available......... 7.50

Claims for Missing Issues

Claims will not be allowed if received more

than 60 days after the day of neges plus time

normally required for postal delivery and

claim. No claims will be allowed because of

failure to notify the Academy of a change of

address.

Notification of Change of Address

Address changes should be sent promptly to

the Academy office. Such notifications

should show both old and new addresses and

zip-code numbers, where applicable.

Published quarterly in March, June, September, and December of each year by the

Washington Academy of Sciences, 1101 N. Highland Street, Arlington, VA 22201.

Second-class postage paid at Arlington, VA, and additional mailing offices.

Journal of the Washington Academy of Sciences,

Volume 80, Number 2, Pages 37-38, June 1990

Human-Computer Interaction:

Psychological Perspectives

Henry L. Taylor

Institute of Aviation

University of Illinois at Urbana/Champaign

The field of Human-Computer Interaction (HCI) has developed primarily

during the last 15—20 years. Concern for HCI, however, was expressed much

earlier by Mauchly in 1947 when discussing the ease of use of machine coding

systems to program EDVAC he stated:

‘“‘Any machine coding system should be judged quite largely from the point

of view of how easy it is for the operator to obtain results.”’ (Randell, 1973).

Hardware concerns such as the design of CRT screens and input devices

dominated the early HCI efforts. More recently, interest has shifted toward

principles of information presentation and ease-of-use concerns.

The introduction of the first time-shared interactive system in 1963, 1.e.,

MAC, and the BASIC programming language stimulated interest in human

factors problems of non-specialists users. The growth in power and speed and

the reduction in size and cost of computers through the reduction in size of

the switching unit from the transistor to large scale integration and very large

scale integration has helped promote interest in human-computer interaction

issues. Indeed, the development of the personal computer during the mid-

1970’s, which made low-cost computers with graphic displays available, pro-

vided an important impetus to the increased use of computers in psychological

studies. Today, the field of HCI is a collaborative endeavor among computer

scientists, human factors specialists, and psychologists.

The importance of the subject of Human-Computer Interaction has been

recognized in the computing and information technology industry during the

last few years. In the United States, the number of human factors specialists

working in the HCI area is estimated to have increased four to five times

38 HENRY L. TAYLOR

during the last decade. The ACM-SIGCHI Conference, first conducted in

1982, has become an annual event; and the attendance approaches 2000 per-

sons. Recent surveys by the Human Factors Society and the Ergonomics

Society indicate that about 50 percent of their membership are occasionally

involved in the design and the evaluation of human-computer interfaces.

With this brief history in mind, the Executive Committee of the Applied

Experimental and Engineering Psychologists, Division 21 of APA, agreed to

sponsor a mid-year meeting with the Human Factors Society Potomac Chapter

in Washington, DC, March 1-2, 1990, on the subject ‘Human-Computer

Interaction: Psychological Perspectives.”’ On March 1, 1990, David E. Kieras,

University of Michigan, presented a tutorial, ‘““An Overview of Human-Com-

puter Interaction.”” On March 2, 1990, the following papers were presented

at the Symposium: “‘Application of the Model Human Processor,” Bonnie E.

John, Carnegie-Mellon University; ‘““The Effects of Rapid Prototyping on User

Behavior in System Design,” by David L. Sewell, Search Technology, and

William B. Johnson, Galaxy Scientific Corporation; “‘Natural Language In-

terfaces,” by Lance A. Miller, Science Applications International Corpora-

tion; ““Knowledge Representations Used by Computer Programmers,” by Scott

P. Robertson, Rutgers University; ‘““The Human Factors of Voice Interfaces,”

John C. Thomas, NYNEX; and “Advisory Materials for Computer Interfaces:

Using Written and Animated Graphical Instructions,” by Jay Elkerton, Uni-

versity of Michigan. Most of the papers from this symposium are presented

in two special issues of the Journal of the Washington Academy of Sciences.

Reference

Mauchly, J. W. (1973). Preparation of Problems for EDVAC-Type Machines. In B. Randell, (Ed.), The

Origins of Digital Computers: Selected Papers (pp. 365-369). Berlin: Springer-Verlag.

Journal of the Washington Academy of Sciences,

Volume 80, Number 2, Pages 39-70, June 1990

An Overview of Human-Computer

Interaction

David E. Kieras

Program in Technical Communication, College of Engineering

University of Michigan, Ann Arbor, Michigan 48109-2108

ABSTRACT

An overview is presented of the field of Human-Computer Interaction (HCI), with an

emphasis on the topic of user interface design. The kinds of activity in HCI are summarized,

and the relationships to the field of Human Factors and Computer Science are discussed.

Two approaches to the user interface design problem are summarized: the standard human

factors approach, and the newer theory-based engineering model approach. The basic

concepts of user interface design within these two approaches are briefly described. Key

references to the literature are included.

Introduction

An informal overview of Human-Computer Interaction (HCI) is provided.

The overview has four parts: First, HCI, and its basic subject matter, user

interface design, is defined with some discussion of its current importance.

Second, HCI as a field of academic and industrial activity is described in terms

of the participants and their orientations. Third, the two basic approaches to

HCI and user interface design are described and compared; these are the

standard Human Factors approach and the newer Engineering approach. The

fourth section is a brief summary of some key principles and concepts from

both approaches. The paper concludes with a brief statement about future

goals for the HCI field.

What is Human-Computer Interaction?

Human-Computer Interaction (HCI) is a broad field. It includes the social

impact of computers, the use of computers in education, and potentially every-

thing else involving humans and computers. This is not a homogeneous field

but a bewilderingly varied one; there is no universally held agreement on

40 DAVID E. KIERAS

research topics, goals, methodology, or a standard curriculum. But one shared

belief is that improving human-computer system performance can be accom-

plished through better design of the user interface. Thus the main emphasis

in HCI is user interface design.

User Interfaces

The user interface is that part of the system that the user directly interacts

with; for example, it is the portion of the system that makes a PC different

from a Macintosh from the user’s point of view. The user interface involves

the specific hardware features of the machine, such as the type of display and

input devices, and the behavior of the software, which includes several aspects:

@ The specific behavior of a particular application program.

@ Conventions or user interface standards for a machine or system.

@ The general environment supplied by the machine.

@ The specific behavior of the operating system.

@ The supporting documentation, training, and online help.

Where do User Interfaces Come From?

Most user interfaces are designed by programmers whose primary concern

is functionality of the software. The software developer’s intuition is the pri-

mary driver of the design, and serious use of human factors is the exception,

rather than the rule. Bennett (1986) has pointed out that the manager of a

software development effort is judged to succeed or fail depending on the

function of the product, the cost of development, and the schedule, whether

the product is shipped on time. The problem is that the manager of a devel-

opment effort is normally not evaluated in terms of the usability of the product.

User interfaces are developed in three important settings which can be

briefly described. The first is the mainframe system software, which is the

kind of software that runs on large administrative systems, typically based on

IBM hardware. Functionality is almost the exclusive concern for this software,

and there is very little concern for usability. Because of the large initial expense

of these systems and their software, there is almost no possibility for change,

sO users must adapt to the system, no matter how clumsy it is. Thus there is

no market pressure for usability, and the users, typically secretarial and clerical

personnel, are a captive audience. Under current standards of usability, these

systems are often a nightmare, even though the tasks involved, such as ordering

supplies, conceptually are far simpler than those the users routinely perform

with high-end word processors on smaller machines.

Personal computer software is developed in two major settings. The first is

OVERVIEW OF HCI 41

the larger companies that have prospered because one or more of their prod-

ucts has become a standard of its type. New releases of these products are

becoming increasingly more complex, but because of the large market share,

they are starting to suffer from the captive audience syndrome. Users have

become so accustomed to these high functionality packages that the cost of

switching to a different one has become high. Thus in some cases, even in the

personal computer market where the usability of software is generally high,

we are beginning to see the complex and confusing interfaces characteristic

of mainframe system software.

However, in the second personal computer setting, software is developed

in the context of small companies that are usually based on an entrepreneur

or creative individual. Such products are based on an inspiration for new

functionality or an especially good user interface for an old idea. A couple of

examples can be given: Cricketgraph for the Macintosh does not really in-

corporate any new ideas about data graphing functionality, but it has become

a best seller simply because this functionality is extremely well delivered,

making graphs very easy to construct. ThinkTank, originally on the PC, is an

example of an idea for a new type of functionality, namely the “‘outline pro-

cessor.” The user can construct and manipulate outlines, manipulating ideas

with much more ease than paper and pencil or conventional word processors.

ThinkTank first popularized this idea, and although it had a very clumsy user

interface, it was still a very functional and important program. Its successor

is the More II program on the Macintosh in which the functionality is coupled

with a much more usable interface. Despite the success of personal computer

software developed in small companies, the development process for the user

interface is largely “‘seat of the pants’’, and the evaluation of the interface is

primarily through user testing of early releases (“‘Beta Testing”) and user

comments and complaints.

Why HCI is Important Now

HCI is a “‘hot” topic now because personal computer technology has made

everybody a user, and this mass market for software makes usability, as well

as functionality, a selling point for software. But in addition, over the last few

years the Macintosh interface has set a new standard for usability, in that

many computer users now expect a consistent and highly usable set of appli-

cation interfaces. These have flourished in the Macintosh software market-

place because of deliberate strategies on the part of Apple in the form of

guidelines and system architecture than encourage applications to conform to

the standard user interfaces.

42 DAVID E. KIERAS

The Future of Usability

But the future is by no means assured. Good user interface design can still

not be taken for granted. The computer industry is still doing user interface

design primarily by the seat of the pants: It is mostly intuitive, with very little

systematic research or testing and evaluation. The consequences so far are

that there have been many wasted opportunities for progress. For example,

new systems, such as the Next computer, are not necessarily any better than

existing systems and might even be worse. Also, the Macintosh has been on

the market for several years now, but no real competitor has emerged. Instead,

various myths and misunderstandings about usability have arisen, leading to

pointless lawsuits over “look and feel’, which are misguided because they

miss the point of the Macintosh user interface. The widespread misunder-

standing of the Macintosh-style graphic user interface (GUI), seems to have

misled many computer industry experts into the belief that simply copying the

superficial aspects of the Macintosh interface will automatically lead to a more

usable system (see Elkerton and Palmiter, this conference). There are also

some great potentials for disaster, such as the drive to make UNIX and some

of its new GUIs a standard. No systematic research or testing has been done

on these new interfaces, and UNIX is well known to be barely usable at all.

The forces within the computer industry pushing for these standards could

thus trap users for quite a few years into marginally usable systems. So progress

is not assured by any means; it is quite possible that in the year 2000, people

will be struggling with systems that are Jess usable than currently available.

What is the Relation Betwen HCI and Computer Science?

Computer science should include HCI, but generally does not, in spite of

the fact that many conventional computer science concepts are actually based

on human performance, and the limiting factor in computer system perform-

ance is often the human user. Three arguments will be given that HCI is a

central part of computer science: First, many key concepts are psychologically-

based. For example, because humans have a capacity for relatively fast input,

but relatively slow central processing, and slow output, time-sharing makes

sense; a Shared computer system can rapidly transmit information to multiple

user displays and then wait for individual users to eventually hit the keys on

their keyboard. If human performance did not have this property, time sharing

as we know it would not make sense. Other concepts, such as structured

programming, or the widespread belief that user interfaces should involve a

minimum number of keystrokes, are also based on beliefs about human per-

OVERVIEW OF HCI 43

formance. A second argument is that most application programs consist mainly

of user interface code, as contrasted with the code that actually does the

computations in question. Thus, much of the software development cost is in

the user interface. It only makes proper economic sense to insure that the

design of the interface is good, so that the code will not have to be rewritten,

and that there is some proportionality between the cost of the code and its

impact on usage of the product. A third argument is that in modern computer

systems, human performance is the basic bottleneck in the total system

throughput. Computer scientists will worry themselves sick over squeezing

every last fraction of MIPS out of computer hardware, but will think nothing

of delivering a user interface that repeatedly stumps the user for seconds at

a time. With modern technology, confusing the user for a second’s time is the

equivalent of throwing away millions of instructions. Computer scientists would

never allow their hardware to be so radically inefficient, but they have not

broadened their scope to include the throughput of the whole human-machine

system.

Thus courses in human-computer interaction or user interface design are

not common in computer science departments. Where they do exist, they tend

to be viewed with some skepticism as being “‘soft”’ or “not really engineering.”

The field of HCI generally has a problem in that it has not established its

credibility as a technical discipline with mainstream computer science. Within

industry, computer specialists are usually not aware of HCI techniques or

results, and often do not take usability seriously; they believe it is simply a

matter of subjective opinion rather than a specifiable design goal like other

aspects of the computer hardware and software.

What is the Relation Between HCI and Human Factors?

Likewise HCI should be a subset of Human Factors, but judging at least

from the standard human factors textbooks and other sources, HCI is either

ignored or given fairly low priority. For example, in the Sanders and Mc-

Cormick (1987) textbook on human factors there is no index or table of

contents entry for computer; the only related topics are cursor positioning

devices, alphanumeric displays, and the physical ergonomic issues in VDT

workstation design. Even the more psychologically-oriented Kantowitz and

Sorkin (1983) textbook discusses computer systems only in the context of data

entry systems and computer programming, which is a reasonable topic, but

certainly not a mainstream one.

Interestingly enough, the Salvendy Handbook of Human Factors (1987)

includes a substantial number of chapters on human factors in computer sys-

44 DAVID E. KIERAS

tems. The topics are ergonomic design of VDT workstations, software psy-

chology, user interface design, input devices, speech I/O, text editors, and

some other miscellaneous topics. Thus within Human Factors, HCI is only

sometimes considered an important topic.

HCI People and Activities

One way to get a picture of the nature of work in HCI is to summarize the

different kinds of people and activities in the field. This summary is somewhat

flippant, but it will convey the overall flavor of the field. A more systematic

and thorough description is a sociological task well beyond the scope of this

presentation.

The Psychologists

Human factors stalwarts. Many human factors people are moving into

HCI under the assumption that improving interface usability is just another

human factors problem, and standard human factors approaches and tech-

niques can solve it. But as will be argued more below, traditional human

factors methods do not appear to be fast enough for the typical product

development cycle, and the traditional strengths of human factors do not seem

to be the main issues in computer usability.

Rapid evaluators and prototypers. ‘This group is attempting to keep the

standard human factors approach in the game by trying to develop better and

faster ways to mock up and evaluate systems. They have had some notable

successes, but the data collection process is still fairly slow and expensive.

These methods require a great deal of understanding of behavioral data, and

so they are not generally usable by the computer science professionals re-

sponsible for developing user interfaces. However, if the human factors spe-

cialist has control of the interface design, these methods appear to work very

well.

Cognitive psychology strip miners. {have a certain affinity for this position

because it is where my own interest in HCI started. This is a group of academic

cognitive psychologists, who recognize that HCI is a good place to study some

basic questions in cognitive psychology, such as how people acquire and use

mental models or procedural knowledge. I am calling them ‘“‘strip miners”

here because the basic philosophy is to exploit the area by mining the useful

results and then moving on to some other research domain. Thus this group

does not have a basic commitment to improving user interfaces.

OVERVIEW OF HCI 45

The problem for the cognitive psychology strip miners is that there is in

fact nothing special about HCI situations from the basic research point of

view. That is, the same basic research questions appear in a variety of other

settings, such as the traditional control panel systems that are very common

in human-machine interaction situations. In addition, HCI situations are in

fact often very difficult to study. One must implement a simulation of a real

piece of software, or instrument an existing piece of software, or use tedious

videotape methods. All of these involve considerable time and expense, which

is rarely repaid in basic scientific results.

Would-be cognitive engineers. ‘This is the group that I currently place

myself in. The concept of this group 1s that the HCI specialist should be able

to do engineering, just like the rest of the design team. HCI engineering

consists of constructing quantitative and predictive models of the HCI situation

that can be used to evaluate user interface designs without empirical testing.

The basic source of these models is current cognitive psychology results and

theory. Despite some research successes, the problem for this approach is that

it has not really been invented yet; there are no convincing demonstrations

that the approach is useful in product design. In addition, the engineering

approach to HCI seems to rub almost everybody else in the field the wrong

way.

The Computer Scientists

Tool builders. This group is primarily interested in developing tools for

building user interfaces; thus they are especially concerned with user interface

management systems and prototyping tools, but they approach it from a com-

puter science orientation. They are primarily interested in the technology of

user interfaces rather than their usability. The concern that I have with this

effort is that they may be merely developing efficient ways for software de-

velopers to continue building poor interfaces.

Artificial intelligence strip miners. Like the cognitive psychology strip

miners, this group is interested in user interfaces only to the extent that this

topic is useful for issues deemed important elsewhere. The argument advanced

by this group is that an adequately intelligent computer would solve most

usability problems; rather than making the user figure out the computer, the

computer could figure out the user! There seems to be a belief among AI

people that HCI is a good domain for building certain AI systems, and they

often justify the considerable cost of AI systems with the usability benefits

that would result. The latter is a weak argument, because industry in general

46 DAVID E. KIERAS

does not support user interface work enthusiastically; the Al-based approach

to usability is thus an expensive way to be ignored, rather than a cheap way.

But there is a serious philosophical problem involved in this approach. It

assumes that the computer would be most useful to humans functioning as a

collaborator rather than as a tool. Thus the intelligent computer would present

itself as basically human-like and will attempt to collaborate or cooperate with

the user, instead of passively submitting to the user’s will. Given that most

people have trouble communicating with their secretaries, spouses, or profes-

sional colleagues, it seems foolish to want to have the same problems with

one’s computer, as opposed to simply telling it what to do.

Time will tell whether the computer-as-collaborator model for HCI is viable.

In the meantime, most user interface work focuses on the computer-as-tool

approach. This capitalizes on the apparent fact that most problems with com-

puter system usability are relatively straightforward problems of bad design,

and thus can be fixed in a relatively straightforward way. There is no need to

use artificial intelligence to make systems usable; most of the problems can

be handled simply by better design practices and techniques.

Visionaries and technology innovators. ‘This group appears to believe that

new ideas about user interface technology, or new computer functionality,

will solve the problem of usability. While it is very clear that a single-minded

emphasis on technology has produced many bad interfaces, and has impeded

usability as often as it has helped, it is also clear that without these visionaries

and innovators there would be no new technology to apply. The problem is

that new technology seems to be applied to user interfaces routinely, without

any thoughtful consideration about whether there will be any net benefit for

the user. The result is a certain amount of wasted effort. Speech I/O, a subject

of intensive development, is a good example; when would the benefits of this

technology compensate for the substantially /ower data transfer rate imposed

on the user?

The Other Disciplines

Sociologists. This group of HCI researchers are focussed on the social

and organizational aspects of human-computer interaction, such as the effects

of computerization on an organization. This work is clearly very important to

everyone concerned with the impact of computers. There are some unexplored

aspects of the social implications of system usability. For example, computer

specialists of the ‘“‘wizard’’ variety take on a certain social role within their

organizations. Perhaps such people have a vested interest in the computer

systems remaining only poorly usable because it enhances their own role. Such

OVERVIEW OF HCI 47

computer experts tend to be male, which is perhaps related to the prototypical

wizard’s desire to have a ‘““macho”’ mastery of the system, or a member of the

secret club or priesthood. Thus there are aspects of usability that relate to the

social organizations associated with computing; these have not been ade-

quately researched. It is possible that such social aspects are a major source

of resistance opposing more usable, efficient software.

Framework seekers. Some groups within HCI apparently believe that the

human-computer interaction situation is special enough, or unusual enough,

that radically new frameworks for understanding human activity are required

to adequately understand how people interact with computers. A more radical

assertion is that approaches based on conventional experimental psychology,

such as standard human factors, are inapplicable to HCI. For example, it is

argued that there are radical effects of context which render standard meth-

odologies inappropriate. It is very hard to characterize these approaches in

any more detail; to the extent that they are radically new frameworks, they

do not translate well to anyone who has not become acclimated to them. But

such assertions are premature; experimental psychology approaches in HF

have not really been applied widely and thoroughly enough yet to support

any judgment that they are not effective. Furthermore, once one trims out

the rhetoric, the arguments of the framework seekers seem to be very straight-

forward within a conventional human factors framework.

Approaches to User Interface Design

The content of the HCI field can be broken into three major segments: the

traditional human factors approach, the engineering approach, and the tech-

nology and techniques for implementing user interfaces. This last topic is fairly

well represented in a variety of sources, such as the Baecker and Buxton

(1987) book of readings, and is often emphasized in HCI courses offered in

computer science departments. This topic will not be dealt with here, since it

is mainly a matter of technology, as opposed to the specifically human-related

aspects of HCI.

Why HCI is Different from HF

HCI is different from HF primarily in the fact that the scope and nature of

user interface design problems are not well handled by standard human factors

techniques and knowledge (Reisner, 1987). The basic problem is that com-

puters are more cognitive in their demands; normally the displays are clearly

48 DAVID E. KIERAS

visible, and the input devices are easy to operate. The difficulties in interacting

with computers are in understanding what to do, not in actually doing it. Thus

key aspects of computer system usability go well beyond traditional human

factors concerns and knowledge. Some of these concerns will be described

more below.

Another source of differences between HCI and HF is that the demands

on the user interface design and evaluation methodologies are more strenuous

than in conventional human factors. The standard human factors approach to

user interface design has only one basic design methodology: First, try to get

human factors specifications included in the overall system design specifica-

tions. Often this cannot be accomplished, at least not in any highly specific

way. Second, the human factors specialist specifies or criticizes a design;

usually he or she is required to criticize an engineer’s design using guidelines

or experience and intuition. If possible the design is tested and evaluated,

using mock-ups if early in the design process, or using prototypes or a first

version of the system, if late. An effort is made to determine if the system

meets the human factors specifications, if any were included, and using in-

formal observational methods to determine if there are any problems. The

human factors specialist tries to get the design revised to solve the problems;

it is normally impossible to make even worthwhile changes if it is late in the

system development process.

It is difficult to use the standard approach of empirical testing of prototypes

for computer interfaces. First, there is an extremely large number of different

possible interface designs, even based on just a standard video display and

keyboard. Thus, it is essentially impossible to carry out a systematic program

of empirical studies that will identify what particular features of interfaces are

good. Another problem is that the current cycle time of product development

within the computer industry is far too rapid for standard evaluation meth-

odology. Software developers usually can not wait for a prototype to be built,

evaluated, and modified. Typically a product is released for sale at the same

time as it becomes testable using conventional methodologies. Notice also

that some forms of empirical comparison are essentially impractical, such as

the assessment of transfer or consistency between two related products. It is

essentially impossible to examine this empirically because construction and

iteration over multiple prototypes is required along with two extensive training

sessions to determine the transfer relationships.

Thus, the basic problem with the standard human factors methodology is

that it is too slow for current software development life cycles. The proposed

solution is the early involvement of human factors considerations in the design

process, and rapid prototyping methodology for quick development. The con-

OVERVIEW OF HCI 49

cept is thus to use the traditional approach, but to get results quickly. But

current experience is that it is still hard to obtain human factors evaluations

rapidly enough to properly drive the development process. Even if the rapid

prototyping methods allow mock-up interfaces to be developed quickly, testing

human subjects and analyzing the resulting data is still a relatively slow process.

Engineering Approach to User Interface Design

The basic concept of this approach is that a user interface should be engi-

neered, that is based on analysis and calculation, rather than on empirical

evaluations of an implemented system. An analogy could be made: Traditional

human factors methods for developing user interfaces correspond to designing

a bridge by following guidelines to construct a bridge that “looks okay” and

then driving trucks over it to test the structural integrity of the bridge. If cracks

appear in the structure, or structural members begin to bend, then the bridge

is either patched up, or torn down and a new one constructed according to

an improved design. Then the test is repeated. This is reminiscent of how

large structures such as cathedrals were constructed in medieval times. Master

builders worked “‘by the seat of the pants” and very often had correct intui-

tions. However, occasionally their intuitions were wrong and they often

overbuilt, and sometimes under-built, resulting in cracking pillars or even

structural collapse. Of course, what we think of as engineering today involves

evaluating a design while it is still in the paper stage; the designer performs

various calculations and looks up general information about strengths of ma-

terials. The calculation shows whether the bridge is acceptably strong for the

intended loads. Only after the paper design has been completed, is the actual

construction undertaken. Normally, and almost always, the constructed bridge

is satisfactory. Rarely, some new phenomenon is uncovered such as the famous

bridge collapse which led to the consideration of aerodynamic factors in bridge

design. However, many thousands of bridges have been designed without

empirical testing of prototypes, but with analysis and calculation before con-

struction. This is what I mean by engineering.

Thus, in the engineering approach to user interface design, analytic models

are used to predict usability from design descriptions or simulations. This is

a rapid approach because working on paper or with computer software or

simulations does not require any mock-ups or prototypes to be built, nor is

there a need to run experiments with human subjects. However, the engi-

neering approach is clearly not complete, so some empirical testing will be

required for reliable system development. So the goal of the approach is

50 DAVID E. KIERAS

modest, in that the effort is to try to get the interface design mostly right

beforehand, working “‘on paper” as much as possible. The slow and expensive

user testing would be reserved for fine tuning, checking, or for protection

against major conceptual errors, rather than being the workhorse for routine

aspects of the design. Notice that traditional human factors has many prec-

edents for analytic approaches, such as visual task design and work analysis

methods. Needless to say, this type of approach is the mainstay of conventional

engineering as well.

Card, Moran, and Newell presented the engineering approach to user in-

terface design in their 1983 book. My long-time collaborator, Peter Polson,

and I were independently educated on this concept by our sponsors and critics

at IBM. That is, for the human factors specialists to hold their own in disputes

with hardware or software engineers, they should be able to supply quantitative

estimates at the beginning of the design stage. These quantitative usability

estimates allow usability to be specified in a testable and objective way, which

is a critical management need (Bennett, 1986; Whiteside, Bennett, & Holtz-

blatt, 1988), and also allow usability to be traded off with other aspects of the

design in a reasonable fashion, on a par with these other aspects. For example,

the design team could make an informed decision whether a 10% increase in

the product’s cost is worth a 10% decrease in the time required to perform

tasks. Furthermore, if these usability estimates can be calculated or predicted

from initial design drafts or specifications, then the design can be iterated

without the construction of any kind of prototype, and without the collection

of actual empirical data. This allows the user interface to be developed on

the same time scale as modern hardware and software engineering practice

requires. For example, the IBM PC was developed and brought to market in

roughly one year; anyone who has participated in any kind of human subject

experimentation knows that no more than a couple of formal tests could have

been accomplished in that amount of time.

As described by Card, Moran, and Newell (1983), the engineering approach

has three aspects: Calculation of quantitative estimates of human performance

needs to be possible. Psychologists have underestimated the extent to which

such calculations are possible; most psychological research has been concerned

with testing contrasting hypotheses and not with estimating parameters of

human information processing. The Model Human Processor, as described

more below, contains pervasive parameter estimates that can be used as a

basis for calculation of human performance. Along with calculation goes the

concept of approximation. Most psychological research has consisted of nit-

picking over very small effects that have strong theoretical implications. For

practical engineering we need to know what kind of approximate calculations

OVERVIEW OF HCI 51

can be made, and what their limits are. Finally, task analysis is a key part of

the engineering approach. Much of human activity is determined by the per-

son’s task, rather than the person’s internal cognitive mechanisms (Simon,

1969). Thus a critical step in designing a user interface will be to describe the

user’s task in great detail, at a level where the task constrains the user’s activity.

That is, the description includes the details of exactly what a user has to do

to accomplish some goal; this is determined by the specifics of the user in-

terface: which commands have to be entered at what time, where the mouse

cursor has to be positioned, and so forth. Any attempt to define a user interface

without such detailed consideration is flawed, since it is not tapping into the

aspects of the task context that in fact most strongly govern the user’s behavior.

Limitations of Engineering Models

There are some important limitations of the current engineering approach

to user interface design. The major limitation is that these methods can deal

only with situations in which the human is tightly constrained by the task;

according to the rationality principle, only here is there a relatively sound

basis for predicting what the user will do. Thus it is useful to distinguish

between the “‘creative”’ parts of a task, such as composing the content of a

document, from the routine parts of the task, such as making specified changes

to a document with a word processor. Both standard human factors and the

engineering approach can assist in designing the routine parts of the task, but

not the more creative parts. Notice that many creative tasks, such as electronic

circuit design, might be considered creative overall, but have many routine

subsections. Notice also that if one is using a computerized tool, such as a

CAD system, to accomplish a creative task, interacting with the tool should

certainly be routine rather than creative. Hence, just because a task has

substantial creative content is no excuse for implementing a poorly designed

user interface. In fact, in creative situations, good design of the user interface

is even more essential, because the user should be free to concentrate on the

truly creative parts of the task, rather than expending effort on trying to master

what is supposed to be a routinely usable tool.

A second limitation of the current engineering approach is that it does not

contribute much when the required information is heavily perceptual or motor

in nature, or parametric in that it depends not on the task or general principles,

but strictly on the parameters of human performance. Examples are the le-

gibility of characters on a screen and the confusability of icons. This is where

traditional human factors results and methods are in fact the strongest. So on

52 DAVID E. KIERAS

the whole, the correct way to view the contrast between the engineering

approach and the standard human factors approach is that they are comple-

mentary; engineering methods based on GOMS task analysis provide a strong

approach to specifying the task-driven aspects of the user interface, whereas

standard human factors methodology is best for choosing those parts of the

interface that depend on empirically determined properties of human per-

formance. |

Survey of HCI Concepts

The following survey of the content of the HCI field is of course very

superficial, and represents a particular point of view. Given the diversity of

the HCI field, many of those active in HCI would disagree with the particular

view presented here.

HCI Sources

Research sources. The single most important book is The Psychology of

Human-Computer Interaction, Card, Moran, and Newell (1983), which pre-

sents an engineering-oriented approach to HCI which will be described more

below. The book is dated, and there are many problems with the experimental

data presented in it. However it remains a unique and indispensable source

in this field because it presents the Model Human Processor, the GOMS model

of HCI, the Keystroke-Level Model for analyzing user interfaces, and an

exemplary piece of work on the quantitative modeling of text selection devices.

These topics will be described more below.

The single book most worth having is the Handbook of Human-Computer

Interaction edited by Helander (1988). This book has a large number of specific

chapters covering most of the scope of human-computer interaction, many of

them directly usable to the practitioner. Two important edited volumes are

the Norman and Draper (1986) volume on User Centered System Design, which

is highly regarded for its conceptual discussion but does not present much in

the way of specific design issues or methods. The volume edited by Carroll

(1987) contains many important early research papers on HCI.

A key research outlet is the Proceedings of the Annual Conference of the

ACM Special Interest Group on Computer-Human Interaction Conference

(SIGCHI). Many researchers present short papers at this meeting, and so the

published proceedings have interesting and important research results. The

problem is that few of these papers go on to be published in a more substantial

OVERVIEW OF HCI 53

form, reflecting the fact that the HCI field has not settled upon a single set

of primary journals closely associated with HCI.

Textbooks. Textbooks normally provide a survey of the core of a field,

but the available HCI textbooks are very weak. Perhaps the earliest popular

textbook is Rubenstein and Hersh (1984); however it appears to be written

for a management audience rather than a student audience. It is extremely

informal, and as a result, is often hard to apply and is not very deep or specific.

But this book does do an excellent job of laying out some important issues

and making some important distinctions that are missing in other sources. The

textbook by Shneiderman (1987) has a good coverage of topics, and is much

more specific and applicable than the Rubenstein and Hersh book. However

the quality of discussion and interpretation of results is often very erratic, and

is not clear who the intended audience is; the treatment of experimental results

is appropriate neither for experimental psychologists or HF specialists, nor

for computer scientists. Comprehensive and coherent integrations of empirical

results are lacking.

Bailey’s (1989) text is an interesting oddity. It is a mixture of traditional

human factors and the basics of HCI. Standard human factors textbook fare,

such as speech communication and the design of knobs and dials (which are

extremely rare on today’s computers) is presented, though it has little direct

relevance to HCI. Finally Baecker and Buxton (1987) is a large and compre-

hensive collection of readings. It includes quite a bit on the design and im-

plementation of user interfaces but some other key material is not well rep-

resented.

General Design Issues

Usability as a design goal. An important activity for HCI specialists is to

document cases where poor user interfaces have essentially destroyed the

functionality of a product. These cases can be used to help persuade computer

scientists and product development managers that user interface quality is

important. A favorite case is Frye and Soloway (1987) in which the user

interface of a popular piece of educational software was so difficult that the

only children who could successfully use the software were those who already

understood the mathematical concepts that the software was supposed to

teach. Thus a poor interface can make a piece of software pointless.

On the other hand, a paper by Goransson, Lind, Pettersson, Sandblad, and

Schwalbe (1987) makes the equally valuable point that sometimes the user

interface is not the problem. There are many system design problems, issues

54 DAVID E. KIERAS

of functionality, and overall usability that are not a result of a poor user

interface, but rather are due to a bad match between the overall user’s task

and the functions offered by the computer system. For example, after rewriting

a database access system used in a business organization to make it more

usable, the developers discovered that in fact the business organization had

no apparent need for the database at all. In another example, the extremely

difficult problems of a multiple-access scheduling system for a medical clinic

disappeared when it was realized that the more fundamental problem was that

the medical clinic was far too large anyway. When the clinic was broken down

into smaller units that were more responsive to patients, the need for an

elaborate scheduling system was eliminated.

Designing for usability. The traditional design process for a user interface

is similar to that in standard human factors. It consists of using guidelines in

the initial stages of the design, followed by some type of user performance

testing of a prototype, which can be as simple as sketches on pieces of paper.

The designer should then iterate the design until a satisfactory result is achieved.

Gould (1988) has an excellent presentation of this process, based on actual

experience, and which presents many specific pointers and suggestions.

A fundamental issue in system design is how users should understand the

system. This is a topic in what can be termed conceptual models or mental

models. This is a difficult issue for theoretical psychology, and many unclear

discussions and muddled terminology are present in the literature. Two good

discussions of this topic are in Rubenstein and Hersh (1984), and in Kieras

(1988a, 1988b). One basic issue is the familiarity of the system, or the value

of the metaphor or overall concept of the system that the user is invited to

learn. For example, in business settings, an interface that resembles traditional

paper forms is often a very easy one for users to understand. In contrast, a

database retrieval system using Boolean expressions does not resemble any

conventional business device, and one can expect users to have more difficulty

learning to use it. A second issue is the extent to which the user needs to

understand the internal workings of the system, as opposed to being able to

work at the “‘outside”’ level of the system. Many traditional computer systems

seem to demand a fundamental understanding of the internals of the system,

possibly at a superficial level of analysis, but directly in terms of the actual

mechanisms and processes in the system. It seems more desirable for the user

interface to hide these mechanisms, so that the user does not need to know

the internal operations of the system. A good example is the Macintosh in-

terface, which goes a long way towards concealing many traditional computer

system concepts; indeed, it seems that many traditional computer experts have

been temporarily confused by the invisibility of the underlying system. For

OVERVIEW OF HCI 55

example, installing an operating system is normally a subtle process; on the

Macintosh it is almost trivial.

User Interface Styles

User interfaces come in different styles, such as menus, command language,

forms, and direct manipulation. Shneiderman (1987) presents a good overall

survey of the different user interface styles, and lists advantages and disad-

vantages of each. This discussion will be elaborated by making use of more

recent results.

Menus. A classic problem in menu design is the question of whether menus

should be broad (many choices), or deep (many levels, few choices). The

traditional advice is to limit the number of choices on a screen to about 7,

but this limit was never given a substantial justification other than it being

related to short-term memory (STM) capacity, but the relation of STM ca-

pacity to menu usage was never analyzed in enough detail to justify this

guideline. A special case of menu interfaces has been analyzed on a quanti-

tative basis by Landauer and Nachbar (1985). The time to arrive at a final

selection of a number in a touch-screen menu interface was a combination of

Fitts’ Law and Hicks’ Law. The specific parameters suggest that if the user

can rapidly locate a menu item on the screen, broad rather than deep menus

would be preferable. However, Landauer and Nachbar point out that the

actual design decision should be based on the quantitative specifics of visual

search time versus choice response time. Somberg (1987) showed that main-

taining menu items in the same position in a mouse-based pull down menu

system was a better solution for long-term use than various clever ways of

rearranging the order of menu items dynamically. Shneiderman (1987) pre-

sents a good discussion of what he terms the BLT menu interface; an example

is the menu interface used in the Lotus 1-2-3 spreadsheet program. In this

type of interface, each menu choice can be selected by typing in the first letter

of the menu item. The items in each menu have been selected so that they

have a unique first letter that is reasonably mnemonic, but this same letter

might be used in other menus. The interface is designed so that these first

letter items can be typed in as a stream, and the menu system immediately

moves from one menu to the other. If the user stops typing and looks up at

the screen, they see the menu corresponding to the last item they typed.

The BLT interface is an especially interesting-and valuable form of menu

interface because the mnemonic single letter items function as commands that

the user can learn incidentally while using the menu system. For example, in

56 DAVID E. KIERAS

Lotus 1-2-3, the user can type RFC2 which corresponds to the command

“Range Format Currency 2 decimal places” and can learn this command string

“along the way” without a specific memorization effort, or referring to the

manual, or learning arbitrary command keys. If at any point the user does

not remember the next letter in the command string, the user can simply look

at the screen and drop back into menu-following mode. A similar type of

interface, except normally only one command at a time, appears in the Mac-

intosh. By convention, single-keystroke shortcut symbols (‘power keys’’ or

“hot keys’) are displayed next to heavily used menu items so that the user

has immediate access to the definitions and can learn them as convenient, and

at any time can revert to using the ordinary menu selection methods. Thus

the user can make a seamless transition from beginning user to “power user”

of an interface, with no penalties along the way.

Experts vs. new users—a false tradeoff. It has long been assumed that

there is a necessary tradeoff between an interface that is suitable for new users

and one that is suitable for experienced experts. The BLT menu, described

above, is one way in which this apparent tradeoff can be circumvented. How-

ever, a key study by Whiteside, Jones, Levy, and Wixon (1985) shows that

in general the tradeoff is not true. Whiteside ef al. studied several different

Operating systems that used different interface styles. Users with different

backgrounds learned how to use the systems and then carried out a set of

benchmark tasks. On the whole, those systems that were easiest to learn were

also those that were fastest and easiest to use. A similar conclusion was reached

by Roberts and Moran (1983) in a classic study of text editors. Whiteside et

al. conclude that the quality of an interface implementation, the many small

details involved in producing a good interface, were more important than the

overall interface style.

Command languages. Command language interfaces are the traditional

form of interface largely because they are relatively easy to implement, having

appeared in the very first batch systems, and then the first interactive systems.

However command languages are typically difficult to learn. The studies in

the field have not made this issue very clear, but a variety of studies suggest

that some command language organizations are fundamentally easier to learn

than others (Shneiderman, 1987). My argument is that the basic difficulty of

a command language is related to the difficulty of synthesizing a command;

that is, users must be able to create commands, not simply perform verbatim

recall of the exact form of the command. Taking advantage of this concept,

a new class of command language interfaces could be defined that are much

easier to learn and to use than existing ones. One example is cross-product

command languages, in which commands are composed by combining a small

OVERVIEW OF HCI 57

set of action-designating verbs with a small set of object-designating nouns.

Thus, instead of an individual special command for each combination of object

or action, the user can instead learn a small set of action verbs and object

verbs and synthesize commands based just on these two small sets. A similar

argument applies to the difficulty of command abbreviation schemes, which

is a heavily researched topic (Shneiderman, 1987, for a review). In both cases

it appears that the basic governor of difficulty is the consistency and extent

of a pattern in the commands or abbreviations that reduces the amount of

special case memorization and verbatim retrieval that the user must perform.

Direct manipulation. Direct manipulation interfaces are those in which

by using some kind of pointing device, typically a mouse, the user controls

the computer by manipulating objects on the display in some form of intuitive

and straightforward way, normally manipulating objects spatially with physical

actions, rather than linguistically structured commands. But direct manipu-

lation is hard to define. There are various discussions in the literature, but on

the whole these comments are theoretically weak or incoherent. But, the basic

principle seems to be that direct manipulation interfaces are organized so that

perceptual and motor activities replace activities that otherwise require ex-

tensive cognition (Elkerton and Palmiter, this conference). Thus, on the Mac-

intosh one can copy a file from one disk to the other simply by locating it

visually and then physically “dragging” the file to its destination. This activity

seems indistinguishable from moving a physical document from one physical

folder to another. In contrast, traditional computers require the user to syn-

thesize an exact string of letters comprising a command, following a defined

syntax which on occasion can be quite convoluted. While direct manipulation

interfaces clearly work very well when there is a concrete physical and spatial

metaphor to be invoked, such as moving documents about a desktop, or

drawing pictures, it is unclear whether they have any advantage in processes

or activities that are not inherently spatial or do not have a clear spatial

metaphor. For example, although there has been considerable interest in visual

programming, there is little evidence that it has substantial advantages com-

pared to traditional text-oriented programming.

Environments. Another set of considerations is interface environment, by

which is meant the entire user’s environment or the user interface of the

system as a whole. For example, workstations often provide access to multiple

windows and various facilities such as a global text editor and window manager.

Card and Henderson (1987) describe the design of a very elaborate system

that provides the effect of very large amounts of organized display space.

Computer systems on a local area network often have other facilities and

features that produce a larger scale environment for the user than the indi-

58 DAVID E. KIERAS

vidual machine would normally have. Likewise, command scripts, history, and

command reentry functions provide a larger scale user interface, in which

users Can sometimes organize their activities at a much higher level than simply

at levels of individual commands. An interesting paper on this subject is

Greenberg and Witten (1988) who found that in UNIX systems a small number

of recently entered commands accounted for the bulk of command usage.

This has strong implications for useful command reentry functions. There are

many different command reentry implementations, but with very few excep-

tions they are extremely crude, often overly complex, and often fail to provide

the type of access that the Greenberg and Witten results suggest. In my

opinion, an exemplary design for command reentry appears in DEC’s VMS

operating system, in which the cursor keys are used to scroll forward and back

through a list of recently entered commands, and to perform simple insertion

and deletion editing on the recalled command before entering it. This design

seems to suffice for most command reentry situations, is trivial to learn, and

quite simple to operate. In contrast, other command reentry facilities, such

as in MS-DOS, seem clumsy and unsuited for routine use.

Input and Output Devices and Techniques

Keyboards. The standard treatment found in Human Factors sources and

the above textbooks on various cursor positioning devices and keyboards are

largely applicable. However, the treatment of the standard Sholes or QWERTY

keyboard is usually erroneous; the textbooks often echo the complaints of

uninformed users about the “‘illogical’’ arrangement of the Sholes keyboard.

The exact origin of the Sholes layout is a problem for historical research, but

it is very clear from the record that Sholes systematically optimized his first

typewriters for speed, engaging in a testing and evaluation program compa-

rable with modern practice. So the most common mythical slur, that Sholes

deliberately attempted to slow the typist, is clearly false. Recent work (Card,

Moran, and Newell, 1983, Ch. 2) points out that the Sholes keyboard is actually

relatively efficient, since it has a high proportion of alternating-hand key-

strokes, which are much faster than within-hand keystrokes (Ostry, 1983).

The more optimal Dvorak keyboard is estimated to be somewhat faster, but

not more than about 20% faster, with some comments that the actual speed

improvement may only be around 5% or so. One can hope that in the near

future we will see an end to the repeated complaints and calls for improved

keyboards, which only serve to distract from other more pressing user interface

problems.

OVERVIEW OF HCI 59

A related issue is the alphabetic keyboard, in which the keys are arranged

in alphabetical order. These are appearing on many small computer-based

devices, such as “‘digital diaries,’ but are known to be fundamentally harder

to learn and slower than the Sholes layout (Norman and Fisher, 1982); the

problems are: (1) almost everybody knows enough of the Sholes layout to

have an advantage; (2) there is no standard alphabetic layout, so users have

to learn each device from scratch, and usually must find letters by slow visual

search; (3) there is little in the way of alternating-hand or other physical

advantages in an alphabetic layout (Card, Moran, and Newell, 1983, Ch. 2,

for an example).

Pointing devices. ‘The comparison of different cursor positioning devices

such as mice, joysticks, hand cursor keys, in most sources is usually presented

in a very superficial fashion. Traditional human factors coverage of different

pointing devices usually presents miscellaneous experimental comparisons which

do not agree on a clear winner and so do not lead to any understanding of

why one device might be better than another. An exemplary piece of work

in this area is Card, Moran, and Newell, (1983, Ch. 7) discussed below, who

constructed a mathematical model of cursor positioning time for each of several

devices; the models provide a much more informed understanding of the

benefits and deficiencies of the individual pointing devices.

Displays. The HF literature has ample discussion of the basic properties

and guidelines for video displays. Several more recent interesting papers can

be mentioned: Burns, Warren, and Rudisill (1986) provides a good example,

with a guideline and experiment-based redesign of the space shuttle displays.

Tullis (1988) provides an analytic tool based on various aspects of the display

density which is clearly useful and important. Gould and his co-workers have

a series of papers comparing reading speed on video displays versus paper,

in which they were able to document that only the most superior computer

displays compete successfully with ordinary paper printing (Gould, Alfaro,

Finn, Haupt, Minuto, and Salaun, 1987). Wise (1986), Eastman, Woods, and

Elm (1986), and Kieras (1988c) discuss the design and properties of graphic

diagram displays.

Helping Users

Errors and messages. ‘There are two key concepts to how a system should

respond to user errors and what messages the system should produce in re-

sponse to an error; these principles are articulated well by Rubenstein and

Hersh (1984). First, wherever possible, the user interface should be designed

60 DAVID E. KIERAS

to eliminate user errors, by making them impossible to produce. For example,

a menu interface can simply refuse to allow the user to select an invalid option.

An example is how the Macintosh interface presents invalid menu choices in

gray and does not allow them to be selected. As explained by Rubenstein and

Hersh, the concept is to ensure that any errors the user makes are ones that

are intelligible to the user in terms of their own problem domain. For example

in a graphing program, they might accidentally select the wrong type of graph.

As soon as they see the cpl they will recognize their error and its cause

immediately.

In contrast, the traditional command language interface allows any input

to be supplied at any time, regardless of whether it is a meaningful or valid

action. For example, one can enter a command to copy a file to a non-existent

directory and is informed of the error only after going to the trouble to create

and enter the command. To a great extent, the designer’s ability to follow

this principle is very limited in command language interfaces. Highly efficient

command reentry and editing is one way to at least alleviate these problems.

The second important principle is that messages produced by the system

should tell the user what to do to recover from an error, as opposed to supplying

a description of the error itself or its effect on the internal operation of the

software. Anyone who has used computers for any length of time has received

unintelligible and unhelpful error messages such as “‘syntax error” or “‘seg-

mentation fault.”’ In some cases, the error information is extremely obscure

even to highly experienced programmers. A well-designed system will include

information either on the screen, or in documentation, that specifically informs

the user what to do for each possible message. An example is the documen-

tation accompanying DEC’s VMS operating system. There is a separate man-

ual of system messages, and the following is a randomly selected sample:

NOBITMAP, no valid storage bit map found on ‘device’

Facility; BACKUP, Backup Utility

Explanation: The Backup Utility encountered an error during an attempt

to search for the storage bit map file [O00000|BITMAP.SYS;1, on the

specified volume. The volume cannot be used as a save set disk.

User Action: Retry the operation using a properly initialized Files-11

Structure Level 2 volume.

The key features of this manual are that every error is explained in a

reasonable amount of technical detail, and then, the user is instructed spe-

cifically what to do to correct the problem. This is the type of information

that should always be supplied for system error messages.

OVERVIEW OF HCI 61

Documentation and online help. Training and reference documentation

and online help is very poorly understood in the HCI literature. As the results

described in Shneiderman (1987) show, online help is a very mixed blessing;

in some studies the online help was actually detrimental. The best current

research is Elkerton’s (1988; Elkerton and Palmiter, 1989; Gong and Elkerton,

1990) which is based on the concept that online help should be based on a

clear and explicit specification of what it is that the user actually needs to

know, organized in terms of what goals the user is trying to accomplish.

Elkerton and his co-workers have demonstrated that online help and training

documentation can be considerably improved by basing it on the organization

and content of a GOMS model, which will be described more below. This

work is exemplary because it is another demonstration of how theoretically

based research, along the lines of the engineering approach, can help clarify

a topic much more powerfully than traditional empirically-based human factors

research.

Some Engineering Models

The Model Human Processor. The Model Human Processor (MHP), pre-

sented by Card, Moran, and Newell (1983, Ch. 2), is a subset of standard

cognitive theory circa 1980 intended to be an engineering model for human

performance. The MHP consists of a set of processors and memories, along

with numerical parameter values for each one. These components are con-

nected in the conventional fashion; the perceptual processor receives visual

and auditory input and deposits the results in the visual or auditory image

store which is defined as a subset of working memory. Working memory is

likewise embedded in long-term memory, a somewhat idiosyncratic arrange-

ment, but consistent with one of the theoretical analyses in the cognitive

psychology literature. The cognitive processor receives input both from work-

ing memory and long-term memory and modifies the contents of working

memory. The cognitive processor is assumed to have a production-system

architecture, in which IF-THEN rules are triggered by the contents of working

memory and long-term memory and modify information in working memory.

However, Card, Moran, and Newell did not explicitly make use of the pro-

duction system architecture. The motor processor is driven by the contents

of working memory, and controls muscle movements. The overall operation

of the MHP is governed by a set of principles. Three important principles

governing human performance are Fitts’ Law, Hicks’ Law, and the Power

Law of Practice. Card, Moran, and Newell demonstrate how Fitts’ Law and

62 DAVID E. KIERAS

Hicks’ Law can be derived at least conceptually from the structure of the

MHP. One other important general principle is the Rationality Principle, which

is essentially a restatement of how behavior can be governed by the task.

Humans try to accomplish their goals efficiently, given the task constraints

and information processing limitations. The claim is that in user interface

situations, the task structure is dominant, which means the specific design of

the computer system will be a primary determinant of the user’s behavior.

The engineering approach to pointing devices. Vhe Card, Moran, and

Newell (1983, Ch. 7) treatment of pointing devices is a good example of the

engineering approach. Most sources in human factors and human-computer

interaction are stymied when it comes to attempting to describe which pointing

device is better in what situation and for what reason. This is because the

standard human factors approach is simply to cite the results of individual

specific experiments. Quite often, the experiments do not agree with each

other, producing confusing results. Card, Moran, and Newell present the

results of a quantitative analysis of a set of different pointing devices in word

processing tasks. The key result is that the mouse follows Fitts’ Law, and the

other devices investigated are slower because movement time is governed

differently, and usually in an inferior way. For example, cursor keys are linear

with the “‘city block” (sum of X and Y) distance, making them on the average

much slower than the mouse. In contrast, the mouse not only follows Fitts’

Law but had the same parameters as the eye-hand system, suggesting that the

mouse, when properly designed, is as good a pointing device as the eye-hand

system permits. This means the mouse can be beaten only in cases where the

eye-hand system is weak, such as small targets (which follows from Fitts’ Law),

or if extra hand movement is required beforehand, such as moving from the

keyboard. But the overall point is that the proper way to evaluate human-

computer interaction situations is with quantitative models of performance,

not gross experimental results.

The Keystroke-Level Model. As an example of a specific engineering tool,

Card, Moran, and Newell (1983, Ch. 8) present the Keystroke-Level Model

for estimating execution times. This method is similar to the work measure-

ment methods used in industrial engineering. It is based on estimating the

overall time for completing a task by summing individual standard values for

the lower-level parts of the task, which are the individual actions. Briefly,

first one determines the sequence of operators required to execute a task, and

then looks up the time for each operator. For example, non-secretarial key-

strokes require about .28 s, while a typical mouse move requires about 1.1 s.

The move between a mouse and a keyboard requires about .4 s. At some

point in the sequence, the user may have to stop and think; this action is

OVERVIEW OF HCI 63

represented with a mental operator with an estimated value of 1.35 s. The

predicted execution time is simply the sum of the operator times.

The Keystroke-Level Model requires a specific task instance, so that the

exact sequence of operators can be tested. However notice that this specific

sequence can be based on a proposed design; it is not necessary to have

implemented anything. Thus, this is an engineering tool in the sense that it

can be used very early in design, as soon as it is possible to specify the sequence

of operations. The main drawback of the method is the need for guessing

where the mental operators are performed. Card, Moran, and Newell provide

some heuristics, but they are incomplete and not adequately general; a better

set of heuristics is badly needed.

The GOMS model. A major contribution of Card, Moran, and Newell

(1983) is to present a general framework for describing the users’ knowledge

of how to operate a system. This knowledge is described in terms of Goals,

Operators, Methods, and Selection rules, from which the acronym GOMS is

obtained. The goals are simply what goals the user can accomplish with the

system, basically what tasks can be performed. The operators are the basic

actions that can be performed on the computer, such as keystrokes or mouse

moves, but also actions on other parts of the task environment, such as turning

the pages of a manuscript. The methods are sequences of operators that are

used to accomplish a goal. Thus they are essentially procedures, but each goal

has at least one method that will accomplish the goal. Selection rules specify

which method should be applied in case there is more than one method to

accomplish a goal. The selection rules contain task- or context-specific infor-

mation to “steer” the user to using the most efficient method. Finally methods

and goals have a hierarchical structure; methods can include operators that

establish sub-goals, which in turn get accomplished by sub-methods. In the

typical computer software task environment, methods and goals have a rich

hierarchical structure. For example, the methods used to move the cursor are

invoked by many different methods for accomplishing different tasks.

Card, Moran, and Newell collected data that supported the psychological

reality of the GOMS categories. However, their mechanisms coupling the

GOMS model to performance were very weak, and the evidence presented for

the validity of the GOMS model as predictor of performance is not impressive.

There is also no approach to how humans learn the GOMS knowledge. It

seems intuitively reasonable that the more difficult or complex a system is to

learn, the more elaborate and voluminous would be the GOMS model needed

to represent the user’s knowledge. Thus there should be some way of making

use of the size or complexity of the GOMS model to predict learning. Finally

Card, Moran, and Newell did not include any methodology for constructing

64 DAVID E. KIERAS

a GOMS model, meaning that it is unclear how and whether it can be used

routinely.

The cognitive complexity model. Kieras, Polson, and Bovair, in a series

of papers (Bovair, Kieras, and Polson, 1990; Kieras and Polson, 1985; Polson,

1987) presented a cognitive complexity approach that is based on using a

production system cognitive architecture to represent GOMS models. This

approach, which was implicit in the Card, Moran, and Newell analysis, but

was developed independently by Kieras and Polson, makes use of the pro-

duction system cognitive architecture to quantify the amount of knowledge

that the user must have. Kieras and Polson saw that production rule repre-

sentations of procedures had essentially the same categories of content as a

GOMS model, and therefore adopted the perspective of representing GOMS

models in terms of production systems. In this approach, a production system

computer simulation model is constructed that can execute the same tasks as

users, interacting with a simulated mock-up of the computer system, and

executing a series of described tasks. The complexity of the production-rule

simulation indicates the complexity of the interface to the user. Bovair, Kieras,

and Polson (1990) provide the specific rules for constructing production system

models that have the desired properties to enable prediction.

There are two main advantages of the cognitive complexity approach:

First, it connects the GOMS model directly into mainstream cognitive theory,

specifically the theoretical work on cognitive skill and learning represented

by the production system cognitive architecture (Anderson, 1983, 1987). Sec-

ond, the production system representation provides quantitative metrics for

predicting certain aspects of usability. Basically, the number of production

rules required in the representation predicts the overall learning time, while

the number of shared rules between two systems or procedures predicts the

amount of transfer of training. The time to execute the production rules, in

terms of the number of production system cycles and the operators involved,

predicts execution time. The amount of information maintained in working

memory predicts the memory load imposed by a task. The current status of

the cognitive complexity work is that the execution time, learning, and transfer

predictions are very well supported across a set of different tasks, different

experiments, and different laboratories (Ziegler, Hoppe, and Fahnrich, 1986).

At this time there has been no work on the memory load predictions (but see

Bovair, Kieras and Polson, 1990).

The applicability of the approach to practical design is more problematic

(Kieras, 1988d, 1988e). Constructing production rule models is a difficult task

that requires substantial expertise in artificial intelligence or cognitive simu-

lation, and is obviously too technically demanding for routine use in practical

OVERVIEW OF HCI 65

interface design. But the real problem is not the production rule programming,

but carrying out the detailed task analysis from which the GOMS model is

constructed. While considerable work has been done on task analysis in the

context of human factors, a GOMS-based task analysis is a specific form in

which the analyst expresses all of the methods that the user requires to actually

carry out the overall task goals. When properly conducted, a GOMS analysis

starts from overall user goals, and goes down the full goal hierarchy until at

the bottom it describes the methods that consist of actual sequences of motor

actions. Most traditional human factors methods of task analysis stop far short

of this level of detail, often being little more than a listing of action-object

pairs, which in a GOMS model framework, is usually only the middle-level

task goals.

Kieras proposed a methodology for constructing and using a GOMS model

(Kieras, 1988d). He provides a set of guidelines for decomposing a task, a

simple notation for expressing GOMS models, and a recipe for constructing

a model in this notation. Calculational procedures are presented for estimating

learning and execution time, based on the relationship of the GOMS model

notation to the production system models. Finally, both in Kieras (1988d) and

in Card, Moran, and Newell (1983, Ch. 12), are suggestions for how a design

can be refined based on properties of the GOMS model representation. These

and some other design principles will be described in the next section.

Some GOMS-Based Design Guidelines

These guidelines are based on implications of the GOMS model and the

concepts of cognitive skill used in the cognitive complexity approach. The

overall concept is that the user acquires and uses procedural knowledge in

HCI situations. The procedural knowledge is organized in terms of a GOMS

model, and is acquired either from explicit descriptions (e.g., instructions, see

Bovair and Kieras, in press) or from problem solving activities based on various

kinds of knowledge, including trial and error (Card, Moran, and Newell, 1983,

Ch. 11). With practice, the procedural knowledge becomes refined and rou-

tinely invoked, as is described in the analyses of the development of cognitive

skill (Anderson 1983, 1987; Card, Moran, and Newell, 1983, Ch. 11). With

extreme amounts of practice, the procedures should become automated and

require very little cognitive processing capacity; however the actual boundaries

of automation in computer usage have not been explored. It is possible that

only the most heavily used activities, such as cursor movement or a few

stereotypical command sequences, achieve an automated state for most users.

66 DAVID E. KIERAS

Interface organization should be in terms of the users’ goals and meth-

ods. The interface should be organized and presented in terms of the user’s

perception of the task, not the programmer’s. This is a standard guideline

within human factors, but if the designer has explicitly developed a GOMS

model that describes how the user is supposed to accomplish goals given a

particular system design, then the designer is in a position to compare the

explicitly developed goals with the structure and specific content of the in-

terface. For example, the designer can compare the menu hierarchy to the

goal hierarchy in the GOMS model, and consider whether the individual words

used in the menus are recognizable as user’s goals. Because the task structure

has been made explicit in the GOMS model, the designer now has a speci-

fication for how the user interface should be organized.

The documentation should present all of the components of aGOMS model

for the task, and provide for access in terms of the user’s goals. This point

is discussed more extensively in Elkerton (1988; Elkerton and Palmiter, 1989;

Gong and Elkerton, 1990). As Elkerton’s work shows very elegantly, explicitly

providing methods in the context of a goal-hierarchical organization radically

improves people’s ability to learn how to perform tasks using documentation

or online help. The major problems with conventional user documentation

become clear from the perspective of the GOMS framework. Users typically

know at least the high level goals that they want to accomplish. If they do

not know how to use the computer system, it is because they do not know

the methods required to accomplish those goals. Thus a user will be entering

documentation or online help with a set of goals in mind, and will be in search

of methods. However, the content of most documentation and online help is

radically unsuited for users in this state of mind. Most documentation in fact

consists essentially of operator documentation. The individual commands are

presented, along with their specific syntax and so forth, but without any

description of what sequences of commands should be used to accomplish

something (methods), or why a particular command would be used (goals),

or in what particular situation (selection rules). Even in the context of Mac-

intosh software, the documentation often consists of a detailed description of

the effect of each individual menu choice. Thus the documentation rarely

presents methods, and rarely presents the document organized in terms of the

higher level goals a user might want to accomplish. Typically the user is reduced

to scanning the documentation, perhaps aided by clever guesses, in an attempt

to find information that can be used to deduce or infer a method. Finally,

documentation rarely includes selection rules. Thus there are often multiple

methods to accomplish a particular goal but the user is in the position of

having to infer or deduce their own selection rules for making use of the

OVERVIEW OF HCI 67

different methods. A common result is that the user will be stuck to using a

method which is extremely inefficient, because they have received no guidance

to the existence of more efficient methods.

Every high-frequency critical task goal should have a simple method. This

is the basic rule for user interface design, but it is often ignored in practice.

The designer can identify the goals which are important and frequently ac-

complished in a task situation and then ensure that the corresponding method

is simple and efficient.

Every goal should have only one method, unless there are specific reasons

for multiple methods. Unnecessary methods just add opportunities for the

user being confused and making errors, as well as increasing the total time

required to learn how to use the system. If there are multiple methods for a

goal, it should be possible to state a simple and clear selection rule for using

each method; if this is not possible, it is a strong clue that the method should

be eliminated—tt is either too specialized, or has no clear function.

Similar goals should have similar methods. A key result from the work

of Kieras, Polson, and Bovair, (Bovair, Kieras, and Polson, 1990; Kieras and

Bovair, 1986; Polson, Bovair, and Kieras, 1987) and related work by Singley

and Anderson (1987-1988), is that the effects of positive transfer in computer

interfaces are extremely large. In fact, a common result found by Kieras,

Polson, and Bovair in various studies was that the extent of transfer, as mea-

sured by the number of shared production rules, was a more powerful predictor

of training time than the individual subject’s own mean training time! This

suggests that developing “consistent”? methods that maximize the amount of

positive transfer is a prime way to reduce the time required to learn a system.

This form of “consistency,” which can be termed method consistency, is bas-

ically that similar goals should have similar methods, where the methods have

been articulated at the detailed keystroke level. Rough similarity at a high

level, or visual similarity in the interface, will not ensure transfer.

Two approaches to ensuring transfer can be listed: first, high-level methods

should share lower-level methods as much as possible. For example, all se-

lection of text should be done in the same way, regardless of what higher-

level operations or mode is involved. This form of consistency reduces the

total number of methods to be learned and prevents mode errors. Second,

conceptually similar goals should have a generalizable method that covers

them all. For example, in many word processors, the move-text goal and the

copy-text goal are accomplished by almost identical methods. A heuristic for

identifying similar methods at this level is whether one can substitute one

name or concept throughout the complete first method to obtain the second.

If the entire method could not be obtained, being able to obtain a sizeable

68 DAVID E. KIERAS

contiguous subset of the second method should be adequate. Based on the

available data, one would have to say if method consistency is not present,

the interface is seriously flawed. However, there are many opportunities for

method consistency in interfaces that often go unexploited.

Error recovery should be possible with routine methods and a minimum of

problem solving. ‘The criteria for good error recovery described above can

be stated somewhat more precisely in the context of the GOMS model and

cognitive skill concepts. If the user can simply learn one method for backing

out or canceling an error, this allows users to recover routinely instead of

having to engage in problem solving. If the error messages supply or identify

a method for recovery, the user can simply read and execute this method

instead of engaging in problem solving to discover a method. Finally, per-

mitting the user to simply retry or reenter a command takes into account the

routine and sometimes automated nature of many methods. If a trivial error

has occurred, it may often be easiest to simply rerun the method. Such schemes

as negotiated error recovery can require the user to learn an unnecessarily

complex set of methods or to engage in problem solving to try to figure out

what the system wants.

What the HCI Field Needs

Traditionally, human factors specialists have always been concerned about

the credibility of human factors within the engineering community, voicing a

standard complaint that designers do not give adequate recognition to human

factors concerns until it is too late. Likewise, user interface design and other

aspects of HCI also have a credibility problem within the computer industry

and academic computer science. What is needed to bolster this credibility is

multiple and highly visible demonstrations of the value of HCI effort in de-

veloping successful systems. In addition, it is critical for practitioners and

researchers in the HCI field to have obvious computer expertise, and more

computer science students should be taught HCI concepts and techniques in

the universities.

My perspective is that there should be more research on the engineering

approach, because this is the approach that will make the most sense to the

technologists, the computer scientists and engineers who normally control the

product development process. The research should emphasize approaches that

are rigorous and simple enough to be taught to computer science practitioners

and to be applied in actual design situations.

Finally, HCI needs more opportunities to validate its proposed methodol-

ogies and concepts in the context of actual product design. Even approaches

OVERVIEW OF HCI 69

as well developed as the cognitive complexity model remain laboratory meth-

ods, and at present we have very little information on whether they have

practical significance in the design of actual systems. The HCI field needs to

gain experience in actual design problems, and take this experience back to

the development of the theory and practice of user interface design.

References

Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press.

Anderson, J. R. (1987). Skill acquisition: Compilation of weak-method problem solutions. Psychological

Review, 94, 192-210.

Baecker, R. M. and Buxton, W. A. S. (Eds.). (1987). Readings in human-computer interaction: A multi-

disciplinary approach. San Mateo, CA: Morgan Kaufmann Publishers.

Bailey, R. W. (1989). Human performance engineering: Using human factors/ergonomics to achieve com-

puter system usability (2nd ed.). Englewood Cliffs, NJ: Prentice-Hall.

Bennett, J. L. (1986). Observations on meeting usability goals for software products. Research Report RJ

5103 (53147). Poughkeepsie, NY: IBM Corporation.

Bovair, S. and Kieras, D. E. (In press). Toward a model of acquiring procedures from text. In Barr, Kamil,

Mosenthal, and Pearson (Eds.). Handbook of Reading Research, Vol. 11. Longman.

Bovair, S., Kieras, D. E. and Polson, P. G. (1990). The acquisition and performance of text editing skill:

A cognitive complexity analysis. Human-Computer Interaction, 5, 1-48.

Burns, M. J., Warren, D. L. and Rudisill, M. (1986). Formatting space-related displays to optimize expert

and nonexpert user performance. In CHI ’86 Proceedings (pp. 274-280). New York: ACM.

Card, S. K. and Henderson, A., Jr. (1987). A multiple, virtual-workspace interface to support user task

switching. In Proceedings of CHI+ GI 1987 (pp. 53-59). New York: ACM.

Card, S., Moran, T. and Newell, A. (1983). The psychology of human-computer interaction. Hillsdale, NJ:

Lawrence Erlbaum Associates.

Carroll, J. M. (Ed.). (1987). Interfacing thought: Cognitive aspects of human-computer interaction. Cam-

bridge, MA: MIT Press. .

Eastman, M. C., Woods, D. D. and Elm, W. C. (1986). Specifying and communicating data structure in

computer-based graphic displays. In Proceedings of the Human Factors Society-30th Annual Meeting (pp.

1034-1037). Santa Monica, CA: Human Factors Society.

Elkerton, J. (1988). Online aiding for human-computer interfaces. In M. Helander (Ed.), Handbook of

human-computer interaction. Amsterdam: Elsevier.

Elkerton, J. and Palmiter, S. (1989). Designing help systems using the GOMS model: An information

retrieval evaluation. In Proceedings of the Human Factors Society 33rd Annual Meeting. Santa Monica,

CA: Human Factors Society.

Frye, D. and Soloway, E. (1987). Interface design: A neglected issue in educational software. In Proceedings

of CHI+ GI 1987 (pp. 93-97). New York: ACM.

Gong, R. and Elkerton, J. (1990). Designing minimal documentation using a GOMS model: A usability

evaluation of an engineering approach. In CHI ’90 Proceedings. New York: ACM.

Goransson, B, Lind, M., Pettersson, E., Sandblad, B. and Schwalbe, P. (1987). The interface is often not

the problem. In Proceedings of CHI+ GI 1987 (pp. 133-136). New York: ACM.

Gould, J. D. (1988). How to design usable systems. In M. Helander (Ed.), Handbook of human-computer

interaction. Amsterdam: Elsevier.

Gould, J. D., Alfaro, L., Finn, R., Haupt, B., Minuto, A. and Salaun, J. (1987). Why reading was slower

from CRT displays than from paper. In Proceedings of CHI + GI 1987 (pp. 7-11). New York: ACM.

Greenberg, S. and Witten, I. H. (1988). How users repeat their actions on computers: Principles for design

of history mechanisms. In CHI ’88 Proceedings (pp. 171-178). New York: ACM.

Helander, M. (Ed.). (1988). Handbook of human-computer interaction. Amsterdam: Elsevier.

Kantowitz, B. H. and Sorkin, R. D. (1983). Human factors: Understanding people-system relationships.

New York: Wiley.

Kieras, D. E. (1988). Mental models for engineered systems and computers. Presented in Workshop on

the Role of Mental Models in User-Centered System Design, under Army Research Institute sponsorship,

hosted by University of Colorado Institute for Cognitive Science, Breckenridge, CO, January 13-16,

1988. (a)

Kieras, D. E. (1988). What mental model should be taught: Choosing instructional content for complex

engineered systems. In J. Psotka, L. D. Massey and S. A. Mutter (Eds.), Intelligent tutoring systems:

Lessons learned (pp. 85-111). Hillsdale, NJ: Lawrence Erlbaum Associates. (b)

70 DAVID E. KIERAS

Kieras, D. E. (1988). Diagrammatic displays for engineered systems: Effects on human performance in

interacting with malfunctioning systems (Tech. Rep. No. 29). Ann Arbor: University of Michigan,

Technical Communication Program. (c)

Kieras, D. E. (1988). Towards a practical GOMS model methodology user interface design. In M. Helander

(Ed.), Handbook of Human-Computer Interaction (pp. 135-158). Amsterdam: Elsevier. (d)

Kieras, D. E. (1988). Making cognitive complexity practical. In CH/ °88 Workshop on Analytical Models,

Washington, May 15, 1988. (e)

Kieras, D. E. and Bovair, S. (1986). The acquisition of procedures from text: A production-system analysis

of transfer of training. Journal of Memory and Language, 25, 507-524.

Kieras, D. E. and Polson, P. G. (1985). An approach to the formal analysis of user complexity. /nternational

Journal of Man-Machine Studies, 22, 365-394.

Landauer, T. K. and Nachbar, D. W. (1985). Selection from alphabetic and numeric menu trees using a

touch screen: Breadth, depth, and width. In CHI ’85 Proceedings (pp. 73-78). New York: ACM.

Norman, D. A. and Draper, S. W. (Eds.). (1986). User centered system design. Hillsdale, NJ: Lawrence

Erlbaum Associates. 2

Norman, D. A. and Fisher, D. (1982). Why alphabetic keyboards are not easy to use: Keyboard layout

doesn’t much matter. Human Factors, 24, 509-519.

Ostry, D. J. (1983). Determinants of interkey times in typing. In W. Cooper (Ed.), Cognitive aspects of

skilled typing. New York: Springer-Verlag.

Polson, P. G. (1987). A quantitative model of human-computer interaction. In J. M. Carroll (Ed.),

Interfacing thought: Cognitive aspects of human-computer interaction. Cambridge, MA: MIT Press.

Polson, P. G., Bovair, S. and Kieras, D. E. (1987). Transfer between text editors. In Proceedings of CHI

+ GI 1987 (pp. 27-32). New York: ACM.

Reisner, P. (1987). Discussion: HCI, what is it and what research is needed. In J. M. Carrell (Ed.),

Interfacing thought: Cognitive aspects of human-computer interaction. Cambridge, MA: MIT Press.

Roberts, T. L. and Moran, T. P. (1983). The evaluation of text editors: Methodology and empirical results.

Communications of the ACM, 26, 265-283.

Rubenstein, R. and Hersh, H. (1984). The human factor: Designing computer systems for people. Bur-

lington, MA: Digital Press.

Salvendy, G. (Ed.) (1987). Handbook of human factors. New York: Wiley.

Sanders, M. S. and McCormick, E. J. (1987). Human factors in engineering and design (6th ed.). New

York: McGraw-Hill.

Shneiderman, B. (1987). Designing the user interface: Strategies for effective human-computer interaction.

Reading, MA: Addison-Wesley.

Simon, H. A. (1969). The sciences of the artificial. Cambridge, MA: MIT Press.

Singley, M. K. and Anderson, J. R. (1987-88). A keystroke analysis of learning and transfer in text-

editing. Human-Computer Interaction, 3, 223-274.

Somberg, B. L. (1987). A comparison of rule-based and positionally constant arrangements of computer

menu items. In Proceedings of CHI+ GI 1987 (pp. 255-260). New York: ACM.

Tullis, T. S. (1988). Screen design. In M. Helander (Ed.), Handbook of human-computer interaction.

Amsterdam: Elsevier.

Whiteside, J., Bennett, J. and Holtzblatt, K. (1988). Usability engineering: Our experience and evolution.

In M. Helander (Ed.), Handbook of human-computer interaction. Amsterdam: Elsevier.

Whiteside, J., Jones, S., Levy, P. S. and Wixon, D. (1985). User performance with command, menu, and

iconic interfaces. In Proceedings of CHI ’85 (pp. 185-191). New York: ACM.

Wise, J. A. (1986). Display systems for electrical system control centers. In Proceedings of the Human

Factors Society—30th Annual Meeting (pp. 1264-1268). Santa Monica, CA: Human Factors Society.

Ziegler, J. E., Hoppe, H. U. and Fahnrich, K. P. (1986). Learning and transfer for text and graphics

editing with a direct manipulation interface. In CHI ’86 Proceedings (pp. 72-77). New York: ACM.

Journal of the Washington Academy of Sciences,

Volume 80, Number 2, Pages 71-89, June 1990

The Effects of Rapid

Prototyping on User Behavior

in System Design

Daniel R. Sewell

Search Technology, Inc., Norcross, GA 30092

William B. Johnson

Galaxy Scientific Corporation, 2310 Parklake Drive,

Atlanta, GA 30345

ABSTRACT

Rapid prototyping is a design, development, and evaluation process that creates suc-

cessively refined hardware or software models representing the current conceptualization

of a product in design. Aeronautical and marine engineeers, for example, use engineering

prototypes of operator displays in simulators for aircraft or ship cockpit/control room

design. Architects use hardware models as partial prototypes of their designs. Today’s

software environments provide system designers with the ability to build system prototypes

with relative ease. This method allows system designers to build demonstrations that are

very effective for users’ formative evaluation. Rapid prototyping research and develop-

ment tools have a significant effect on how system designs are perceived by potential users

of the designs and the systems. Prototyping and its goals are defined and descriptions

are provided of user participation in system design, the impact of rapid prototyping on

user behavior, and examples of how rapid prototyping has been used to impact user

expectations. The examples focus on intelligent tutoring systems, a pilot-vehicle interface,

and graphical displays for maintenance problem solving.

Introduction

Prototyping is a design, development, and evaluation process used to create

hardware and/or software models that represent the current conceptualization

of a system or object design. Rapid prototyping refers to the development

process whereby several iterations of prototypes are generated, evaluated,

and revised before the final system is implemented. The terms prototyping

and rapid prototyping will be used interchangeably throughout the paper. This

paper will focus on prototyping as it relates to complex human-machine system

72 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

design. In particular it examines the effects rapid prototyping may have on

the behaviors of individuals involved in the design, development, and oper-

ation of a system. Such systems carry out functions and have an interface

through which a human may interact. These systems span a range of possi-

bilities including information systems, training systems, vehicle systems, pro-

cessing plant systems, and manufacturing systems.

For example, in vehicle systems, the U.S. Navy uses a full-scale mockup

of submarine command and control centers for the design and layout of the

complete physical environment. This includes everything from console and

other equipment placement to new user interfaces on the console screens

(Wallin, 1988). Similarly, the U.S. Air Force relies on rapid prototyping for

research related to advanced avionics systems (Duke, et al. , 1989). In advanced

avionics systems environments, control law concepts, alternate vehicle control

algorithms, pilot controls, and cockpit displays are implemented and tested

in a series of advanced-systems prototypes. In these and other such prototyping

environments, it is possible to iterate through many different configurations

for evaluation and subsequent design (for other examples see Cieslak, et al.,

1989; Hays, 1989; Lewis, et al., 1989; Saxena & Kaul, 1986).

These and other current development environments provide system de-

signers the ability to construct system prototypes with relative ease. These

prototypes may contain as much or as little of the purpose, function, and form

of the system as is desired for evaluation. The latest development methods

promote incremental design and development in an iterative process. These

methods allow today’s designers to build demonstrations that are very effective

for evaluation. Such rapid prototyping research and development tools have

a significant effect on how designs are perceived by everyone in the design

process, from on-line users through system maintainers and managers.

Rapid prototyping presents several prospects for the design and develop-

ment of complex systems. Designers are now able to provide concrete ex-

amples of their designs to users for feedback, to try out new design ideas with

minimal investment, to clarify their own ideas through successive prototype

iterations, and to involve end-users and managers in design; to name but a

few of the prospects (cf. Belardo & Karwan, 1986; Floyd, 1984; Stevens,

1983). Designers also are able to obtain, from themselves and others, eval-

uations of efficacy, efficiency, utility, and overall acceptability of their design

concepts before the designs or implementations become too expensive to

change (Cerveny, Garrity, & Sanders, 1986). In effect, prototyping provides

a fertile ground for shaping both the system under development and the

behaviors of all people involved in the design, development, and use of the

system. However, to make the most of this potential, it is necessary to develop

RAPID PROTOTYPING 73

a clear understanding of prototyping and its behavioral implications. The rest

of this paper examines prototyping as a methodology, discusses the users and

behaviors affected by prototyping, and provides examples of rapid prototyping

as it impacts user behaviors. The paper closes with a discussion of directions

for rapid prototyping and resulting behavioral change in the system devel-

Opment process.

General View of Prototyping

The following discussion examines literature on prototyping and integrates

this literature into a general framework defining goals of prototyping, kinds

of prototyping, individuals affected by prototyping, and the behaviors of these

individuals as they are affected by prototyping. Since the complex systems

under consideration are primarily software-driven, much of the literature ex-

amined is software-based work.

Goals of Prototyping

Each particular definition of prototyping and prototypes contains the theme

that a prototype is a model of the operational system. However, these and

other characterizations of prototyping (Carey & Mason, 1983; Harker, 1988;

and Jordan, et al., 1989) all modify the definition to align with the goal of the

prototype. For example, if an automobile manufacturer’s goal is to test con-

Sumer appeal then the prototype of a new automobile must be a fully oper-

ational vehicle that can be tested for qualities such as handling, aesthetics,

comfort, and power. Other automotive prototypes may place emphasis on

other factors such as manufacturability. From another perspective, if the goal

is to evaluate display utility then a prototype for evaluating a computerized

display for a nuclear power plant operator’s panel need only have the display

concept depicted. Such a prototype would attend less to interface software or

to simulations of the plant instrumentation systems. Consequently, we must

specify the goals of prototyping before we can define the different kinds of

prototyping and their effects on individuals.

There are three distinct views of rapid prototyping. These are discussed in

turn and then integrated. Bally, Brittain, & Wagner (1977) provide the earliest

and simplest view in their discussion of three goals for prototyping. One is to

increase user confidence in the system. Another is to increase the learnability

of the system. The third is to provide the user early experience with the system.

Developing a more comprehensive view, Floyd (1984) discusses several

goals which may be clustered into three primary goals that she also classifies

74 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

as kinds of prototyping. One goal is to explore ideas, examples of which are

clarifying requirements, developing new features, and structuring implemen-

tation. Another goal is experimentation, examples of which are determining

efficiency of the system or demonstrating technical feasibility. The third goal

is evolutionary adaptation which consists of a system adapting gradually to a

changing environment. This view incorporates that of Carey & Mason (1983)

who said the goals of prototyping are to improve the development of and

definitions of requirements.

Mosty recently, Verrjin-Stuart & Anzenhofer (1988) suggested the goals of

prototyping are:

. evaluate organizational impact

. establish requirements and data structures

. develop human-machine interface

. establish data definitions

. anticipate possible system changes

. determine operational efficiency

NM B WN FR

Their ideas are a general integration of several views. One view is Alavi’s

(1984) five goals which are to obtain user input and feedback, to increase user

commitment to the system, to promote relations among developers, operators,

and supporters, to increase likelihood of a “right”? system, and to clarify

requirements and functions. Another is Belardo & Karwan’s (1986) goals

which are to pique the user’s interest, to increase user satisfaction, and to

increase managerial dedication. The final view is Cerveny, Garrity, & Sanders’

(1986) which includes increased system quality, decreased resistance to change,

increased user commitment, increased sense of ownership, increased effective

system use, and increased user attitude toward the system.

Examination of these three views shows that there are two distinct and

possibly equally important goals: 1) Affect system design, or 2) Affect system

design users. Each of these has several subgoals. The goals and subgoals may

be active alone or in combination with one another. This combined view yields

six potential goals of rapid prototyping. The goals and subgoals are named

and characterized as follows:

1. Affect system design.

a. explore system design ideas—this includes establishing requirements, defining

data structures, identifying system functions, developing interface concepts, etc.

b. evaluate system design ideas—this incorporates experimentation and includes

determining functional utility, operational efficiency, technical feasibility, etc.

c. adapt system design ideas—this incorporates evolutionary adaptation and in-

cludes anticipating possible system changes, requirements changes, environ-

mental changes, etc.

RAPID PROTOTYPING 75

2. Affect system design users.

a. impact organization—this includes improved team participation during devel-

opment, increased user input and feedback, etc.

b. educate users to the design concept—includes increased user learnability, in-

creased experience with system, increased user interest, etc.

c. proselytize system users (and development participants )—this includes increased

user commitment, increased management commitment, increased sense of joint

ownership, increased likelihood of system being considered right, etc.

These goals account for all the activities described above and may be used as

the starting point for developing characterizations of prototyping, and the

behaviors affected.

Definition of Prototyping

The orientation of definitions of prototyping is guided explicitly by the goal

of affecting the system design. The definitions ignore the goal of affecting the

system design participants. Consequently, that goal will not be addressed until

the next section of the paper.

Floyd (1984) defines prototyping as a ‘“‘well defined phase in the production

process, where a model is produced in advance, exhibiting all the essential

features of the final product, for use as test specimen and guide for further

production”. Morrison (1988) borrows from Boar’s (1984) definition calling

prototyping a “method for extracting, presenting, and refining a user’s needs

by building a working system. By increasingly refining (the prototype), as

problems are uncovered and solutions emerge, prototyping can efficiently

solve the definition problem’’. Tozer (1987) uses examples, saying that pro-

totypes can be screen or report mockups, simulations, or a complete model

of the final software system. Finally, Tanik and Yeh (1989) call prototyping

a “process of developing a scaled-down version of a system to use in building

a full-scale system. ... The final product of the prototyping activity is a

working model that can be used for many purposes, such as requirement

validation, feasibility study of a complex system, behavioral specification of

a system, and functional specification of a system design’.

Extending and integrating this earlier work, Luqi (1989) defines prototyping

as the process of creating one, or a series, of concrete executable models of

selected aspects of a proposed system. The model is created as part of a larger

design process which includes requirements specification followed by design,

then prototype development followed by user validation. This is an interleaved

process in which traditional activities such as requirements definition and

functional decomposition lead to the development of an initial version of the

system which will be evaluated and refined. The versions or kinds of prototypes

may take on a number of forms including mockups, simulation, or complete

models of the final system.

76 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

Kinds of Prototypes

There appears to be a common theme among the different kinds of pro-

totypes available to developers. The theme, in general, is that kinds of pro-

totypes are differentiated by the extent to which they embody or clarify the

purpose, function, and/or form of a system under development. In a widely

used scheme, Carey and Mason (1983) suggest three categories of prototypes:

1. scenario or simulation prototypes

2. demonstration system prototypes

3. Version 0 or working version prototypes

The first kind produces a scenario or simulation. The prototype only simulates

the software system without the application logic. This embodies the purpose

and some function of the system. The second kind produces a demonstration

system. The prototype includes the user interface with enough background

application logic to make the system work. This embodies the purpose, func-

tion, and some form of the system. The final kind of prototype is producing

a Version 0 prototype, which 1s the first working release of the software. This

embodies the complete purpose, function, and form of the system. Therefore,

alternative descriptions of prototypes can be organized according to what is

embodied in the prototype and then described in finer detail.

. purpose only prototype—requirements list, statement of need, etc.

. function only prototype—conceptual design

. form only prototype—static mockup

. function and purpose prototype—scenario, storyboard characterization

. form and purpose prototype—models, simulations

. form and function prototype—models, simulation

. form, function, and purpose prototype—working version, Version 0

NYDN NBWN FR

If the prototype embodies system purpose only, then a requirements document

or a statement of need serves as the prototype. If the prototype embodies

system function (and may include purpose) then a scenario or storyboard

serves as the prototype. If the prototype embodies system form (and may

include purpose and/or function) then a model, working prototype, or op-

erational system serves as the prototype. Each of these is a candidate for early

iteration and refinement with users and other design participants. Conse-

quently, each can be affected by the participant; and, each can affect the

participant.

Summary

Based on existing work, we have defined prototyping, its goals, and its

kinds. In short there are goals related specifically to affecting system design

and to affecting system design users. These goals may be pursued by the

RAPID PROTOTYPING 77

development and iteration through several kinds of prototypes which vary in

the degree to which they demonstrate some combination of purpose, function,

and form. Each kind of prototype can be expected to have different effects

on the system design and the behavior of system design users depending on

who is being affected and what goals are being sought. The focus of the rest

of the paper is on those effects—what they are, who they occur to, and what

behaviors they impact.

Effects of Prototyping

Positive & Negative Effects of Prototyping

Prototyping has both positive and negative effects. There has been some

work delineating these effects. Morrison (1988) has focused on the effects of

prototyping on an artifact under development. He concludes good effects are:

1) prototypes provide immediate impact on design due to tight feedback loop,

2) dynamic interactions and development are gained, and 3) development of

a working deliverable. Bad effects fall into a general category of seducing the

developer into thinking a design is good when there are still problems. Some

bad effects of prototyping that produce this result are that prototypes conceal

system structure, tend to maintain the status quo, defer full implementation,

hide exception handling, and hide complex manipulations. —

Alavi (1984) has focused on the effects of prototypes on users during the

development process. He concludes that good effects are increased commit-

ment among users, better relations among users, and increased likelihood of

the produced system being accepted. At the same time bad effects are possibly

overselling to yield unrealistic expectations, losing management control and

losing user enthusiasm (as prototypes fail or disappoint).

Bally, Brittain, & Wagner (1977) focused on the users and the overall

process. They found good effects are increased user confidence in the resulting

system and early learning about the system. The negative effect is that the

method can be very expensive as clear development goals can get lost. Along

these lines, the system may never be completed and the “‘best”’ possible design

may never be reached.

Combining across these and other studies shows prototyping, as a system

design method, produces three general, positive effects. One is to increase

communication between the system user and the designer. The increased

communication will result in clearer definitions of requirements and specifi-

cations. The ultimate benefit is lower development and modification costs

because extensive rework is avoided. Increased communication also promotes

78 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

closer working relationships and enhanced commitment among the users in

the development process. The relationships and commitment can result in a

greater sense of ownership in the system developed. 5

A second general, positive effect is to compel a tighter feedback loop during

the development process. This feedback may be among some or all participants

in the development process and results in immediate design impact both from

and on the participant. Such an impact may yield improved design charac-

teristics and, consequently, an increased likelihood of producing a good sys-

tem. The increased feedback may also produce a sense of ownership from the

participants thereby resulting in greater commitment.

Finally, prototyping has the promise of providing a working example of the

system during development. Providing this example for various users will result

in a working deliverable that may be studied, evaluated, demonstrated, and

possibly carried over to production. Study and evaluation provides users op-

portunity to learn. Demonstration helps to keep management committed to

the development and may provide completion of contractual requirements.

Any carryover from prototypes may reduce time and cost for final production.

Prototyping has three general, potentially negative effects. One is the po-

tential for failing to meet expectations. When a prototype system is applied

with a naive user population there is often the perception that the prototype

is the final product. Sometimes such users cannot understand why there is

such a long delay between the prototype, which seems to work, and the

completed system, that looks just like the prototype. This problem is exac-

erbated when the user perceives that the prototype, which appears real, was

completed for a very small portion of the available funding resources. It begs

such questions as ‘‘you appear to be quite far along, are you sure that you

need all that remaining time to complete the system’’?

Related is the problem of using prototypes with naive users and the resultant

feedback. Users in the earliest stage of formative evaluation may feel that the

prototype system is completed. Therefore, they feel that it is not subject to

modification and may develop a negative attitude toward the final system

based on the prototype. The other danger is that such users cannot differentiate

between the prototype and the finished product. Consequently, they may not

recognize or believe that a finished product is no longer a prototype which

may be changed. The result would be that users may try to suggest major

changes to the finished product and be disappointed when such changes are

not forthcoming.

Another negative aspect of prototyping is that it has the potential to lead

to incomplete development. Incomplete development may be exhibited as de-

RAPID PROTOTYPING 79

ferring full implementation in favor of prototypes or never being able to

complete the implementation. During this process development goals may be

lost and the management of the development process may go out of control.

Consequently, there can be numerous add-on effects related to, or resulting

from, incomplete development thereby making recovery exceptionally diffi-

cult.

The final negative aspect of prototyping is its potential for producing un-

satisfactory designs. There may be tame designs which maintain the status

quo. These may be ad hoc designs which oversimplify problems and/or ignore

problems thereby preventing reaching the “‘best’’ design. These may be unclear

designs which conceal system structure, hide exception handling, and hide

complex manipulations from the developer thereby preventing adequate eval-

uation before final development and delivery.

The effects discussed above focus on the positive and negative effects of

prototyping on the product or artifact being created. This means that the work

from which these were drawn was aimed at the goal of affecting the design

instead of the goal of affecting users. We now turn to the problem of examining

users and their behaviors as affected by prototyping.

Users Affected by Prototyping

Users consist of all people who impact or are impacted by the system under

development. The following classes of users and general activities of users are

an extension of Morrison (1988) and Rockart & Flannery (1983) and are

further described below.

. sponsor—decisions about initiating and continuing system development.

. Manager—requires system outputs and controls operator activities.

. Operator—carries out all system functions.

. supporter—provides all training and maintenance for system.

. developer—designs and implements complete system.

Wn B&B WN rR

The following descriptions characterize the level at which each of these users

might respond to system development.

Sponsors have the power to authorize expenditures for and require speci-

fications of a system to be developed. According to Rockart & Flannery (1983)

these are indirect users who understand the purpose of the system and may

only need peripheral outputs that serve those ends. This means they are

possibly affected by, most interested in, or think about the system in terms

of its purpose. Consequently, they may respond well to prototypes embodying

system purpose.

Managers have the power to require outputs from a system to be developed

80 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

and need to understand the limits, capabilities, and interface to that system

when developed. According to Rockart & Flannery (1983) these are inter-

mediate users who understand the system function well enough to specify

direct outputs that others will have to produce. This means they are possibly

affected by, most interested in, or think about the system in terms of its

function. Consequently, they may respond well to prototypes embodying sys-

tem function. |

Operators have to handle all input and output for the system to be developed

and require training and understanding of the interface and all its relationships.

According to Rockart & Flannery (1983) these are direct users who understand

the form of the system (i.e., the look and feel of the interface) but may or

may not understand the function and purpose. This means they are possibly

affected by, most interested, or think about the system in terms of its form.

It also means they may not respond well to prototypes based purely on purpose

or function.

Supporters have to provide all training for and/or maintenance of the system

to be developed. According to the Rockart & Flannery (1983) scheme these

individuals must be both direct and intermediate users who understand the

system form and function well enough to specify direct outputs that others

will have to produce; and, to teach others how to produce it. This means they

are possibly affected by, most interested in, or think about the system in terms

of both its form and its function. Consequently, they may respond well to

prototypes embodying system form and/or function. This reasoning leads to

the conclusion that individuals who must support systems might be strong

candidates to use during the design process since they may have a broader

range of understanding of the system than anyone else.

Developers have to design and produce the system to be developed. They

are required to understand the support, operation, and management of the

system. Ideally, these individuals should think about the system in terms of

its purpose, function, and form. In addition, it is desirable that their thinking

be along the same lines as the other four classes of users in order to ensure

a match between the developer’s efforts and other users’ needs.

Behaviors Affected by Prototyping

A review of the goals sought through prototyping, the effects of prototyping,

and previous explorations of prototyping yields 14 specific user behaviors

which appear to be affected by prototyping. There is not much difference

among these behaviors so they will organize into more general behavioral

RAPID PROTOTYPING 81

classes for further discussion. Twelve behaviors are those carried out by users

with respect to the system and two are those carried out with respect to other

users. The behaviors are listed below:

1. With respect to the system:

. learning about (Bally, Brittain, & Wagner, 1977; Stevens, 1983)

. operation of (Harker, 1988)

. understanding of (Stevens, 1983)

. commitment to (Alavi, 1984; Belardo & Karwan, 1986; Cerveny, Gerrity, &

Sanders, 1986)

. confidence about (Bally, Brittain, & Wagner, 1977)

. interest in (Belardo & Karwan, 1986)

. involvement in (Belardo & Karwan, 1986)

. Satisfaction with (Belardo & Karwan, 1986; livari & Kayalainer, 1989)

acceptance of (Cerveny, Garrity, & Sanders, 1986)

. ownership of (Cerveny, Garrity, & Sanders, 1986)

. expectations of (Harker, 1988)

attitude about (Cerveny, Garrity, & Sanders, 1986)

2. With respect to other individuals:

a. communication among (Alavi, 1984; Floyd, 1984)

b. relationships among (Alavi, 1984)

aAo7 &

ne Ben eere

It appears that communication and relationships among individuals are more

general classes of behavior comprised of many specific behaviors. In a similar

way, the 12 behaviors engaged in with respect to the system can be organized

into three general sets of behaviors for further discussion. These are under-

standing of system, commitment to system, and attitude toward system.

Understanding of the system is comprised of the behaviors indicating how

well users can learn about the system, explain the system, teach about the

system, and/or operate the system. Commitment to the system consists of

dedication to system development, interest in system development, and in-

volvement in system development. Attitude toward the system is comprised

of the users’ acceptance of the system, confidence in the system, satisfaction

with the system, ownership feeling toward the system, and expectations of

the system.

The behaviors just described may also be characterized as psychological

States. If viewed as such, then one job of the rapid prototype is to try to affect

the psychological states such that they are positively oriented toward the

system under development. Even if the prototype developer does not attempt

to actively change these states, he or she must be aware that the behavior or

State of the users ultimately determines when system development is com-

pleted. Consequently, the prototype developer should at least be aware of

the states and try to minimize any potentially negative impacts of their work.

82 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

Summary

Based on existing work, we have defined the effects of prototyping, the

users who are most concerned, and what behaviors they impact. In short there

are three general positive and three general negative effects that may be

achieved with prototyping. These effects may be achieved with sponsors,

managers, operators, supporters, and/or developers of systems. The effects

impact their understanding of, commitment to, and/or attitude toward the

system. To understand and/or evaluate a prototyping effort from a behavioral

perspective (as opposed to a system design perspective) one must examine

each of these factors explicitly. Furthermore, to plan for a prototyping effort

that will be integrated with a system design effort one should also examine

the goals and kinds of prototyping available to the developer. This would

allow the developer to determine what kinds of prototyping would achieve

particular effects therefore allowing the developer to better plan his or her

efforts.

Examples of Behavioral Effects in Prototyping Applications

The remainder of this paper describes a variety of research and development

applications in which rapid prototyping was used to design and develop a

software application. Three distinctly different applications are used. Each

application was worked on by either one or both authors. The first set of

examples is based on experience in the design and development of intelligent

tutoring systems for technical training. The second example is related to the

design of a pilot vehicle interface for tactical aircraft. The third is related to

the development of concepts and principles for electronic presentation of

graphical information to maintenance technicians.

Prototyping for Intelligent Tutoring Systems

The primary purpose of the prototypes described here is to increase the

communication between the ultimate system user and the designer of the

software system. The prototype has the primary goal of insuring that system

design is matched to user expectations. The prototypes described here prevent

the unpleasant “‘surprises” that can be associated with system design.

Intelligent tutoring systems (ITSs) are computer-based instructional systems

that capitalize on artificial intelligence technology to deliver training in a

variety of applications. ITSs are characterized by having independent models

of a system expert and pedagogical expert along with a dynamic model of the

student. Descriptions of ITSs are treated elsewhere (Polson & Richardson,

1988; Psotka, et al., 1988; Wegner, 1987). Johnson (1988a,b,c) has argued

RAPID PROTOTYPING 83

that the multi-disciplinary team that must work together to design and build

ITSs must work very closely with the end users. Such personnel are technical

instructors, instructional developers, job incumbents, students, and managers

of technical training. Prototyping is the ideal method to ensure clear com-

munication among the parties that must be involved in design and development

of an ITS. This section will describe how prototyping has been used to insure

that the finished ITS is an efficient and effective addition to an operational

training program.

The examples used here draw from two experiences. The first is a project

entitled, Microcomputer Intelligence for Technical Training (MITT). The sec-

ond project is entitled, Advanced Learning for Mobile Subscriber Equipment

(ALM).

Microcomputer Intelligence for Technical Training

MITT is a project that was completed in cooperation with the Air Force

Human Resources Laboratory and NASA Johnson Space Center. The project

was scheduled to build a prototype training system in a relatively short time.

The short development cycle dictated that all parties must have a very clear

understanding of the development goal at the start of the project. MITT was

envisioned to be a new generation of an evolved approach to computer-based

diagnostic training (Johnson, 1987). Therefore, one of the older systems called

DGSIM (Johnson, et al. , 1986) was used as a vehicle to help the subject matter

experts to understand what could be done. DGSIM was not a prototype, as

defined earlier in this paper. However, it did serve to show the NASA in-

structors and subject matter experts ways that the new MITT ITS could be

structured. The behavior of the instructors changed as they became active

designers of the new system rather than passive subject matter experts that

merely provided technical information as requested by the ITS scientific staff.

The first MITT prototype was completed approximately four months after

the technical domain was identified. This prototype was complete, fitting the

definition of Tanik and Yeh (1989) and the more sophisticated level of pro-

totype offered by Tozer (1987). MITT was complete in program logic and in

terms of user interface. The first MITT prototype was an example of what

Carey and Mason (1983) called ‘‘Version 0”’.

The prototype of MITT was evaluated for acceptance by astronauts, flight

controllers, and instructors at Johnson Space Center. Due to the sophistication

of the users, it was appropriate that the prototype system be as complete as

possible. Because the prototype was complete, it was very useful in obtaining

a substantive list of specifications for subsequent versions of MITT.

84 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

Prototype for MITT Writer

MITT Writer is an authoring system that will permit technical training

personnel to build MITT intelligent tutoring systems without using program-

ming languages. As with the MITT system, a prototype was used to dem-

onstrate the capabilities of the envisioned system. Since this prototype was

used only to demonstrate and clarify the requirements of a complete system,

it was only a scenario or simulation of the planned system. The simulation

demonstrated the interface and functionality of the system in design. Such a

prototype can also be called a user interface prototype.

Advanced Learning for Mobile Subscriber Equipment

ALM is the ITS for the Mobile Subscriber Equipment (MSE). MSE is the

largest electronic equipment acquisition in the history of the U.S. Army. When

fully fielded in 1993, it will replace nearly all the tactical (i.e., front-line)

communications radios and telephones in the Army. During the transition

period, from 1989 through 1993, the Army must have soldiers prepared to

install and support the new MSE as well as all equipment in the current

inventory. This presents a sizable training challenge for not only the active

Army but also for such units as the Army Reserve and National Guard.

The rapid prototyping method was used on this project to accomplish a

variety of goals. First, a simulation/scenario prototype was designed to show

training system managers and General officers what an intelligent tutoring

system for MSE might look like. Using the development environment afforded

on the Apple MacIntosh with Supercard™, the development team was able

to build a MSE simulation with less than a person-month effort. This system

is called Advanced Learning for MSE (ALM) because it is designed to provide

recurrent training and practice to personnel who have received an introductory

MSE course. The prototype was instrumental for making the plan to develop

a completed system. The prototype permitted the customer to see what was

being proposed. Therefore, they had a concrete idea of what the finished

product would look like.

ALM was transported from the Macintosh-based prototype to a Version 0

of ALM on the Electronic Information Delivery System (EIDS), an 80286

computer that is in abundance in the U.S. Army. This Version 0 is in user

acceptance evaluation before the fully operational training system is delivered.

The use of prototyping on the ALM research and development has ensured

that the finished product will evolve to meet the expectations of the Army

customer.

RAPID PROTOTYPING 85

Prototyping for Pilot’s Associate Pilot-Vehicle Interface

The pilot-vehicle interface for the pilot’s associate (PA-PVI) refers to the

human-machine interface between the pilot and the aircraft in an as-yet unbuilt

jet fighter aircraft. The architectures and implementations of elements of the

system have been described elsewhere (Andes, 1987; Hammer & Geddes,

1987; Howard, Hammer, and & Geddes, 1988; Rouse, Geddes, & Curry,

1987). The research and development of the PA-PVI is part of a larger program

to research and develop a distributed intelligent system that will reside in the

avionics of the aircraft. This project serves to show the role which rapid

prototyping can play when it has been an explicit part of the system design

process from its inception.

The general goals of using prototyping in developing the PA-PVI were to

affect both system design and system users. The primary stated goal was to

convince the sponsors of the feasibility of the technologies and concepts in-

volved. In turn this was expected to lead to greater commitment from the

sponsors. The goal with respect to the operators was to develop a positive

attitude, some understanding, and willingness to accept new ideas. Other goals

were to evaluate system functionality, enhance communication among a geo-

graphically distributed development team, and to promote a deeper, shared

understanding among developers working on different components of the

system.

After an initial architecture for the system had been developed and while

the first hardware/software prototype was being developed, an initial proto-

type evaluation was planned. In this evaluation, two scenario based mission

prototypes were developed. One had all system functions, pilot actions, and

interface displays described for the PA-PVI. The other had all system func-

tions, pilot actions, and interface displays described for a current jet. These

were presented to pilots for comparative performance evaluation and the

results are described in Sewell, Geddes, & Rouse (1987).

After this purpose- and function-level prototype evaluation a series of com-

puter-based functional prototypes were developed and presented to sponsors,

potential operators, and the complete development team for review. As the

system evolved through these prototyping cycles, a low fidelity simulator was

brought into the plan. This provided a medium in which to develop form-level

prototypes that could be used for systematic evaluation by potential system

users. Current plans include redeveloping the system in high fidelity simula-

tions for formal evaluations as a Version 0 prototype.

Rapid prototyping in the PA-PVI has been successful in several respects.

Many of the goals have been achieved. Sponsors have demonstrated greater

86 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

commitment to the system by asking for continued development past research

and initial development. New sponsors have demonstrated commitment by

asking for the technology to be applied to other domains. Pilots have begun

to demonstrate understanding and acceptance of the system. Some have served

in advisory/testing roles and some have moved into research and development

jobs to help further development. Developers have been able to use the

prototypes for evaluation and as a medium for increased communication.

As a result of integration and evaluation efforts, significant redesign and

redevelopment has occurred to date. At the same time, it is only through

intense communication that redesigns among the different components could

and were brought into compatibility. This also raises the only significant neg-

ative effect. Prototype integration was infrequent enough to allow divergence

in development. This divergence was sometimes subtle and often required

tremendous efforts to reconcile. This could, perhaps, have been avoided through

traditional design practices using complete decomposition and design before

development; or, through more frequent prototype integration.

Prototyping for Graphical Display Presentation for Maintenance

For maintenance problem-solving the development of graphical displays

that can be shown on small computer-based display surfaces as an alternative

to blueprint-sized hardcopy is a high priority. Currently, the manuals for

maintaining complex systems often make up thousands of pages which must

be revised and updated in addition to being used by maintenance technicians

in their everyday work. A research and development project to compose and

evaluate principles which will drive this display development is currently un-

derway. The early research and development has been reported elsewhere

(Sewell, Rouse, & Johnson, 1989). The work will transition to system design

and development in future projects. For this paper, this project serves to show

how rapid prototyping can play a role even in the most preliminary stages of

system development—in this case during early research.

There were two general goals of the prototyping effort involved in this

project. One was to affect the system design through the development of

principles and displays for evaluation. It was only through prototypes that

these could be evaluated. The other was to affect the potential system op-

erators (maintainers) in two ways. One desired effect was to have them accept

the researchers in their environment. The other was to have them accept the

researchers as having potential to benefit maintenance jobs.

The goal of being accepted by the maintenance technicians was probably

the more important goal. It certainly took precedence since achieving all other

goals depended on this one being accomplished. The initial response of the

RAPID PROTOTYPING 87

maintainers was to tell the researchers that they “already had what they

needed”, they “‘had previous, failed attempts sitting unused at their work

site’’, and so forth. In response, the researchers developed prototype com-

puter-based graphical displays to demonstrate some of the possibilities of their

approach. Based partially on this prototype, the maintainers enthusiastically

consented to participate in the research.

The goal of affecting system design by developing principles from which to

generate elements of a new system that is a complete departure from existing

maintenance systems is a lofty goal. Unfortunately, it is one that the main-

tainers find difficult to relate to. They are forced to work with what they have

and have little time to devote to exploring underlying issues. Yet, it is necessary

to extract information from the maintainers in the form of existing knowledge

and feedback on ideas that are generated. Since these maintainers find it

difficult to evaluate the new ideas in the form of purposes, requirements, and

principles, researchers must develop concrete examples embodying those pur-

poses and principles.

The researchers generated paper-based sets of prototype displays (sized for

computer screens) from initial requirements and principles. These displays

were then used by the maintainers as they talked through solving a mainte-

nance diagnosis problem. For each prototype display the maintainer was probed

for underlying reasons why it was good or bad for the maintainer’s activity.

The results from these sessions provided much information that was used in

the continued development of display principles.

In addition to changes in the design ideas, the researchers were seeking to

increase the maintainers’ confidence that a system could be built eventually

and to generate realistic expectations about what kind of system might be

built. This was much more difficult and only met with mixed sessions. The

maintainers held strong opinions about possible technology, about the com-

plexity of their work, and about the capabilities of non-maintainers. These

opinions were generally negative with respect to the future of the desired

system development. By the end of the current research, these maintainers

felt positive about the capabilities of the researchers and the prospect of

developing useful materials. However, they still felt negative about the tech-

nology required for such an effort and about the prospect that the current

work would eventually be turned over to other designers/developers who were

not acceptable to the maintainers.

As a result of this early prototyping and research effort, the designers and

the users appear to have moved closer to a common view of what might be

developed to support the maintainers even though they still differ on the

feasibility of such a system. Future efforts in the research will be to develop

88 DANIEL R. SEWELL AND WILLIAM B. JOHNSON

computer-based prototype implementations of principle-driven displays for

experimentation. These should also provide the opportunity for further impact

on future design concepts and to make inroads with the maintainers to show

that the technology is feasible.

Conclusion

Prototyping is a design, development, and evaluation process that creates

hardware or software models that represent current conceptualization of the

products in design. We have defined prototyping by its goals, kinds, and

behavioral effects. In addition, we have presented specific application ex-

amples examined in the framework provided by the definition.

The most important elements are the extensions of prototyping to consider

explicitly the behavior of the different participants in the system-design process

and the impact of prototyping on that behavior. After all, the evaluation and

fate of any newly developed system is determined, not solely by the charac-

teristics and qualities of the system, but also by the behavior that system users

demonstrate toward the system. To define those behaviors and the factors

that affect them is to make them available for manipulation. Future system

developers should take into account these potential effects when planning

prototyping activities.

References

Alavi, M. (1984). An assessment of the prototyping approach to information systems development. Com-

munications of the ACM, 27(6), 556-563.

Andes, R. C. (1987). Adaptive aiding in complex systems: An implementation. Presented at the IEEE

Systems, Man, and Cybernetics Society Conference, Alexandria, VA.

Bally, L., Brittain, J. and Wagner, K. H. (1977). A prototype approach to information system design and

development. /nformation and Management, 1, 21-26.

Belardo, S. and Karwan, K. R. (1986). The development of a disaster management support system through

prototyping. Jnformation and Management, 10, 93-102.

Boar, B. H. (1984). Application Prototyping. New York: John Wiley and Sons, Inc.

Carey, T. and Mason, R. E. (1983). Information system prototyping: Techniques, tools, and methodologies.

Information, 21(3), 177-191.

Cerveny, R. P., Garrity, E. J. and Sanders, G. L. (1986). The application of prototyping to system

development: A rationale and model. Journal of Management Information Systems, 3(2), 51-62.

Cieslak, R., Fawaz, A., Sach, S., Varaiya, P. and Walrand, J. (1989). The programmable network pro-

totyping system. Computer, 22(5), 67-76.

Duke, E. L., Brumbaugh, R. W. and Disbrow, J. D. (1989). A rapid prototyping facility for flight research

in advanced systems concepts. JEEE Computer, 22(5), 61-66.

Floyd, C. (1984). A systematic look at prototyping. In R. Budde, K. Kuhlenkamp, L. Mathiassen and H.

Zullighaven (Eds.), Approaches to Prototyping (pp. 1-18). Berlin: Springer-Verlag.

Hammer, J. M. and Geddes, N.D. (1987). Design of an intelligent monitor for human error in a complex

system. Presented at the 1987 American Institute of Aeronautics & Astronautics Computers in Aerospace

Conference, Wakefield, MA.

Harker, S. (1988). The use of prototyping and simulation in the development of large-scale applications.

The Computer Journal, 31(5), 420-425.

Hays, N. (1989). Biosphere II—A prototype for the future. Computer, 22(5), 11.

RAPID PROTOTYPING 89

Howard, C. W., Hammer, J. M. and Geddes, N. D. (1988). Information display management in a pilot’s

associate. Proceedings of the Aerospace Applications of Artificial Intelligence Conference, 1, 339-348.

livari, J. and Kayalainer, M. (1989). Impact of prototyping on user information satisfaction during the IS

specification phase. Information and Management, 17, 31-45.

Johnson, W. B., Wiederholt, B. J. and Maddox, M. E. (1986). Diagnostic training demonstration: Instructor

and student manuals and sample diskettes (EPRI NP-4493P). Palo Alto, CA: Electric Power Research

Institute.

Johnson, W. B. (1987). Development and evaluation of simulation-oriented computer-based instruction

for diagnostic training. In W. B. Rouse (Ed.), Advances in Man-Machine Systems Research: Vol. 3 (pp.

99-127). Greenwich, CT: JAI Press, Inc.

Johnson, W. B. (1988). Developing expert system knowledge bases for technical training, In L. D. Massey,

J. Psotka and S. A. Mutter (Eds.), Intelligent Tutoring Systems: Lessons Learned (pp. 83-92). Hillsdale,

NJ: Lawrence Erlbaum Associates, Inc. (a)

Johnson, W. B. (1988). Intelligent tutoring systems: If they are such good ideas, why aren’t there more

of them? Proceedings of the Tenth Annual Interservice Industry Training Systems Conference, 399-406.

(b)

Johnson, W. B. (1988). Pragmatic consideration in development and implementation of intelligent tutoring

systems. In J. R. Richardson and M. C. Polson (Eds.), Foundations of Intelligent Tutoring Systems (pp.

189-205). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. (c)

Jordan, P., Keller, K. Tucker, R. and Vogel, D. (1989). Software storming: Combining rapid prototyping

and knowledge engineering, JEEE Computer, 22(5), 39-48.

Lewis, T. G., Handloser, F., Bose, S. and Yang, S. (1989). Prototypes from standard user interface

management systems. Computer, 22(5), 51-60.

Lugi. (1989). Software evolution through rapid prototyping. JEEE Computer, 22(5), 13-25.

Morrison, W. (1988). Communications with users during system development. Software Technology, 30,

295-298.

Polson, M. C. and Richardson, J. J. (1988). Foundations of Intelligent Tutoring Systems. Hillsdale, NJ:

Lawrence Erlbaum Associates, Inc.

Psotka, M. C., Massey, R. and Mutter, S. (1988). Intelligent Tutoring Systems: Lessons Learned. Hillsdale,

NJ: Lawrence Erlbaum Associates, Inc.

Rockart, J. F. and Flannery, L. S. (1983). The management of end user computing. Communications of

the ACM, 26, 766-784.

Rouse, W. B., Geddes, N. D. and Curry, R. E. (1987). An architecture for intelligent interfaces: An

outline of an approach to supporting operators of complex systems. Human-Computer Interaction, 3.

Saxena, K. B. C. and Kaul, M. (1986). A conceptual architecture for DSS generation. Information and

Management, 10, 149-157.

Sewell, D. R., Geddes, N. D. and Rouse, W. B. (1987). Initial evaluation of an intelligent interface for

operators of complex system. In G. Salvendy (Ed.), Cognitive Engineering in the Design of Human-

Computer Interaction and Expert Systems. Amsterdam: Elsevier.

Sewell, D. R., Rouse, W. B. and Johnson, W. B. (1989). Initial evaluation of principles for graphical

displays in maintenance problem solving. (Technical Report No. ST-TR-8817-001). Norcross, GA: Search

Technologies, Inc.

Stevens, G. C. (1983). User-friendly computer systems? A critical examination of the concept. Behavior

and Information Technology, 2, 3-16.

Tanik, M. M. and Yeh, R. T. (1989). Rapid prototyping in software development. JEEE Computer, 22(5),

9-10.

Tozer, J. E. (1987). Prototyping as a system development methodology: Opportunities and pitfalls. Jn-

formation and Software Technology, 29(5), 265-269.

Verrjin-Stuart, A. and Anzenhofer, K. (1988). Information systems user-designer communication problems.

Information and Management, 14, 133-142.

Wallin, C. R. (1988). Concept of operations experiment (COOPEX 88-1) (Tech Memo TM-88-2194).

Newport, RI: Naval Underwater Systems Center.

Wenger, E. (1987). Artificial Intelligence Tutoring Systems. Los Altos, CA: Morgan Kaufmann Publishers.

AUTHORS’ NOTES

Preparation of this paper was partly supported by the Office of Naval Technology and the Naval Training

Systems Center under Contract No. N0001489C0047. We thank Bob Glushko for his review of and helpful

suggestions for this paper.

The views expressed about the example projects where prototyping took place reflect the opinions and

understanding of the authors. They are not meant to reflect the opinions or understanding of the sponsors

of or other participants in those projects.

Daniel Sewell’s currrent address is 818 Springdale Rd., Atlanta, GA 30306.

ca feats aiinlierias abi |

owebs seakcisolp

“WCeTOUTL at baad reshleyeacy bonwrsiyi- seclbabinate' By, ‘iia Ere isetalite oa ee

19) 4 Vo “Apne Ph arena se, aol Ae aah ave, mir Sicha 4 nly 2 pias vt 6 ow ol ae r

Me ? 3 EE deo LAL TO doh

eeiiukdhs | nt syoinign | setbst gt ee ) ago mee ee mea PAGND } atest reo €

ADR (OP-Fe om paryo.) late fe mits ney gastaly Pe LAL ft SE) 490 ie as a

BPmv 44 7 nel rt Pedursee ni “Lote hcaitlh pase aul

yee brad?” Mast whe Ans i x Halie ty 7 Cee Haare ip ae i aa ie

TMS DOLL > STEN Bea 2 Watiiowls askob gia Harrys! SUN |

fe a , . 2} o Bisa, a

PPO. Loeb Haat ed i rath Gai Dns sig Weak 3 pee advyitstens ate :.

i 9 serianitin oaehthast ton doud saan 84. 58 by RR aes eae} ek bene odd at: ne

biel ‘ersinsel ashok toh caias, Padacetall a) St aa he sna En aS Pate.

WORT rote Wit ante Seite th yg Sate OH Batge Mesh St Sa ead om,

ie sa)

[he most iraportant elem 2 ile ixamerapees hy ECM 54

ator! ety bvaieety mor! 4 ip cOFT ‘Gai i ir eyeet tim 2 “see a

Pi : n

, vg ak war es Fi ef «

Fr a Na Shs Se” tga bea rit eee By at ae

RS . ; Gore bt ah hd PT ihe ot © ‘ip wg te Dee. git: ganns ppt ahaa ¥ ay

(Pe MASONS Wve WTO, RD od ater & % ah Ra al UH aeilesi esDG >) SBS

‘ i > | Z J x 2 P + Seek ms Ru as a ‘ < F te te ey ,

4 or V4 feia? badd aye oe Aa) 2 wi es hia ap, Laas grevh of CHL A opeienpe

per | , ae Septem

CAORDAIEHS Usihid Sr uneiaMk, baee t eat ell ance et oR Hl

ths . ra pei » t ie =e vial Me mh vaioere fag 7

We Liat se is th 4 if gx) vote Ste goih apy any Ad a & q, ee L

aA. tape byes, ta jails) i 0) mot onetite dA ime , aR pew at a « e -

a Mahi WeuiyetesS- Amare) 2ervatiogk TREYB INO Puln ot: yeiegaus at toe

Pan eoteenote | .nolinrenon CAO aot peat AK Falta Bey pel (OOTY OF dana Sine Ze

oor. Ont +

16! acagel miagdlaloi mito. cotuders tei: hes) og ware F _ . soba

radteit) Wa toy) alt 2c PINGRA AS SNE ee ied Viowrely HORN re

manip og stages om a

io Pre * ' ‘ ‘ t ty } Bo A ae, 4.

Wy 6) Ney > mana Loser Nese Je € " a apceunl, ity “a e

2, iste iid, cope a IT) Tada ne mA i a? FH Ati Hh : 3 ahs Ae Oy, ONG ey

wey

‘ he , < j t i be iat . oh : Th ‘ ee , 7“ wete

etary | sb nay 2 pl aah Rha, a see ede ‘oon SE. Kod nth nalts , Pe mal Glen cs

i 1?

~~ a NSN Git Dh

(¢ Oh Stree Mae, packin: att aemgta ae gts bagels Rr Ierynnng fihegn Sh (UH FS 3 a fs or)

a bid ara (pulngtee 2, Sh Bae 43 ae i ms gaiguongr’ «Tet t

Ny "le ge ie oe aia ‘yuss tae

soho 158M cites Lage weattaciye elit Ti? iP ¥ valodaesnd berate: hy

i )4 \ *

as Iedixs Agta: rye! tat) Bay

ore

’ is ie Sele ee i r 3 irs aber x

d i! Sy Gairite ri ty fi % 5 a “;

tA a Adel ort Troi deed ie Ve aus NUNS ae heii ‘aes Savy ge, a eee

ar ie

Ly 8 a 4, ew ing OT ee HA &, gat sab btiah Serta per

fetgyy grained pagiotd sh) PL an whasay. Bent ft’ Sorvulaned ber

Be cower, OW, ead Urine a eee Wins crepgmngl Tag sily Me ¢

| A wphtems (onmere. (FEC Cogiingat aomtne |

ei : i) . ie es i) iis at ty'e cy?

Jub i * meet gl

' ‘

a0 sinks hie eM ods Die yr onaae jolaeelll ae .

iqlotrbes le. ata} patel Ones | Asqira WA FOG

MPN ish} SOG, TRY EWE i pies maven! uf, het? useley ab Actegt tice

tutg gsi nie oats esfian, apain, tout ‘aniqurotone smh Hoofony auiexe Rees wwii The

Megat arta de yaa Ms h A ARS tN

uy. OT ara

ye, CMO, Elbe ott =r tk Apt aor mae ne 2

75 Years of Scientific Thought

The Washington Academy of Sciences, one of the oldest scientific organiza-

tions in the greater Washington, DC area, has published a book entitled “75

years of scientific thought’’ commemorating the first 75 years of the existence

of the Journal of the Academy.

This compilation, generally aimed at a broad-based scientific readership, con-

tains 25 of the most significant Journal articles, each being of truly enduring

value. Eight of those landmark papers were written by Nobel laureates in-

cluding such preeminent scientific giants as Hans Bethe, Percy Bridgman,

Harold Urey, and Selman Waksman.

This book is the product of an intensive two-year study conducted by a blue-

ribbon multidisciplinary Committee on Scholarly Activities which was chaired

by Dr. Simon W. Strauss, the Academy’s Distinguished Scholar in Residence.

The subject matter, which includes papers on topics such as Theories of Heat

and Radiation, Chemical Nature of Enzymes, High Pressure in Physics, Cul-

tural Implications of Scientific Research, and Separation of Isotopes, covers

a wide variety of scientific fields, including physics, chemistry, biology, an-

thropology, and general science. The 25 papers provide a classic portrayal of

scientific thought over the past three-quarters of a century. For a complete

listing send a self-addressed stamped envelope to the Academy address shown

below. | |

1987, 374 pp., author and chronological title indexes, softbound.

Price for Academy members is $15, and for non-members it is $30.

Send orders to the following address:

Washington Academy of Sciences

1101 N. Highland Street

Arlington, VA 22201

A ee a iy ,) pe iS

i i ' 5 aK

/ Lee « 5

a Pathe bs al

y ; — id neue ‘ , eae ’

, "7 Tay a}

e Ae

; ; fd p F vere ¥ r Sol ; ay

S vel Lp pth te ews eee ee eine

; . “ er ot te r rae oH ve | ee fe

heen. ae Pa tne ee Re Rm a ey PR cs Sd ee ib

a y mt hee Oa A, ah un ¥ . WI he Et bs Ras nei ‘AN rh - ie <i ne

on . Ph Ok: ee a aa Dc ed ee as é nie f

An ee aA heks hate aOR a? Bat) kA" marr es be

" . Nd Mi if) hy

s hs Ae

Pai ia zi Bae

“ rr aise Pde

’ & Eb Pik Ale ine

‘ ‘ ia t f

ig = 7

"1 : = |

5 ce

“aged oie Ww rast a

“SSUES, 0 sth gida Whablo. ott say esonsiae jo Vato) EA. for

rea ee» alitins dood « b artaatcting en Bote ICL: Oty, reidlags 1239

‘SS TuKS ol Pts easy CY tai ott gniiesomeaant iw Mr, yor: oii

: Hae Rie bnoaA aoe Joh os

“Ga ister: sbi Siti inside boawd ‘bald # ta bontie er Sonne POMBE

pinvbAS viet to bbe 1283 satbines inrricl Jnporting pe f ‘)

ber entaowsl lado" yd noir atey emeqey wiearbnnl deod? 1O8

-agmngbnd yo. Sita 20Gb ee “ee winngise ere

| nome nsrulad bow ,

ould 2 yd -betoubno ybute 1694 =D gvinmetn me lo jauborg of Se

bows " Baw “oirh i 2aitivitoa chatosze RO Solio yiamilqioal

fie) ee bte9 Hit reload hed matics ¥ Vina on} RUG Wa

inal % esmosdt ae dove > iqod mo diag esbulons mond. rose ‘F

ly) ened at opens ioe asmnysnd he cue levied! ae

reid agin io nolan’ Dia sioines oilansind Yo enor Re

vaotord ERarenny” ) CQizeitg gpibisl Mit zblott anitnsied I Mie

19 Lagat Mag Di2ahlo & ObIVONG MORES ST concaize leretroy, bus,

Steiqmia F104 soujunney b tp erot wah yp youty ae ont Ta¥O tigue

aware ie sibbe yarsbasA od? af PR Svis sia bores bbs ‘eee 9

i, |

Ms Leah :

ant See ie

= bh

ww eo a).

» ug }

ouciftoe esxober Si ioigelo HON ae bas soda | atl

xy \

PAG ’

a1 :" <A vt ;

8? 2: 1 rociino-noe Wie. 212 2; gradimonm comb A

aor is gaiwotlot adi one :

esonntne We quobas A worgniian -

wort tanita TO)

LOSES ay RODIN A

: i i a oe

Shs my, an

ab ty :

whey Pay uae mas! eal oa

s ty Te ye al va

ee CR a Hat Rit

tsi ae id nave he

Te 7% \y as i

‘ F ere Bint was vy

fe hae

ayy edd Th Ae 4

DELEGATES TO THE WASHINGTON ACADEMY OF SCIENCES,

REPRESENTING THE LOCAL AFFILIATED SOCIETIES

Smeg Med) SOctety Of WaSMINPtOM. 60... 6.6. ee pee vekiec ewes ew eeeneecterees James F. Goff

eeaerncapolagical Society Of Washimgton.......... 00.0.0 .cen cece ede canes Edward J. Lehman

merc) Society OF WaSMINGION .. 0.6.0.6... eee beens enact alnegewes Austin B. Williams

Sea SOCICLY OF WaSNiNOtON. 20... ee cee eta sees eseoeteaccune Jo-Anne A. Jackson

Smemeniiovical Society Of Washington .-..5: 2.0.0... 6. eee ence eees Manya B. Stoetzel

PERE AVCTCOCRADIIE SUCIETY ey 0c 2 oie Gee sca av aees vale swteaeeceaeciwes Stanley G. Leftwich

Bee tes Society Ot WaSHiNPtOM .. 2.2... 2... ieee eee eta ns evn eanene James V. O’Connor

maa socicty Or tic District of Columbia ..... 20... 065.0006. c ca cee cece cceeeees John P. Utz

Pear HIB COME AN OOCICDY 25055). 6. ob. sek dive wands koe den cv view diaie vivian adie sowie Paul H. Oehser

Pa SOCICLY OF WVASDINSION -. 2.2... 5 5.06. c ee cence deen ewan cnbaedeseeeus Conrad B. Link

msocicty of American Foresters, Washington Section ................... Forrest Fenstermaker

DPEMeNTSOCICIY Oli PMGINCEIS 02... is. seer e esse teensees ede veuusreenvewes Alvin Reiner

_ Institute of Electrical and Electronics Engineers, Washington Section...... George Abraham

American Society of Mechanical Engineers, Washington Section................ Michael Chi

memuntaolopical Society of Washington......-........0.0c.cccdeeecaeeeee Kendall G. Powers

American Society for Microbiology, Washington Branch................... To be determined

Society of American Military Engineers, Washington Post............. Charles A. Burroughs

American Society of Civil Engineers, National Capital Section.......... Herbert A. Pennock

Society for Experimental Biology and Medicine, DC Section ............ Cyrus R. Creveling

mumerican Society for Metals, Washington Chapter ...............0.......0- Pamela S. Patrick

American Association of Dental Research, Washington Section............ J. Terrell Hoffeld

American Institute of Aeronautics and Astronautics, National Capital

2 PID E cage ool SPOR eRe IN a Ue eee Reginald C. Smith

amemea Metcorolopical Society, DC Chapter.................00.0 cee eeee A. James Wagner

een mre SOCICtY OF WaSMIMISLON. . 2.2... e cee ce eee cee et es ebeceencteone Ralph Webb

peoustieal Society of America, Washington Chapter.....................-. K. Richard Cook

eaemcan Nuclear Society, Washington SEction....................6ceceeeeee ces Kamal Araj

bastitute of Food Technologists, Washington Section..................0ce eee es Elvira L. Paz

American Ceramic Society, Baltimore-Washington Section .............. Joseph H. Simmons

MRC SOICL VY ce eae ce oes sek sacs cas seveeceesaeeeas Alayne W. Adams

remus Infistory Of seicnce CMD ya. oi. i ec kee eee cece Albert G. Gluckman

American Association of Physics Teachers, Chesapeake Section ............ Peggy A. Dixon

Optical Society of America, National Capital Section..................... William C. Graver

American Society of Plant Physiologists, Washington Area Section ... Walter Shropshire, Jr.

Washington Operations Research/Management Science Council .............. John G. Honig

istrument Society of America, Washington Section.................0.eeeeeeeeees Carl Zeller

American Institute of Mining, Metallurgical and Petroleum Engineers,

RE rate RIONIN CEMA eh Seta tuts <2 2 whid sss andes an eoe pans sodes canes Ronald Munson

Mectitmicl: Apvital AASIPOMOMETS. 624. f 5. eee eee ee ce eee ee eee ees Robert H. McCracken

Mathematics Association of America, MD-DC-VA Section ................ Alfred B. Willcox

Pysuact Of Columbia Institute of Chemists......................0c0e0s Miloslav Rechcigl, Jr.

misimice of Columbia Psychological Association ......./.....0.6. 000. see esceweeenas Jane Flinn

Pesminecon Pat Pechineal (Groups. csc cea seen eee cee en tengeemene Robert F. Brady, Jr.

American Phytopathological Society, Potomac Division................... Deborah R. Fravel

Society for General Systems Research, Metropolitan Washington

ee ee rt ie aE Sk sick id wets Yul ein Balas diame Ronald W. Manderscheid

hibits Pactors Socuety, Potomac Chapter. ........-.<..0.0<cesrececeene Thomas B. Malone

inebican hisheiies Society, Potomac Chapter... .......... 66.00 0ceceseeanawens Robert J. Sousa

mesaciavon ton science, Technology and Innovation... ............0s.s.0aeeeee Ralph I. Cole

le rem SOCIO SICA SOCICUY: cc. eeg se cece nd eee escent eenenenens Ronald W. Manderscheid

Institute of Electrical and Electronics Engineers, Northern Virginia Section.... Ralph I. Cole

Association for Computing Machinery, Washington Chapter............. Charles E. Youman

Se eA PO MS EMISTIC Al SOCICLY 2)..55 52 2ecccs seni eees ceeds cesses teen vevens's Robert Jernigan

Society of Manufacturing Engineers, Washington, DC Chapter ............. James E. Spates

Savicty or Industrial Engineers, Chapter 14.............6.c.002000. ee ee sense John Larry Baer

Delegates continue to represent their societies until new appointments are made.

Washington Academy of Sciences 2nd Class Postage Paid

1101 N. Highland St. at Arlington, Va.

Arlington, Va. 22201 and additional mailing offices.

Return Requested with Form 3579

on VOLUME 80

Number 3

Journal of the September, 1996

WASHINGTON

ACADEMY..SCIENCES

ISSN 0043-0439

Issued Quarterly

at Washington, D.C.

CONTENTS

Conference on Human-Computer Interaction, jointly sponsored by the Human

Factors Society, Potomac Chapter, and the American Psychological Association,

Division of Experimental and Engineering Psychologists, in Arlington, VA, March

1-2, 1990.

PART II

Articles:

LANCE A] MILLER. Natural Language Interfaces” .... 622)... 6 0.00 sess

SCOTT P. ROBERTSON, ‘Knowledge Representation Used by Computer

Ries ISN RUNNIN STS te seem MN re ys ie NG child's laa Ses GS eh A emi ae ete La Gi

JOHN C. THOMAS, “‘The Human Factors of Voice Interfaces” ..........

Washington Academy of Sciences

Founded in 1898

EXECUTIVE COMMITTEE

President

Armand B. Weiss

President-Elect

Walter E. Boek

Secretary

F. K. Mostofi

Treasurer

Norman Doctor

Past President

Robert H. McCracken

Vice President, Membership Affairs

Marie Bourgeois

Vice President, Administrative Affairs

Grover C. Sherlin

Vice President, Junior Academy Affairs

Marylin F. Krupsaw

Vice President, Affiliate Affairs

Edith L. R. Corliss

Board of Managers

R. Clifton Bailey

Jean K. Boek

James W. Harr

Betty Jane Long

Thomas N. Pyke

T. Dale Stewart

REPRESENTATIVES FROM

AFFILIATED SOCIETIES

Delegates are listed on inside rear

cover of each Journal.

ACADEMY OFFICE

1101 N. Highland Street

Arlington, VA 22201

Phone: (703) 527-4800

EDITORIAL BOARD

Editor:

John J. O'Hare, CAE-Link Corpora-

tion

Associate Editors:

Albert G. Gluckman, University of

Maryland

Joseph H. Neale, Georgetown Uni-

versity

Marc Rothenberg, Smithsonian Insti-

tution

Marc M. Sebrechts, Catholic Univer-

sity of America

Edward J. Wegman, George Mason

University

The Journal

This journal, the official organ of the Wash-

ington Academy of Sciences, publishes orig-

inal scientific research, critical reviews,

historical articles, proceedings of scholarly

meetings of its affiliated societies, reports of

the Academy, and other items of interest to

Academy members. The Journal appears

four times a year (March, June, September,

and December). The December issue con-

tains a directory of the current membership

of the Academy.

Subscription Rates

Members, fellows, and life members in good

standing receive the Journal without charge.

Subscriptions are available on a calendar year

basis, a eget in advance. Payment must ‘be

made in U.S. currency at the following rates:

Claims for Missing Issues

Claims will not be allowed if received more

than 60 days after the day of mailing plus time

normally required for postal eae and

claim. No claims will be allowed because of

ae to notify the Academy of a change of

address.

Notification of Change of Address

Address changes should be sent promptly to

the Academy office. Such notifications

should show both old and new addresses and

zip-code numbers, where applicable.

Published quarterly in March, June, September, and December of each year by the

Washington Academy of Sciences, 1101 N. Highland Street, Arlington, VA 22201.

Second-class postage paid at Arlington, VA, and additional mailing offices.

Journal of the Washington Academy of Sciences,

Volume 80, Number 3, Pages 91-115, September 1990

Natural Language Interfaces

Lance A. Miller, Ph.D

Science Applications International Corporation, McLean, VA 22102

ABSTRACT

Following an introductory review of natural language processing activities and analysis

technologies, natural language interfaces are defined and their advantages, disadvantages,

and desirable features are discussed. Four major clusters of natural language interfaces

are identified: Query, Conversion, Commentaries, and Control. A number of different

examples of these clusters are presented and discussed from the viewpoints of cognitive

and computer processing. Finally, four examples of remaining frontiers for practical nat-

ural language interface applications are presented and discussed.

Introduction

A somewhat broader perspective on natural language interfaces is taken in

this review than the customary view that natural language interfaces are pri-

marily input-communication channels to the computer, speech recognition

being the most cogent example. This view takes any natural language pro-

cessing activity supported by the computer as being a candidate interface

activity depending primarily on what processing is accomplished by the com-

puter and how quickly. Thus, today’s ‘‘batch-processing” language applica-

tions could well be tomorrow’s interfaces given appropriate changes in tech-

nological implementation. For example, we don’t ordinarily think of translation

of large texts from a foreign language into English as an “‘interface’’ appli-

cation. However, if the translation were to be accomplished quickly and surely

enough to support an English-speaking-only user’s random hyper-text browsing

of foreign texts in a seamless immediate fashion, we would surely view the

activity to be a natural language interface one. Given the extraordinary ad-

vances of computer performance on the one hand, and the enormous advances

in computational linguistics on the other, such a transition is quite conceivable.

Adopting this orientation we then devote careful attention to a wide variety

of human-computer activities involving natural language—briefly emphasizing

speech recognition. We focus on the differences between natural and non-

natural languages and the nature of the various types of computational tech-

92 LANCE A. MILLER

nologies involved in processing natural languages. The scene having been set,

we then define a subset of natural language applications as being interfaces,

and these provide the context for the remainder of the paper.

Natural Language Interfaces, at the present imperfect state of technology

capability, have a number of disadvantages to offset their obvious, and not

so obvious, advantageous features, and we appropriately dwell on both of

these.

Given a Natural Language Interface, we consider how we might evaluate

its effectiveness, particularly its cognitive adequacy, by enumerating and re-

viewing an extended set of desirable features.

Within all of the above context we then identify a set of four clusters of

possible and actual natural language interactive interface activities, as char-

acterized by their input-output characteristics. We provide a number of ex-

amples of these, focusing on a few, including that set which combines natural

language with non-language gesturing actions.

We conclude the review with a proposal for several new types of practical

natural language interfaces whose development would provide great practical

advantages.

Types of Natural Language Applications

Types of Natural Language Input/Output Applications

Almost all computer processing applications involving natural language are

concerned with language as either input or output (the alternative being to

employ language forms for internal reasoning activity independent of I/O).

For language as input, the applications differ as to ultimate function but all

involve the analysis at various levels of the natural-form input.

In Table 1 are listed 8 examples of areas involving input-language analysis.

There has been some interest in using particularly voice input in command

situations, such as of industrial robots. Analysis of natural language software

requirements and designs, expressed as text, has occasionally been raised as

a possibility, but there has been little computational work. Analysis of meta-

level queries, such as “‘what do you know about’’, have also been cited, but

the extensive query work involves analysis of specific concrete queries. Some

knowledge acquisition work exists in which users enter simple natural language

assertions about properties of entities and entity-relations. There continue to

be an increase in the number of programs which provide analysis of stylistic

characteristics, and to a lesser degree the thematic and other content, but

there is very little work attempting to provide deep analyses of substantial

coherent text.

Concerning the 11 types of natural language output generation also listed

Table 1.—Human and Computer Activities Using Natural Language (NL) Processes

NATURAL LANGUAGE INTERFACES

Extent of

NL Process Activities Computer Research

NL Input Analysis Command Following Some

Software Design/Programming Little

DB Query Analysis Extensive

Meta-level Query Analysis Little

Single Assertion Comprehension Some

Style Analysis Extensive

Content/Theme Analysis Some

Deep Understanding of Coherent Discourse Very Little

NL Output Generation Status Informing (speech) Some

Instruction Generation Little

Response to Query Some

Conversational Response Some

Report Generation (from DBs, etc.) Moderate

Abstracts, Synopses, Paraphrases Little

Language Translation Extensive

Sketchy stories, plans, etc. Little

Short Explanations Little

Extended Reasoning Very Little

Large, creative, coherent productions (e.g., novels, None

instruction manuals, position papers)

in Table 1, some work involves synthesized speech status reports or directions,

and text query responses, and conversational output, but most of the output

work has been concerned with bulk text translation. Aside from some data-

base-driven report generation, there has been little study of other forms of

output and essentially none concerning large creative productions.

Only some of the applications listed in Table 1 would today be considered

as involving natural language interfaces but we argue later that all of these

are potential candidates in the future, given suitable technological improve-

ments.

Physical Characteristics of Natural Language Input/Output

When natural language is input or output by human or computer it has a

variety of physical characteristics as described in Table 2. For example, the

generic input activity of reading involves some type of natural alphabet or-

thography when presented to humans and some form of digital encoding

standard for computers. The acquisition of computer-readable text is typically

accomplished by human transcription, used various word-processing or text-

processing software packages. However, more and more capability is being

provided by acquisition directly from physical text via scanning devices coupled

with Optical Character Recognition (OCR) capability. The computer activity

comparable to human listening is that of speech recognition which involves

processing of digitized speech samples either represented as time-varying sig-

nal waveforms or as Fourier-transforms of power vs. signal frequencies.

94 LANCE A. MILLER

Table 2.—Basic Natural Language Input/Output Processes and Their Associated Signal

Signal Characteristics

Computer-Equivalent Computer Activity

I/O Process Human Representation Known As

Input’

Reading Natural-alphabet orthography ASCII/EBCDIC codes Text processing

Listening Complex speech, frequency Digitized speech samples and Speech recognition

range ~ 100-2000 Hz. their transformations

Output?

Writing Printed, scripted, typed ASCII/EBCDIC Text generation

orthography

Speaking Speech utterances Digitized synthesized speech Speech synthesis

'Because the vast majority of study and development is on the input in the table, we exclude from

consideration such other legitimate input processes as braille-reading, interpretation of signed gestures,

ete:

*For the same reason we do not consider signing and other types of non-standard output.

Human language communication involves a variety of other types of signal

encodings besides those listed, including use of sign-language, representation

of text as braille, and generation of special coded signals involving such devices

as radio morse-code and signal lanterns and flags. Aside from signing, these

forms involve relatively uninteresting non-linguistic transformations. On the

other hand, the interpretation of signed gestures produced by a human (or

the generation of understandable animated signing output) provides a variety

of so-far unstudied provocative challenges: recognition of the properties and

relations of spatially-separated entities established during sign discourse, char-

acterization of signing in terms of a grammar of signs, elucidation of the signing

mechanisms corresponding to the complex features of natural discourse (e.g.,

topicalization, passivication, “‘lexical’’ collocation, anaphora, etc.), etc.

Natural Languages vs. Non-Natural Languages

The discussion of signing provides an introduction into the issue of what is,

and what is not, a ‘natural’ language. For the purposes of this paper a natural

language is one (1) which is commonly and long-accepted as a language of

communication of a human culture in its every-day activities, (2) involving an

established orthographic representation of spoken utterances composed from

a base alphabet of symbols, and (3) for which there is no evidence of its having

been artificially devised or created. Such a definition removes from consid-

eration for the moment such potential candidates as whale language, com-

munications from primates, and even human signing itself. What remains are

the unequivocal languages such as English, Russian, Latin, Spanish, etc., and

we will simply side-step issues such as the status of dialects, pidgins, and

creoles.

NATURAL LANGUAGE INTERFACES 95

With respect to computer processing of language, there are 3 categories of

language and non-language that have been considered, as represented in Fig-

ure 1. The first is true natural language. Second is a layer of natural languages

which are constrained in more significant ways than just by vocabulary re-

strictions (prohibiting relative clauses, for example) or an artificial language

claimed by its proponents to have all the features of the natural model (e.g.,

Esperanto). These layer languages, which are not considered truly natural

here, do however cover most of the important forms of natural language.

Finally, there are a variety of non-natural languages which have been partic-

True

Natural Language

Other Siig Forms

Queries Assertions

Commands

PSEUDO LANGUAGES AND CONSTRAINED NATURAL LANGUAGES

QUERIES COMMANDS ASSERTIONS

0) Natural-Like ) Robotics Languages 0) Knowledge Base

Query Languages Formal isms

0) 4GLs 0) Procedural Programming o Logic Programming

Languages Languages

ty) (SQL) 0) SQL tf) Logics (e.g.,

First-Order)

Algebras, Calculi

NON-NATURAL LANGUAGES

Fig. 1. Examples of non-natural high-level computer-input specification languages corresponding to three

natural language forms of queries, commands, and assertions. A layer of pseudo natural languages (e.g.,

Esperanto) and constrained natural languages separates the natural from the non-natural languages.

96 LANCE A. MILLER

ularly developed for one or another type of form: query, command, or as-

sertion.

An important characteristic that all of the languages shown in Figure 1

possess in common is the requirement that the input must undergo various

levels of linguistic analysis in order to be utilized in the ultimate application

of providing data to respond to a query, issuing a machine motor-control

instruction, or correctly capturing and representing some kind of knowledge

assertion. Such analysis is discussed in the next major section on Technologies,

but it can be said that all require a grammar of at least context-free power

(in the 4-level range from regular, to context-free, to context-sensitive, to

Turing-Machine). Where natural language differs primarily from the others

in this grammatical regard is in terms of the diversity of acceptable parse-tree

structures, relating to the much more flexible set of allowable variations in

the surface input than is acceptable to non-natural languages.

Natural Language Speech Recognition

In this general section on natural language applications it is appropriate to

focus briefly on the now venerable task of getting computers to recognize

human speech. This is probably the most popularized of all the natural lan-

guage interface applications, and it has reached a stage of impressive maturity

in the past five years.

Key to understanding the difficulties of speech recognition and the recent

progress is an appreciation for the three first-order variables used to classify

speech systems: (1) whether the system handles discrete vs. continuous speech,

(2) whether the system is speaker-dependent or -independent, and (3) whether

the recognized vocabulary is relatively small (e.g., <2,000 words) or large

(e.g., >10,000 words). Five to ten years ago there was still quite active research

in speech systems at the lowest end—handling only discrete speech (speech

with speaker-inserted pauses between words), for small vocabularies, and

requiring special training for each new speaker/user (speaker dependence).

Today, hardware boards to accomplish that level of recognition are routinely

available from several manufacturers for a few thousand dollars.

Today’s R&D speech systems are focused on the high-end of all these

variables: speaker independence, large (to very large) vocabularies, and un-

restricted continuous speech format. In contrast to the past, most of today’s

promising systems place heavy emphasis for the determination of speech input

on language technologies, particularly phonetic-alphabet grammars employing

dictionary look-up and natural language word-string parsers. In Table 3 are

listed some of the speech recognition systems under development which fit

the above characteristics. Of these, some efforts so emphasize the natural

NATURAL LANGUAGE INTERFACES 97

Table 3.—Selected Speech Processing Systems

System Where Purpose Features’ Reference

Sphinx CMU Speech Recog. Cont, SI, LV Lee, 1989

Pundit Unisys NL Parser NP emphasis Dowding & Hirschman, 1987

SUMMIT MIT Speech Recog. Viterbi Search Zue et al., 1989a, b

DECIPHER SRI Speech Recog. Cont., SI, LV, Pr Murveit et al. , 1989

TINA MIT NL Parser Best-first search Seneff, 1989

MINDS CMU Word Predictor Context Knowledge Young, 1989

BYBLOS BBN Speech Recog. Context-dependent Chow et al., 1987

HARC BBN Speech Recog. Chart parser, semantics Boisen et al., 1989

'Cont = continuous speech input, SI = speaker-independent, LV = large vocubulary, >10K words,

NP = noun-phase, Pr = probabilistic

language parsing aspects of guiding the recognition systems that a great deal

of the effort may be spent on just these aspects; examples are Unisys’ Pundit

and MIT’s TINA. A key research question for these systems is how to conduct

the lexical-search and parsing such that useful results are obtained in real-

time to help confirm or redirect the ongoing hypotheses of the front-end audio

analyzer components.

While no robust full-function commercial speaker-independent continuous-

speech large-vocabulary systems are yet available, there is considerable reason

to suspect that yet another five years will see their presence as flexible input

interfaces for a host of practical applications.

Natural Language Technologies

We are accustomed to characterizing many ‘“‘hard” application areas in terms

of the variety of technologies that underlay and make possible the achievement

of the application functions. For example, for the area of Advanced Manu-

facturing (flexible manufacturing systems, automated assembly cells, etc.) the

key technologies include numerically-controlled machines, Computer-Aided

Design software, Computer-Integrated-Manufacturing control packages, etc.

However unfamiliar it may be, it is equally appropriate to characterize natural

language processing applications in terms of the language “technologies” which

underlay them.

We provide this characterization for the first of the two types of natural

language applications portrayed in Table 1, the analysis of natural language

input.

Description of Language Technologies

We show in Table 4 five types of language technologies which can be (sim-

plistically) thought of as applying successively to transform the original lan-

LANCE A. MILLER

Table 4.—Levels of Natural Language Input Analysis

Processing Example

Level Unit of Analysis Major Activities Output (‘Take a chair’’)

Lexical Word Segmentation, affix- | Data records(s): take: vb, pres, sing

stripping, inflec- Part-of-speech (POS) take: n, sing.

tional analysis, person, number,

lookup TENSES)...

Syntactic §Phrase/Sentence Parsing: POS assign-. Parse-tree(s), prob- SENT

ment government, lem-report

Pinel pecans VERB —_ DIROBJ

straint-checking,

SVO case-assign-

ment take aa

ART N

a chair

Semantic _Parse-tree, annota- Scoping, Quantifica- Logical form takel (P1, IMP,

tions tion, sense-selec- ?you, Chair,

tion, predicate-for- ?dest)

mation

Discourse Logical form, Dis- Anaphora resolution, Variable substitution (Previous: “Jim, let’s

course context argument-filling, in logical form, play cards. .”

sentence fragment additional predi- ?you = Jim

coordination cates ?dest = in (Liv.

Rm.)

Pragmatic Discourse Represen- Hypothesize, goals, | Augmented logical goal (P1, provide __

tation, goal infor-

mation, real world

knowledge

intentions, beliefs

form

seating), intended

__act (play4),

believe (speaker,

Jim, X) X = want

(Jim, play4)

guage input into a fully-analyzed pragmatically-understood form. We arbi-

trarily assume that the initial technology of lexical processing begins with a

preprocessed string of segmented tokens each of which, under ideal conditions,

represents either a word or a punctuation symbol (in practice the tokenizer

process may well have to interact with the lexical process to achieve this).

The purpose of lexical processing is to develop a data-record for each word

that contains the feature information needed by the syntactic processing which

follows. In the Table 4 example of ‘“Take a chair’’, the lexical analyzer will

typically report features for all of the different parts-of-speech it can assign

to each word; for “‘take”’ both a verb and a noun part-of-speech are identified,

along with some respective features. Perhaps the most important of the lexical

processing activities (a sub-technology in its own right) is the stripping of

affixes (prefixes and suffixes) until a remaining stem is discovered (e.g., the

stem “‘establish”’ in ‘“‘antidisestablishmentarianism’’). |

The beginning stage of syntactic analysis is a string of word records primarily

characterized by their part(s) of speech. These parts of speech are represented

NATURAL LANGUAGE INTERFACES 99

as so-called terminals in a grammar of rewrite rules which specify how these

terminals can be combined into larger linguistic units (non-terminals, like

noun-phrase and prepositional phrase). The desired output of the syntactic

analyzer is that parse-tree which most appropriately describes the grammatical

governing and binding characteristics of the part-of-speech elements; in the

example, ‘“‘take”’ is recognized as a verb, not a noun, with the direct object

of its imperative form being the noun-phrase “‘a chair.”

Whereas the actual words and their meaning are of secondary importance

in syntactic processing, in semantic processing the word meanings become

critically important. Suppose that the context for the example sentence ““Take

a chair” is a social evening in a home, with two couples getting ready to play

cards by setting up a card table in the living room and gathering chairs for it

from the dining room, and the host—pointing to a particular chair—utters the

request with the clear idea that the visitor is to transport it into the living

room. Thus, of the many identifiable senses for the verb to take, the one of

transporting is intended (not the one of “behold!” or “‘consider’’, for example,

or of ‘‘deduct” as in “take 5 away from 17. . .’’). This sense may be repre-

sented as takel, and it can be represented as being associated with additional

pieces of information concerning (1) an identification of this unique propo-

sition, say ‘‘P1”’; (2) who the addressee is in “‘(You) take the chair’’, repre-

sented by ?you; (3) what the object of taking is, Chair; and

(4) what the destination is of this action, ?dest. These are the ‘“‘arguments”’

or parameters that need to be filled in for the takel sense of “‘take”’ in our

supposed processing system (another processing system might want fewer

arguments, eliminating “?you” or possibly even more, adding one for the

speaker). Semantic processing typically arrives at an estimate of the underlying

word-sense for words in the utterance by consulting the semantic-forms—

often called case-frames—associated with a dictionary entry, and comparing

the constraints for the various arguments to elements in the utterance; for

example, if a case-frame direct object is supposed to be an inanimate thing,

as with takel, then ‘“‘chair’—being inanimate—fulfills that restriction and the

takel sense is still a possibility.

The next technology, discourse processing, brings in consideration of the

preceding text or dialogue to add more information to the semantic logical

form by determining things like pronoun referents (there are 3 in ““He gave

it to him’’)—anaphora resolution—helping resolve some of the unknown ar-

guments in the logical form (?you and ?dest), deciding what the topic and

focus are in the input, checking for clues which reveal the speaker (or hearer)

attitudes or argument structure, resolving so-called cohesion mechanisms which

link parts of the input together, and so forth. Thus, the previous dialogue

100 LANCE A. MILLER

might have included the host saying “Jim, let’s play cards . . .”” with a sub-

sequent reference to the living room, and processing of this previous discourse

could resolve the values for the missing arguments in the semantic logical

form.

Finally, the last stage of natural language input analysis involves the tech-

nology of pragmatics—attributing to the agents involved in the language in-

teraction various goals, beliefs, intentions, and plans. Up to this point the

focus has been on what is being said; here the emphasis is on why. Our

sophisticated pragmatics processor might thus conclude that the goal of the

stated imperative proposition P1 is to provide seating, where the intended

action is a particular type of “‘playing’’, play4; and the pragmatics processor

further concludes that the speaker of ““Take a chair” believes that Jim wants

to engage in play4.

The maturity and competence of these five technologies is highest with the

lexical and syntactic processing and falls off markedly after semantics. With

respect to non-natural languages, no differentiation is made among the last 3

technologies, nor are the semantics—in the sense of constraints on argu-

ments—often made explicit in formal defining assertions; the constraints are

usually implicit in the procedural compiler code.

Very large dictionaries of simple word strings are now very common because

of the interest in PC-level spelling checkers, but full dictionaries with extended

word-features for multiple senses, and senses represented as case-frames with

selectional constraints, are also becoming more available, at least in research

settings—for other languages in addition to English (cf. Byrd, et al, 1987).

The trend is very much towards putting more and more information in the

dictionary, particularly in contrast to representing the same information in

the form of grammatical or other rules; some syntactic approaches very much

rely on this strategy (e.g., Lexical Functional Grammer; Kasper, 1987).

While there is considerable agreement among computational linguists con-

cerning desirable dictionary features, there is much less agreement concerning

how the syntactic processing should be accomplished. In the early days of

parsing, there were basically two approaches: bottom-up assembly of part-of-

speech terminals into phrases, and these into clauses (e.g., phrase structure

grammars; cf. Heidorn et al., 1982), and top-down hypothesization of high-

level clauses or phrases decomposed down into parts-of-speech (e.g., Aug-

mented Transition Networks; Woods, 1980). Today, given the wide variety of

natural language applications being studied, there is a great diversity of gram-

mar approaches and formalisms, each having particularly useful properties for

its intended span of applications. Table 5 lists some of these approaches which

may be very briefly sampled: Government and Binding Grammar is proving

NATURAL LANGUAGE INTERFACES 101

Table 5.—Some Popular Grammar/Parser Formalisms and Approaches

Grammar Reference

Government and Binding Grammar (GB) Berwick & Weinberg, 1984

Lexical Functional Grammar (LFG) Kasper, 1987

Tree Adjoining Grammar (TAG) Joshi, 1987

Augmented Phrase Structure Grammar (APSG) Heidorn et al., 1982

Augmented Transition Networks (ATN) Woods, 1980

Categorical Grammar Pareschi, 1988

Head-Driven Phrase Structure Grammar (HPSG) Maeda et al., 1988

Chart Parser Allen, 1987

Definite Clause Grammar (DCG) Pereira & Warren, 1980

String Grammar/Functional Grammar Sager, 1981

Functional Unification Grammar (FUG) Kay, 1985

Modular Logic Grammar (MLG) McCord, 1985

to be particularly useful in our BRIDGE Tutor project for multi-language

training of Army interrogators (cf. Berwick & Weinberg, 1984); the bottom-

up Augmented Phrase Structure Grammar was very useful for the EPISTLE

text-critiquing project (cf. Heidorn et al. , 1982); and Modular Logic Grammar

is proving to be an effective approach for language translation (cf. McCord,

1985).

Syntactic processing and the other language technologies discussed are per-

vasive throughout all types of natural language applications, not just analysis

of input. However, there are additional technologies that apply more to output

generation that are also identifiable in the research literature. For example,

the process of planning a discourse or text involves very complicated consid-

eration of the goals balanced against the nature of the audience, the time and

resources available, etc. At the end of this planning process are the detailed

decisions concerning what actual words to use, how much information to give

in One versus multiple sentences, etc. A related not entirely subsidiary process

is establishing cohesion between the next planned output and the previous

discourse as well as the discourse environment. As our understanding of these

processes matures, and as they are better able to be represented in software,

then it is likely that they will emerge, like dictionary and syntax, to be full-

fledged additions to the arsenal of linguistic technologies.

Interaction of Technologies with Application Effectiveness

An illustration of how the language technologies work together and how

one gains additional application capability as they are added on is given by

the Table 6 example of text-critiquing for various kinds of spelling, grammar,

and stylistic errors (see Miller, 1990). Considering spelling errors, certain of

these need only a dictionary look-up to determine that they are wrong—as

with “‘myne”’ for example, which is clearly not a word. However, when the

102 LANCE A. MILLER

Table 6.—Types of language technology needed to detect 3 different classes of text-production errors

Language Technology

Text Errors Dictionary Syntax Semantics

Spelling “This is myne.” “Don’t press fo hard. “The pipe collapsed due to

mental fatigue.”

Grammar “He ain’t happy.” “The purpose of these meet- ‘The age of these men

ings are...” which are unmarried.”

Style “Our arrangement has now “He who has not knows what ‘Coming around the corner,

been finalized.” he who has possesses.” the building shone red in

the sunset.”

substituted word is also a word—as in ‘“‘Don’t press to hard’’—dictionary

technology alone is insufficient to detect the error; syntactic analysis is needed

to reveal that ‘“‘to” and “‘hard”’ can’t be put together to form a prepositional

phrase or an infinitive or any other acceptable construction. In the third

example concerning “‘mental fatigue” (instead of ‘‘metal fatigue’) both dic-

tionary and syntax are necessary but insufficient to detect a problem; semantics

really are necessary to determine that mental fatigue doesn’t apply to pipes!

In general, improvements in one technology can compensate for deficiencies

in others, and one is often able to solve a difficult application problem by

bringing to bear various aspects of several technologies, not just the one that

seems most to apply. Thus, for example, ways of handling certain kinds of

complex grammatical constructions have been developed by putting additional

information into the dictionary or increasing the power of the semantic com-

ponent.

Natural Language Interfaces: Considerations

We are finally prepared to deal with the main topic of the paper, having

laid the necessary groundwork in terms of applications and technologies. In

this section we deal with various considerations of Natural Language Interfaces

(NLI), leaving to the next the discussion of specific examples.

Definition of NLI

NLI have two forms of realization, input and output, and we propose the

following definition to cover both:

‘“‘Natural Language Interfaces are communication channels between human

users and computer systems and involve the dynamic processing of coherent

natural language, either as analyzed input or as generated output, with suf-

ficient speed and accuracy to support an interactive task.”

NATURAL LANGUAGE INTERFACES 103

Essential phrases are those of communication channels (to insure that the

language has a key input or output role), dynamic (to eliminate pre-planned

or canned processing), coherent (to indicate the requirement to guide pro-

cessing via a larger language context), natural language (to indicate the need

for language technologies and exclude artificial ‘“‘languages’’), and interactive

(to eliminate one-shot or batch-type applications). Thus, translation of a

speaker’s typed input from English into French text is an example of NLI if

it occurs rapidly, such that it could be dynamically reviewed by the user (or

read by a Parisian and replied to), and it isn’t if the translation is delayed

until all the text is input and then translated at some later time.

Advantages of NLI

NLI are usually touted for the fact that users don’t have to learn them, as

shown by the initial advantage entries in Table 7. Less frequently mentioned

is that fact that, along with the language itself, come mechanisms for using

the language that are extraordinarily adaptive and productive. A person,

looking at a drawing of an unfamiliar mechanism, has several well-established

strategies for generating acceptable names of parts of the mechanism for use

in discussion; and has well-used strategems for generating descriptions of the

mechanism’s operation or formulating queries. There are, in addition, a half-

dozen or more specific natural language mechanisms for referring to things—

e.g., extrinsically (“that one over there, third from the left’), anaphorically

to prior dialogue (“‘she didn’t know whether she had said that, but that wasn’t

the point. . .””), and even cataphorically to dialogue yet to come (“‘The points

I want to make are these three: .. .’’).

Table 7.—Advantages and Disadvantages of Natural Language Interfaces

Advantages Disadvantages

@ Naturalness, cognitively not demanding @ Verbosity is tiring for experienced users

®@ Highly Portable e Ambiguity is always a problem (sense,

® No learning required reference, figurativeness, scope, quantification,

@ Easily remembered attachment, goal. . .)

@ Flexible, adaptive to a host of situations @ Presupposed, entailed, and implied general

e@ Thus, very big reusability knowledge is so great

@ Easily productive for new requirements @ Subtleties of speaker’s attitude as expressed in

(naming, new procedures) the utterance/text can’t yet be appreciated

®@ Extraordinary facilities for reference (and @ Users don’t have analytic knowledge of

indexing) and description language processing to help computer out

@ Provides (esp. via semantics) for all desirable @ Ellipsis, fragments and ill-formed but (human)

features of ideal programming languages— comprehensible input cause difficulty

information-hiding, data abstraction, operator @ User’s high expectations of computer’s

overload, feature inheritance, each entity with capability can’t be met

its own methods, etc. @ Expertise needed to develop and maintain NL

system is not readily available

104 LANCE A. MILLER

Other advantages accrue from the same reasoning used in the field of soft-

ware engineering to argue for one programming language or methodology

over another. Thus, natural language is as reusable as things get, and all of

the special features of object-based and object-oriented methodologies so

popular these days have their cognates in natural language features.

Disadvantages of NLI

On the other hand, non-enthusiasts will point at a number of supposed

problems with natural language, particular as input. These points are well

taken with respect perhaps to existing NLI implementations, but most of them

would lose their force were the analytic capability of the computer to be

substantially improved. Thus, it is certainly true that natural language can be

verbose if the input analyzer doesn’t support all the normal ways that natural

language provides for terse rapid communication—if it can’t handle ellipsis

and fragments, doesn’t handle relative referencing and indexing, and has little

semantic/pragmatic capability to resolve the meaning from the context. Sim-

larly, ambiguity can occur in a variety of forms and can be truly debilitating

if the processor is primitive in its capability to handle it.

More fundamental is the fact that people typically have little appreciation

for what is and is not difficult to process, and they will be of little help in

assisting the computer in resolving their input unless the problem is stated

just right. Similarly, people often overcome communication difficulties via

their vast shared knowledge about the world, and it’s unlikely that this aspect

of NLI will approach this capability for many years; this in turn contributes

to users’ being frustrated with what the computer can do vs. what they expect

it to be able to do.

Concerning the subtleties of engaging in conversation, taking turns, de-

tecting key attitudes towards the discussion topics, and identifying commu-

nication problems, etc., there is a great deal of progress on all of these issues

towards finding ways of formalizing them in software, such that these aspects

can be imagined to be well-supported in future systems.

The general solution to the problems of using NLI, then, is to improve their

processing sophistication and capability; there appear to be few truly inherent

disadvantages of natural language when it is implemented in its full range of

mechanisms. Given the progress in all of the language technologies, this ap-

proach appears to be genuinely feasible within the next decade.

Desirable Features of NLI

Having decided in favor of NLI, what then should one look for or insist

upon? Here one must take into account general psychological principles as

NATURAL LANGUAGE INTERFACES 105

well as specific knowledge concerning language-use. We provide some 14

criteria in Table 8 for evaluating NLI, and we indicate our evaluation of their

progress to date. These desiderata are listed roughly in order of decreasing

priority, from the point of view of insuring the most support to the widest set

of NLI applications.

First on the list is robustness. Every human editor appreciates how difficult

it is to eradicate every last ungrammaticality, typo, punctuation error, not to

mention the more serious problems of non-sequiturs, poor organization, un-

even content, etc. NLI systems must be able to handle the ill-formed input

of all kinds that is certain to occur, if only to come back with clarification

requests, just like people do all the time.

Next on the list are good coverage by all the language technologies. NLI

with toy vocabularies, skeletal grammars, and inadequate higher processing |

cause the greatest frustration and provide poor application support because

of inadequacies and errors. Although off-the-shelf computer dictionaries and

grammars are not yet common, there is today a much higher incidence of

reusability of language technologies than ever before, and coverage problems

will certainly become more tractable.

Table 8.—Desirable Features of Natural Language Interfaces

General

Desirable Feature Explanation Status’

Robustness Can do its job despite ungrammaticalities, misspellings, punctuation 3

errors, etc.

Good Word Cover Vocabulary is completely sufficient. 3

Good Grammar Cover Deals with wide variety of grammatical constructs. 3

Good Semantics Cover Develops acceptable meaning interpretations. 2

Good Pragmatic Recognizes intended meaning and purpose. 1

Interpretation

Graceful Failure Doesn’t just quit, provides information on its difficulties, suggests 2

alternatives.

Explanation Facility Gives the user and the developer some kind of trace or explanation yi

capability.

Handles Ellipsis/ Accepts (syntactically) incomplete but contextually comprehensible 3

Anaphora input. Correctly interprets referring expressions.

Reasonably Extensible Most importantly its vocabulary can be added to with modest 3

difficulty.

Input Facilitation Uses variety of means to ease input, such as ‘“‘auto-completion”’ of 2

previously used words based on first few letters.

Problem Detection Sensitive to user difficulties or, especially, misunderstandings. 1

Paraphrase Production Can produce pragmatic equivalent to an expression to aid user 2

understanding.

Supports Extended System can engage in a “‘conversation” with user, lasting a number 2

Dialogue of turns, making the appropriate integrations and inferences.

Supports Multi-Modal Integrates natural language text (or speech) input with gestures, eye 2

Input

movements, etc.

'Status codes are on a 5-point scale from 1 = very little progress to 3 = substantial progress to 5 =

problem essentially solved.

106 LANCE A. MILLER

The next two points, graceful failure and explanation facility—and also

paraphrase production—provide the user with some insight when difficulties

are encountered, something they are more and more accustomed to in most

of the non-NLI computer applications. A related, and more difficult capability,

is to monitor user performance for problems in the use of the NLI and provide

feedback.

The handling (and production) of language fragments is essential to sup-

porting fast-moving dialogue, as is the capability to handle all the varieties of

referencing, whether via pronominal anaphora, definite noun phrases, or deic-

tics (‘‘this, that’’).

A lot more could be done to permit users to add capability to NLI systems,

particularly vocabulary. Similarly, for typed input, greater facilities can be

provided to reduce the amount of typing required. One useful technique is

to have the system automatically complete the input with words it knows given

the first few input characters, changing the candidate as necessary with each

new character input.

The final criteria provide extensions of the interface both in terms of time/

coverage, for dialogue support, and also in terms of modality, to incorporate

information from gestures and eye movements in particular to be used as

referring sources.

Natural Language Interfaces: Types and Examples

Four general classes of NLI can be identified, as shown in Figure 2. Query

NLI typically involve a typed-text question augmented in some cases with

information from a gesture or eye-movement. The response is primarily in-

formation from a database. Conversion NLI involve the mapping of input into

some transformed type of output, either a shift in modality from speech to

text (or vice versa) or a translation within the same modality (only text, so

far) from one natural language to another. Commentary NLI have a much

broader input capability, taking in almost any object for examination and then

providing some natural language reaction to it—comments, critiques, anno-

tations, summaries, or brief reports. Finally, Control NLI use natural language

speech or text commands to control the computer system, applications, or

some attached robot or mechanical operation. Examples of these types of NLI

are shown in Table 9 (pure speech recognition systems were presented in

Table 3).

The table examples are really just a sampling of a much larger set of NLI

which are undergoing substantial development, not to mention the numerous

additional 1-2 person university projects. A decade ago a listing of all systems,

including 1-person projects, would have hardly filled the page. This is due

NATURAL LANGUAGE INTERFACES 107

Input Output

Touch

Text Query + Point

Data Summary

Text Query Menu Selection

Extended Text

Speech Text

Extended EE Translated Text

Dialogue Speech

Extended Text

DB + Changes Critiques

Software Programs Summaries

Knowledge Representations Reports

System Commands

Commands Robot/Mfg. Commands

Speech Appli'n Control

Fig. 2. Four Clusters of Natural Language Interactive Interface Activities

partly to the broad expansion of effort in the NLI arena in this time period.

It is also due to the improvement in computer software efficiency and hardware

speed such that these applications are capable of being truly interactive.

A number of the examples are much richer than they may appear from the

tabled descriptions. For example, the Bridge Tutor (BridgeT) is being devel-

oped to provide second-language support and training for Army interrogators

(MOS 97E; Miller, 1990). In the simulation mode, personnel are to type in

LANCE A. MILLER

108

sjuoua[q AianO

E861 ‘79 1a yuRUDL, nus IN JO UONI9[9g-Nus|] (1)ussugq ad I Bed € Alan} payouysoy

(O)youery IXOL,

KasskpO Vuo [eIoIOUIWIOD (O) ysysuq ax ‘ad ¢ aouajUag-HINY] = ad

(soidoy) sopn1y ‘puodaq

BBg6I ‘AAO ANITAVd saselg ‘peg ‘pooH (O)usIsuq =6 gM + SO S sousjUog-HINY] = jenjdasu09

THOIN

wody ‘AIOOYL, onbiyta

egg6t “AAOH = (Wd) NVINNAd sINJONNS WoyY (O)ustsuq ay Ss s0u9}Ueg-HINY] = weis01g SIT

9861 ‘79 Ja UUZ|DH SISVO suoneoyddy jomuoD (usysuq ynduy = sulyorlL, C OkF 2 Sd10A

suonjeoyddy

OJEIDOSSY S,}O]Id suoneoyddy ydyp05 (1)usysuq induy 9 OUT] [POY € d10A,

[omUu0D yonoy,

6861 “JO3STYEM VuLXx [2POW! esINOIsIq (Duswsuq yonoL/gd oT ‘Iddy eyeq 7 2p anduy 1x9],

uoneIS9}U] JomuoD yonoy

8861 ‘o1ldeys 2p [PON Noorano PIPSA-DINN (Dysysuq yonol/qd i ‘Iddy eyeq p 29 induy 1x9],

eloydeuy yonoy,

6861 ©7212 UsYOD M1VLdOHS [NjJoMog UOINTOsoY (Duswsuq = yonoL/gd I eyeq v 29 Alan© 1x9],

6861 sunulog WOT}da]2g

‘Ayyedeuedy 2p IOUNaAA (2A0]D) aaoineied TdA (ustsuq /ad I -nuay| ig BUNUIOg 2 Id10A

uonelodi0D [Vv 09] [93UT "WWOD ‘UdALIG-"}0IG ({)ystsuq aad if eed Z AIONQ 1X9]

L861

‘IOUEM 2) XUpusH] V0 |. WWIOD “jsnqoy (ysisuq ad I Bed p AYan© IxX9L

L861 * 1) 12 BMEXIYS] di [2POW PHOM. ({jesoueder dd I eyed 7 AJan© 3x9],

9861 ‘OUT IN “suey [einen WO] [eso] (1) ysisuq dd I eyed p AJaNn© 3X9],

8861 ‘prelled ITAL JOWWeID [es11089}e_) (usisuq ad I eyed 7 AJand 1X9

QOUdIOJOY SUIBNY WId}sAS soinjeay O/] osensuey] ¢90IN0S 7SUISSIIOIg jndjno [240] ynduy

UO eWIOJUT jndjno SulsIeg

jndjno

UOIIUSOIIY BI0OA LIPI[A[-I[SUIS ULY) 19YjIQ SUId}JSAG JdvJ1IjIUT VSeNSULT] [eANjeA! VWIOGS—"6 J]qQU],

109

NATURAL LANGUAGE INTERFACES

aseg ospojmouy = gy ‘oseqeieq = gq ‘s9poD Indjnoc,

JoNUOD uoneoyddy = 9g ‘asuodsay

QOUdIUIS-INJY poyesrou9H = ¢ ‘Wag/aseiyd pojyeiou9H = p ‘suUONeUIOJsUBIT, UONPIOUNUOIg = ¢€ ‘[RADINIDY 1X9] = Z ‘[RADNOY eIVG = [| :s9poD surssod01g jndjnoC,

‘SuIsieg DUueWIaS

jonorqudcs [ny = p ‘IRWUWPIO 991J-}X9}UOD JO Ie[NsIYy 0} pouleNsuOD = ¢€ ‘SUISSOD0Ig [edIxoT] A[UeWIg = Z ‘SuNI0dg oseIyg/PIOM, = | :SOPOD JOAoT] Bursieg jnduy,

“‘WISSY J0QOY Jonuo) Alang/

9861 ‘28PLY[OS AHduNW ‘SuIpueysiopuy) isngoy (O/1) usisuq ay 7 ‘IOMSUY TH v puewwoD x9]

spuewwo)

LL61 “ouery SOd IN sofo1g Oqiny, (Dusysug = qy/induy ia SOa-sW 7 puewwo xa

L861 ‘']) Ja ouRWIY oesurII-SV jonpolg eqrysoy, (Q/1) der--3uq ynduy v uoTesueL TL, v onsojeiq IN

Jom], WosTT[I}U] (ues10y) asuodsoy

0661 ‘19IUN Jom], sspug Ioyeso1ojuy AWIy (Q/])ueWIAyH ay 7 onsojeiq IN 7 onsojelq IN

8861 (O)usisuq

‘uoa] 2 SO[J2oUOIseA, WVNVdS NVdSONG OSTV (O/I)ustueds ynduy ¢ IXOL 7 IxXO], popuayxy

8861 ‘79 Ja oeseN (O)usisuq

QQ6] ‘UINdO[S aprin JOJSueIT, [ROIXOT ({)osoueder yndu] S 1XOL, p }XO], popusaixq

ystueds-uewisy (O)ysysuq

8861 “WNd0[S TV.LAWN yousrq-Yyond os[y ([juew195 induy ¢ XO] Vv XO], popusixy

anbnig soIpnis

7861 * 7) 2 UIOpIoH ATLSIda afAyg /TeWTUR ID) (O/D)usisuq jynduy v sonbyi v XO], popusixy

L9LV SdIOA, 0} 1X9 podojaasq sqe] 119d (O/Dusisuq ynduy € yooods G XO], popucixy

[elo1sWIWIOD

sjONpOIg [IOMZINY JIOMZINY puq-3u014 YOO (O/Dusysuq ynduy € yoseds Z 1XO], popucixq

p86 ° 72 12 suUBARTY SNITEMA, AyyenO ystH (O/Dusisuq ynduy € yooods @ IXOL popuoixy

6861 yxoqyodAYy SUISSODOIg (axo}10d 474)

‘yore 2 usTeuUA JeuonesIoAUOZD IN sovenuits (O/Dussuq Xd, po103s fe yxoyIOdAY I AIONO ...1N,,

sjulog sojeIq

6861 “JINT 2? Yyouyorz SoIAPV ITV 32 UONSsT9g-nus|] (O/Dusisuq 1X9, Por0js G TOMSUY “IN (4 AlanO poyouysoy

IOUdIIJOY QUIeN WiaiskS soinjeoy O/] osensuey] .901N0§ 2SUISSIDO1g jndjno [2A9T ynduy

UONeWIOJUT jndjno sulsieg

jndjno 3

(ponuU0D) UONIUZOIIY Id10A VIP2yA-2]/SUIG UBY) 10YyjIO SUId}SAG JdeJIDJUT OSENSULT [VANjeN] DUIOS—"6 9]/qR],

110 LANCE A. MILLER

questions they would pose to a detainee, pictured on the display from a

videodisc player, according to a particular style of interrogation. The typed

input—in German, first, then Korean and Spanish—is passed to a natural

language component for analysis. Based upon the correctness and adequacy

of the question (as evaluated also by an intelligent ‘‘master tutor’ component),

a typed response or a digital pre-recorded audio output is generated, or else

the display can show a puzzled prisoner, not understanding, saying ‘“‘Bitte?”’

The projects which combine natural language input with other modalities—

particularly pointing and eye-movements—are also very exciting. These NLI

provide a much higher degree of naturalness than with natural language alone,

and this also extends to brevity. In Table 10 we identify 7 strategies for referring

in natural language to an element on a computer display. The simplest of

these is a simple statement like “‘here’”’ accompanied by a pointing gesture.

For natural language alone, especially for relative referencing expressions,

the length of the referring phrase—as well as the difficulty in processing it—

increases dramatically over the language plus gesture mode. In this combined

mode the display coordinates activated by the gesture only need to be coor-

dinated to the “closest”? referring expression in the natural language input.

Thus, a speech input spread sheet with touch capability could tie data cells

to processing instruction via commands like “Change this, this, and this to

zeroes’’, where in association with each ‘“‘this’” was a screen touch onto a

displayed data area.

These and the other tabled applications illustrate the health and vigor of

NLI development and the opportunity for improved computer-user interface

capabilities utilizing natural language facilities.

Remaining NLI Frontiers

We conclude by sketching four types of extensions to existing NLI which

could provide particularly useful application functions, as shown in Table 11.

The first two are predicated on the assumption that when a user is having

difficulty and needs to ask a question, natural language is the most immediate

and easiest means of supporting this need. The first, Meta-level Dialogues,

occur when the user wishes to talk about the application itself, not about

particular transactions within it. A major difficulty for supporting this exten-

sion within an existing NLI is that of recognizing that the user’s input is a

meta-level comment. Self-references and “‘you” references to the system may

be reliable markers, as may question markers like “why”. On the response

side, formulating reasonable bounds on the scope of the question and gen-

erating an appropriately general response are two major difficulties. Never-

NATURAL LANGUAGE INTERFACES 111

Table 10.—Natural Language Strategies for Referring to Elements on a Computer Display

Strategy Example of Natural Language Output

Absolute Location (pointing)

Absolute Coordinates

Relative Coordinates

“Right here”

“Fourth line, 12th column’, “Map coordinates A13”

“In the second paragraph, third line, 2nd word”

“About 2 inches below and left of center”

“The section headed ‘Performance’ ”

“Field 12 ‘Marital Status’ ”

‘The first word after ‘disengage clutch’ ”

“In the section for ‘occupation’; the second question”

“The rear door in the side view drawing”

“The third primary input line from the top”

“The area just above the main entrance”

‘““A point midway between the top of his upper lip and the bottom of

his nose.”

Labeled Text Target

Relative Text Target

Well-formed Graphics Target

(for an image/ graphic)

Relative Graphics Target

theless, such a facility would provide users the capability to bypass much

descriptive and instructional material and develop the appropriate concept of

the system’s goals when the need was recognized.

Situated Help refers to the situation in which a user asks for help with respect

to a particular problem situation she finds herself in, particularly when avail-

able help sources weren’t sufficient. We conjecture that the desire to use

Table 11.—Some Remaining Frontiers for Practical Natural Language Interface Applications

Application Precipitating Condition

Examples

User needs to understand systems,

goals, intents, overall conceptual

structures, and knowledge

limitations, etc.

User can’t get information she needs

from on-line help or from manuals,

wants to get assistance for this

particular situation

Meta-level Dialogues

Situated Help

In-Depth Interviewing System has detailed knowledge-

acquisition objectives re an

informed human source. System

needs to cover a lot of ground in

interview resulting in integrated

picture, with all contradictory areas

followed-up

User and system (or robot) share a

view of some extrinsic reality

(display screen, image, etc.) and are

performing cooperative examination

or manipulation of it. Need to not

only handle individual extrinsic

references but coordinate world

views.

Extrinsic-Referencing

Dialogues

‘Why are you asking me this?”

‘What kinds of things do you know

about?”

“How am I doing?”

“T can’t figure out how to get out

here!”

‘“‘What’s wrong with my command—

that it isn’t being accepted?”

‘How do I change this back?”

‘“‘How do you know when a pie is

done?”

“Is bread-baking similar to baking

pies?”

“T thought you said you can’t tell

when it’s ready just by looking at

it,?

“I think this doorway here (gesture) is

too wide. Make it a bit smaller. . .

No, that doesn’t look right, put it

back like it was.”

“Pan around to your left. Stop! OK,

continue. Wait, go back!

“Change the name to ‘Smith’

throughout. Now move the Ist

paragraph to follow the second.

Now, in the very first line. . .”

112 LANCE A. MILLER

natural language here is very strong, through frustration and the desire to get

on with the task. There are several major difficulties for this NLI possibility

also, both linguistic and extra-linguistic. The language problems are particu-

larly those of finding the exact referent in the referring expression, especially

as this may be a past action, not just an entity. In addition there will probably

be the choppy “emotional” style associated with writings in which the author

is expressing some kind of upsetment—use of exclamation marks, dashes,

ellipsis periods, etc. These will almost certainly not be interpretable. Outside

of the language problems there is the reasoning associated with determining

what the user’s goals were, her activities, and particularly her beliefs about

the ways things work in this application.

The third example deals with the possibility of having the computer act as

an interviewer, to instigate Knowledge Acquisition procedures with the user

as the source, to gain an understanding of a user domain of expertise (or

requirement). One application might be to develop completely knowledge-

based documentation for a system by interviewing the developers who played

various roles in building it. The language difficulties of such a NLI application

are manifold, including those of maintaining a coherent extended dialogue,

somehow recognizing all of the various topics, understanding the user’s qual-

ifications, and following accepted “‘implicatures”’ for this type of interaction.

Whereas the first two examples deal with two levels of help for the user,

and the third deals with helping the system acquire information, the fourth

possibility concerns cooperative interaction between machine—and robot—to

accomplish shared activities in the world. The most challenging problems here

are not linguistic but rather those concerning beliefs: beliefs concerning each

other’s goals and immediate intentions, beliefs about the capabilities and

available resources of the other, and beliefs about the present state of the

world—and whether the other shares the same beliefs!! Instigating and co-

ordinating belief statement and revision, particularly to maintain a shared

world model, appear to be far more difficult than the linguistic problems of

resolving referring expressions, communicating states, and issuing clear in-

structions.

It is a tribute to the remarkable progress in the natural language processing

field—and to the memory/performance features of today’s PCs and worksta-

tions—that suggestions for new NLI such as the above four can be so calmly

and un-selfconsciously described; it’s quite likely that they will be similarly

received. With such progress it is likely that the key technical limitations to

realizing these additional futuristic NLI won’t be found in the computational

or pure linguistic fields. Rather, they will probably come from the limitations

of theory, method, and findings concerning the psychology of communication

NATURAL LANGUAGE INTERFACES 113

and its intersection with pragmatics and semantics in various other fields.

What an exciting challenge!

References

Allen, T. Natural Language Understanding. Menlo Park, CA: Benjamin/Cummings Publishing Co., 1987.

Baker, J., and J. Baker. “Dragon Systems’’, In Proceedings of the Speech and Natural Language Workshop,

Philadelphia, February 21-23, 1989, 143.

Ballard, B. ‘“‘A General Computational Treatment of Comparatives for Natural Language Question An-

swering’’, In Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics,

Buffalo, June 7-10, 1988, 41-47.

Berwick, R., and A. Weinberg. The Grammatical Basis of Linguistic Performance. Cambridge, MA: MIT

Press, 1984.

Boisen, S., Y. Chow, A. Haas, R. Ingria, S. Roucos, R. Scha, D. Stallard, and M. Vilain. ‘Integration

of Speech and Natural Language: Final Report’, Report No. 6991, BBN Systems and Technologies

Corporation, Cambridge, MA, 1989, 111.

Byrd, R., N. Calzolari, M. Chodorow, J. Klavans, M. Neff, and O. Rizk. “Tools and Methods for

Computational Lexicology.”” Computational Linguisics, 1987, 13, 219-240.

Chow, Y., M. Dunham, O. Kimball, M. Krasner, F. Kubala, J. Makhoul, P. Price, S. Roucos, and R.

Schwartz. ““BYBLOS: The BBN Continuous Speech Recognition System’, IEEE ICASSP-87, paper

Sot 1987, 105.

Chow, Y., and S. Roukos. “Speech Understanding Using a Unification Grammar’’, In Proceedings of the

IEEE International Conference on Acoustics, Speech, and Signal Processing, Glasgow, Scotland, May

23-26, 1989, 247.

Cohen, P., M. Dalrymple, D. Moran, F. Pereira, J. Sullivan, R. Gargan, Jr., J. Schlossberg, and S. Tyler.

“Synergistic Use of Direct Manipulation and Natural Language,” In Proceedings of the Computer Human

Interaction Conference, Austin, Texas, April 30—May 4, 1989, 227-233.

Davis, J., and T. Trobaugh. “Direction Assistance”, Technical Report, MIT Media Technology Lab,

December, 1987.

Dowding, J. “Reducing Search by Partitioning the Word Network’’, In Proceedings of the Speech and

Natural Language Workshop, Philadelphia, February 21-23, 1989, 272-276.

Dowding, J., and L. Hirschman. “Dynamic Translation for Rule Pruning in Restriction Grammar’’, Pre-

sented at the 2nd International Workshop on Natural Language Understanding and Logic Programming,

Vancouver, B.C., Canada, 1987, 276.

Fischer, G., and H. Nieper-Lemke. ““HELGON: Extending the Retrieval by Reformulation Paradigm’’,

In Proceedings of the Computer Human Interaction Conference, Austin, Texas, April 30—May 4, 1989,

357-362.

Frohlich, D., and P. Luff. ‘‘Conversational Resources for Situated Action’, In Proceedings of the Computer

Human Interaction Conference, Austin, Texas, April 30—May 4, 1989, 253-258.

Glenn, F., III, H. Iavecchia, L. Ross, J. Stokes, W. Weiland, D. Weiss, and A. Zaklad. ‘“Eye-Voice-

Controlled Interface”, In Proceedings of the Human Factors Society—30th Annual Meeting, 1986, 322-

326.

Grishman, R. “A Very Brief Introduction to Computational Linguistics’, In Proceedings of the Speech

and Natural Language Workshop, Philadelphia, February 21-23, 1989, 37-52.

Heidorn, G., K. Jensen, L. Miller, R. Byrd, and M. Chodorow. “The EPISTLE Text-Critiquing System’’,

IBM Systems Journal, 1982, 21, 305-326.

Hemphill, C., and J. Picone. ‘““Chart Parsing of Stochastic Spoken Language Models’, In Proceedings of

the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 250-255.

Hendrix, G., and B. Walter. ‘The Intelligent Assistant’, Byte, December, 1987, 251-258.

Hobbs, J., W. Croft, T. Davies, D. Edwards, and K. Laws. ““Commonsense Metaphysics and Lexical

Semantics.”” Computational Linguistics, 1987, 13, 241-250.

Hovy, E. “Planning Coherent Multisentential Text”, In Proceedings of the 26th Annual Meeting of As-

sociation for Computational Linguistics”, Buffalo, June 7-10, 1988, 163-169. (a)

Hovy, E. “Two Types of Planning in Language Generation”, In Proceedings of the 26th Annual Meeting

of the Association for Computational Linguistics, Buffalo, June 7-10, 1988, 179-186. (b)

Ishikawa, H., Y. Izumida, T. Yoshino, T. Hoshiai, and A. Makinouchi. “KID Designing a Knowledge-

Based Natural Language Interface’, IEEE, Summer, 1987.

Joshi, A. “An Introduction to Tree Adjoining Grammars’’. In Manaster-Ramer, A. (Ed.), Mathematics

of Language. John Benjamins, Amsterdam, 1987.

114 LANCE A. MILLER

Kasper, R. “Systemic Grammar and Functional Unification Grammar’. In J. Benson and W. Greaves,

(Eds.), Systemic Functional Approaches to Discourse, Norwood, New Jersey: Ablex (in press). Also

available as USC/Information Sciences Institute, Technical Report RS-87-179, May, 1987.

Kay, M. “Parsing in Functional Unification Grammar”’. In D. Dowty, L. Kartunen, and A. Zwicky (Eds.),

Natural Language Parsing. New York: Cambridge University Press, 1985, 251-278.

Klavans, J., J. Nartey, C. Pickover, D. Reich, M. Rosson, and J. Thomas. ‘““WALRUS: High-Quality

Text-to-Speech Research System’’, In Proceedings of IEEE Speech Synthesis and Recognition, 1984,

19-28.

Lalit, R., F. Jelinek, and R. Mercer. “A Maximum Likelihood Approach to Continuous Speech Recog-

nition”, [EEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-5, No. 2, March,

1983.

Lane, A. “DOS in English’, Byte, December, 1977, 261-264.

Lee, K-F. “Automatic Speech Recognition: The Development of the SPHINX System’”’, Kluwer Academic

Publishers, Boston, 1989, 130.

Lee, K-F, H. Hon, and M. Hwang. “Recent Progress in the Sphinx Speech Recognition System’’, In

Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 22.

(a)

Lee, K-F, H. Hon, and M. Hwang. “Recent Progress in the Sphinx Speech Recognition System’’, In

Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 125-

130. (b

Lenat, a. M. Prakash, and M. Shepherd. ‘““CYC: Using Common Sense Knowledge to Overcome Brit-

tleness and Knowledge Acquisition Bottlenecks.” AJ Magazine, 1986, VI, 65-85.

Linebarger, M., D. Dahl, L. Hirschman, and R. Passonneau. ‘Sentence Fragments Regular Structures’’,

In Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics, Buffalo,

June 7-10, 1988, 7-16.

Maeda, H., S. Kato, K. Kogure, and H. lida. “‘Parsing Japanese Honorifics in Unification-Based Gram-

mar’, In Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics,

Buffalo, June 7-10, 1988, 139-146.

Mann, W. “An Overview of the Nigel Text Generation Grammar’, USC/Information Sciences Institute

Research Report RR-83-113. (a)

Mann, W. “Rhetorical Structure Theory: A Theory of Text Organization’, USC/Information Sciences

Institute Research Report. RR-87-190. (b)

McCord, M. “Modular Logic Grammars’’, American Journal of Computational Linguistics, 1980, 6, 31-

43.

McCord, M. ‘“‘Modular Logic Grammars”’, In the Proceedings of the 23rd Annual Meeting of the Association

for Computational Linguistics, Chicago, 1985, 104-117.

Miller, L. ‘Design of an Intelligent Tutor: The Bridge Tutor for German Language Training.” Internal

Report, Science Applications International Corporation, 1990.

Moore, R., F. Pereira, and H. Murveit. “Integrating Speech and Natural-Language Processing”’, In Pro-

ceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 243-247.

Murveit, H., M. Cohen, P. Price, G. Baldwin, M. Weintraub, and J. Bernstein. ‘““SRI’s Decipher System”’,

In Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21—23, 1989,

238-242.

Nagao, M., J. Tsujii, and J. Nakamura. ‘“The Japanese Government Product for Machine Translation’,

Machine Translation Series, J. Slocum (Ed.), Cambridge University Press, Cambridge, England, 1988.

Natural Language Inc. Natural Language Retrieval of Database Information, 1986.

Neal, J., and S. Shapiro. “Intelligent Multi-Media Interface Technology”. In Architectures for Intelligent

Interfaces: Elements and Prototypes, Monterey, California, March, 1988.

Neuburg, E. ‘Speech Tutorial’, In Proceedings of the Speech and Natural Language Workshop, Phila-

delphia, February 21-23, 1989, 53-71.

Olive, J., and M. Liberman. Text to speech—An overview. Journal of the Acoustical Society of America,

Suppl. 1, 78(3):s6, Fall 1985.

Pareschi, R. “A Definite Clause Version of Categorical Grammar”’, In Proceedings of 26th Annual Meeting

of the Association for Computational Linguistics, Buffalo, June 7-10, 1988, 270-277.

Pereira, F., and D. Warren. ‘Definite Clause Grammars for Language Analysis—A Survey of the For-

malism and a Comparison with Transition Networks”’, Artificial Intelligence, 1980, 13, 231-278.

Price, P., W. Fisher, J. Bernstein, and D. Pallett. ““The DARPA 1000-Word Resource Management

Database for Continuous Speech Recognition”, Proc. ICASSP, New York, April, 1988, 255.

Sager, N. Natural Language Information Processing: A Computer Grammar of English and Its Applications,

Reading, MA; Addison-Wesley, 1981.

Salton, G. ‘Syntactic Approaches to Automatic Book Indexing”, In Proceedings of the 26th Annual

Meeting of the Association for Computational Linguistics, Buffalo, June 7-10, 1988, 204-207.

NATURAL LANGUAGE INTERFACES 115

Selfridge, M. “Integrated Processing Produces Robust Understanding”, Computational Linguistics, Volume

12, Number 2, April-June 1986, 89-106.

Seneff, S. ““TINA: A Probabilistic Syntactic Parser for Speech Understanding Systems”, In Proceedings

of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 168-178.

Slocum, J. (Ed.), Machine Translation Series. Cambridge, England: Cambridge University Press, 1988.

Tenant, H., K. Ross, R. Saenz, C. Thompson, and J. Miller. ““Menu-based Natural Language Under-

standing”, In Proceedings of the 21st Annual Meeting of the Association for Computational Linguistics,

Cambridge, Massachusetts, June, 1983, 151-158.

Vasconcellos, M., and M. Leon. ““SPANAM and ENGSPAN: Machine Translation at the Pan American

Health Organization”, Machine Translation Systems, J. Slocum (Ed.), Cambridge University Press,

Cambridge, England, 1988.

Wahlster, W. “User and Discourse Models for Multimodal Communication”, Architectures for Intelligent

Interfaces: Elements and Prototypes, J. Sullivan and S. Tyler (Eds.); Palo Alto, CA; Addison-Wesley,

1989.

Webber, B. “Discourse. Deixis: Reference to Discourse Segments’, In Proceedings of the 26th Annual

Meeting of the Association for Computational Linguistics, Buffalo, June 7-10, 1988, 113-122.

Webber, B., and T. Finin. “In Response: Next Steps in Natural Language Interaction”, in Artificial

Intelligence Applications for Business, W. Reitman (Ed.), Norwood, NJ, Ablex, 1984, Chapter 12, 211-

234.

Weimer, D., and Ganapathy, S. ““A Synthetic Visual Environment with Hand Gesturing and Voice Input.”

In Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989,

235-240.

Weischedel, R., R. Bobrow, D. Ayuso, and L. Ramshaw. “Portability in the JANUS Language Interface”’,

In Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989,

112-117.

Whalen, T., and A. Patrick. “Conversational Hypertext: Information Access through Natural Language

Dialogues with Computers”, In Proceedings of the Computer Human Interaction Conference, Austin,

Texas, April 30—May 4, 1989, 289-292.

Winograd, T. “Language as a Cognitive Process”, Reading, MA: Addison-Wesley, 1983, 255.

Woods, W. “‘Cascaded ATN Grammars”, American Journal of Computational Linguistics, 1980, 6, 1-12.

Young, S. ‘““The MINDS System: Using Context and Dialog to Enhance Speech Recognition’, In Pro-

ceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-23, 1989, 131-136.

Zue, V., J. Glass, M. Phillips, and S. Seneff. ‘Acoustic Segmentation and Phonetic Classification in the

SUMMIT System’’, Presented at the International Conference on Acoustics, Speech, and Signal Pro-

cessing, Glasgow, Scotland, May 23-26, 1989, 276. (a)

Zue, V., J. Glass, M. Phillips, and S. Seneff. ““The MIT SUMMIT Speech Recognition System: A Progress

Report’, In Proceedings of the Speech and Natural Language Workshop, Philadelphia, February 21-

23, 179-189. (b)

Journal of the Washington Academy of Sciences,

Volume 80, Number 3, Pages 116-137, September 1990

Knowledge Representations Used

By Computer Programmers!

Scott P. Robertson

Psychology Department, Rutgers University, Busch Campus,

New Brunswick, NJ 08903

ABSTRACT

Composition and modification of computer program code are important skills requiring

considerable cognitive effort. Psychologists have begun to study these skills both in order

to understand complex cognitive phenomena and to contribute to the design of program-

ming tools and aids. Research on the cognitive representations that programmers use is

discussed. Many of the parallels between research and theory in natural language com-

prehension and programming language comprehension are highlighted. The procedural

nature of program text and the fact that programmers are highly goal-directed makes for

interesting contrasts with natural language. We describe our own work in this area which

has led us to the opinion that code comprehension and modification are best understood

in a problem-solving framework. The paper concludes with suggestions for programming

tools that aid abstract reasoning about how to solve problems, not specific reasoning about

how to write code.

Introduction

A computer program is a text, referred to as “‘code,”’ written in a carefully

formalized language, describing a set of instructions for a machine to follow.

Often a program is accompanied by documentation, written in a less carefully

formalized language, explaining to humans what the code ‘‘means.” A pro-

gram may also be accompanied by graphical information like flow charts or

call structure diagrams that represent other aspects of its functionality.

We routinely accept the idea that a computer translates code into internal

structures that are useful for guiding the machine through desired states.

Considerable effort goes into the design of languages and physical architectures

for the representation and realization of procedures that computers carry out.

On the other hand, we have paid less attention to the representation of these

same procedures by human beings. This despite that fact that the generation

and modification, not to mention the underlying purposes, of computer pro-

grams still originate with humans.

116

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 117

The mandate for understanding the representational structures that humans

use belongs to cognitive science, especially cognitive psychologists. Recently

researchers in this field have turned their attention toward computer pro-

gramming. There is much to be learned in this effort from prior research on

natural language comprehension. In this study the knowledge utilized by pro-

grammers when they work with code is examined. In many places contrasts

are made between the issues that arise in the study of natural language com-

prehension and computer program comprehension.

The first section discusses the representation of program statements and

the processes involved in comprehending individual lines of code. This may

be referred to as the “‘microstructure”’ level of representation. The next part

covers the representation of aggregates of code. This is commonly referred

to as the “macrostructure” level of representation. The third section treats

knowledge outside of the programming language itself. This includes knowl-

edge in the task domain and memories of other programs. The last section

discusses the role of programmers’ goals and strategies on representation and

manipulation of code. The conclusion contains ideas for new directions in

programming tools that support reasoning using the knowledge structures

discussed throughout the paper.

Microstructure Representation

The basic unit of natural language is a sentence and the basic unit of a

program is a line. In natural language a simple sentence expresses a single

concept or relation. Complex sentences may contain several elementary con-

cepts. Cognitive psychologists use propositions to represent simple concepts.

A proposition is centered around a relational term that describes the associ-

ation of several objects to each other. The proposition is derived from rules

mapping symbols in the natural language to internal symbols in the represen-

tation language.

As an example, consider the sentence “John gave a sandwich to Bill.”’ In

natural languages verbs typically provide the relational term around which

propositions are built, so a proposition for this sentence is derived from a

general form like the following:

(GIVE, actor, object, recipient).

This form serves as a representational frame for all sentences of the same

type. Thus, the example sentence would be represented as:

(GIVE, John, sandwich, Bill).

118 SCOTT P. ROBERTSON

The advantage of propositional notation is that it allows sentences with

different surface forms to be represented conceptually the same way. This is

one theoretical way of handling the problem of synonymy. More importantly,

however, this permits general rules to be written that utilize the conceptual

information (memory searching rules, for example). If surface forms were

maintained in the internal representations of concepts, then rules for reasoning

and memory search processes would have to be specific to each possible

representational variation.

Schank (1972) has claimed further that there are only a few canonical forms

underlying the microstructure of a natural language. Thus, in Schank’s con-

ceptual dependency theory there is a primitive action (called “ATRANS’’)

that represents any transfer of possession as the proposition:

(ATRANS, actor, object, recipient)

regardless of the verb used to express the concept. This allows all of the

following sentences to be represented by the same logical form:

John gave a sandwich to Bill.

John gave Bill a sandwich.

Bill was given a sandwich by John.

John handed a sandwich to Bill.

Bill got a sandwich from John.

Evidence for propositional representations in natural language comes from

several sources. One is confusion errors in recall for synonymous surface forms

(Flores d’Arcais, 1974; Sachs, 1967). A subject who reads ‘“‘Mary was given

a rose by John” may mistakenly identify “John gave a rose to Mary” or ‘““Mary

got a rose from John” on a later recognition test, for example. In a study by

Anderson (1974), subjects asked to judge whether a sentence meant the same

thing as a previous sentence initially did so more quickly when the two sen-

tences matched verbatim. After 2 minutes, however, a sentence with a dif-

ferent surface form but the same meaning was judged just as quickly as a

verbatim match. Many researchers agree that information about the surface

features of a sentence are encoded and persist for a limited time in short term

memory but that only propositional forms are encoded in long term memory.

Other evidence for propositional representations in natural language comes

from reading-time studies. In natural language reading times are predictable

partly by the number of underlying propositions that are contained in a sen-

tence. The sentence ‘“The car lurched and chugged”’ contains two propositions,

one for “The car lurched” and one for ‘“The car chugged.”’ Controlling for

sentence length, Kintsch (1974) and Kintsch and Keenan (1973) have shown

a linear relationship between the number of propositions in a sentence and

the reading time for that sentence.

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 119

Is there a propositional microstructure for programming languages? There

are two reasons that this is a difficult question to answer. First, by design,

programming languages are economical and it is rare to find more than one

way of expressing a basic programming action. We might expect that a prop-

ositional form exists for assignment, for example:

(ASSIGN, variable-name, value),

but we would be hard pressed to find more than one way of expressing this

in any one language. In a sense, the highly constrained syntactic rules for a

programming language constitute its microstructure.

A second reason that it is hard to specify the microstructure of a line of

code is that the line represents a procedure, not just a static concept. Mayer

(1987) argues that a line of code is represented conceptually as a set of “‘trans-

actions” to be carried out by the machine on which it will run. The transaction

steps are determined by the programmer’s mental model of the machine ar-

chitecture and rules. For example, a statement of the form

variable-name = value, (e.g., A = 0)

would be represented as the following sequence of transactions:

(1) Find the value in the expression (e.g., read 0).

(2) Store the value in temporary memory (e.g., store 0).

(3) Find the value currently in the memory location of variable-name (e.g., find the

current value of A).

(4) Erase the value in that memory location.

(5) Store the value now in temporary memory in the memory location associated with

variable-name (e.g., associate 0 with A).

(6) Move to the next line.

(7) Do what the next line says.

Mayer (1987) has performed a reading time experiment on very short (3-

line) BASIC programs, and Dyck & Auernheimer (1989) have generalized

the experiment to Pascal. In Mayer’s study subjects were shown a few lines

of BASIC and the reading times were measured for each line. The number

of “transactions” that were implied by each statement was a significant pre-

dictor of reading time, accounting for 56% of the variance in their data.’

We recently completed a reading time study for a long (135 line) Pascal

program and our findings pose several questions (Robertson & Davis, 1990).

One thing that we noticed is that “‘reading,” in the sense that we use the term

for natural language, is not what programmers do when they look at lines of

code. When allowed to search through the code in any manner they choose,

programmers move as easily backward through the code as forward. They

return to lines over and over. They spend too little time on some lines to even

120 SCOTT P. ROBERTSON

process the elementary features of the text, and they spend so much time on

other lines that they could have studied a whole procedure instead.

We ran two groups of subjects, showing each the 135 lines of the program.

One group saw the lines in a scrambled order while the other saw them in

the coherent order of a program. We wanted to contrast these two situations

because we reasoned that the reading times for the scrambled code would be

‘microstructure only” reading times, while the reading times for the coherent

code would reflect other comprehension processes.

We used a simple model for deriving the code microstructure. The model,

like Kintsch’s simple model for natural language, assumes that a proposition

is built for each concept in the line. The program line:

mean: = sum/n

can be represented by two propositions, one assignment proposition:

(ASSIGN variable-name, value)

and one arithmetic operator proposition:

(DIVIDE, variable-1, variable-2).

In our example one proposition is embedded within the other as follows:

(ASSIGN (mean, (DIVIDE, sum, n))).

In order to construct this proposition in memory, elementary concepts must

be constructed for each variable (mean, sum, and n) and each operator (AS-

SIGN and DIVIDE). Other items that appear in lines that might be relevant

to propositions are command names (function calls, conditional statements,

etc.) and delimiters (parentheses, semi-colons, etc.). From this simple model

we quantified the number of steps required to construct each component of

the proposition for each statement in the program. We used each component

of the proposition as a separate predictor as follows:

Reading time = #variables + #command names + #operators +

#delimiters°

We predicted the reading times for each subject separately and here report

the means across subjects.

This simple equation accounted for an average of 54% of the variance in

reading times across subjects who saw the scrambled lines, comparing favor-

ably with the 56% that Mayer found using the more complex transaction

analysis. However, the equation accounted for only 23% of the variance in

reading times for the group which studied coherent code.* While the regression

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 121

equation is a significant predictor in both cases, it is clear that much more

variance is left unaccounted for when real code is being read. A closer look

at what programmers were doing helps to explain this.

When programmers study a program they do not read through it in order

as they would a text. Rather they skip back and forth through the code and

view lines multiple times. In our study, the average line was looked at 5.6

times with some lines being returned to as many as 13 times. While subjects

who read the scrambled lines spent an average of 6 minutes and 20 seconds

in the experiment, the subjects who studied the program spent an average of

50 minutes and 16 seconds. Remember that both groups looked at the same

135 lines of code.

Virtually every subject who studied the coherent program made repeated

regressions in the code, re-reading sections. When we categorized encounters

with a line of code according to which direction the subject was moving when

he or she read it, we discovered a major source of variation in reading times.

We divided line encounters into four major categories. If a subject moved to

a line from the previous line and then moved on to the next line, this was a

forward encounter. If a subject was moving backwards through the code,

arriving at a line from the subsequent line and then moving on to the previous

line, this was a backward encounter. If a subject arrived at a line from the

previous line but then returned to the previous line again on the next move,

this was a forward-to-backward switch. If a subject was moving backward and

arrived at a line from the subsequent line and then returned to the subsequent

line on the next move, this was a backward-to-forward switch. Table 1 shows

the average number of moves observed in each category. Seventy-two percent

of the moves were forward moves, like normal reading of prose, seventeen

percent were backward moves, and about eleven percent were switches in

direction. Sometimes subjects read a few lines of code, switched direction and

backed up to an earlier line, switched directions again and then read through

the code a second time. We called these sequences of movements “episodes”

and further subdivided our reading times into “within episode” and “‘between

episode” sequences.

Table 2 shows the mean reading time per syllable for lines of code in each

Table 1.—Mean number of moves per category and proportion of total moves in each category

Movement Category

Forward-Backward Backward-Forward Total moves

Forward Backward Switch Switch per subject

Mean 593.2 138.0 44.0 44.0 819.2

Proportion .724 .168 053 053 1.0

122 SCOTT P. ROBERTSON

Table 2.—Mean reading times per syllable (ms) for movement categories within and between episodes

Movement Type Between Episodes Within Episodes

Forward 408 —

Backward 141 146

Forward-Backward Switch 1581 1027

Backward-Forward Switch 569 410

First forward pass Hs 387

Second forward pass == 225

movement category and each episode category. This subdivision of reading

times revealed that very different processes were taking place in each move-

ment category. In particular, the long times at switches suggest that consid-

erable cognitive processing was taking place when switching occurred, espe-

cially forward-to-backward switching. Also, it is clear from the rapid reading

rates in the backward category that a very elementary analysis was taking

place during backward movements.

Table 3 shows the application of our simple microstructure equation to

reading times in the different movement categories. The mean proportion of

variance across subjects accounted for by the equation is shown for each

category. Also shown are the numbers of subjects (out of five) for whom

various components of the equation were significant (p < .05). That number

is marked with an asterisk when the predictor was significant for a majority

of the five subjects.

The proportion of variance accounted for by the equation was highest for

the subjects who saw scrambled lines, and all of the components except de-

limiters were significant for the majority of subjects in this group. We take

this as evidence that a straightforward translation of the lines of code into an

underlying propositional form is taking place.

Table 3.—Proportion of variance accounted for (R’) by the microstructure equation in both statement

presentation conditions and different movement categories. Also shown are the number of subjects (out

of 5) for whom each predictor was significant (V = variables, C = commands, O = operators, D =

delimiters). Asterisks indicate that the predictor was significant for a majority of subjects. The double

asterisk indicates a negative coefficient for that predictor.

Predictor

Type of Statement Presentation R? Vv C O D

Scrambled Lines 54 2 4* 3° 1

Coherent Program

Movement Categories Between Episodes

Forward AT, 53 2 3* sf

Backward cae, Sih 2 4** 0

Movement Categories Within Episodes

First Forward Pass .34 2 1 4* 1

Backward ait 5° 0 1 1

Second Forward Pass 22 1 2 3* 1

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 123

The proportions of variance accounted for by the equation when predicting

reading times in the coherent program group were all less than the scrambled

condition. Further, different components of the equation were significant in

the coherent program condition, and these components varied widely across

movement types. Variables and operators were significant for a majority of

subjects in the non-episode reading times, suggesting that cognitive processes

occurring during this activity were more like those in the scrambled condition,

i.e., more like normal reading. Within episodes, however, variables were not

as consistently important.

We view this as evidence that the cognitive processing taking place at each

line varies considerably depending on the overall goal of the programmer at

the time. Programmers’ goals vary as evidence is gathered about how the

program works. In a later section we will discuss our view that comprehension

of programs is problem-solving behavior, not normal reading, and present

further data on this point. First, however, we turn from a discussion of the

representation of lines to the representation of chunks of code.

Macrostructure Representation

In natural language a paragraph or a story is more than the sum of its

sentences. Propositions derived from sentences are connected to each other

in a meaningful way, and inferences are generated during comprehension that

provide a complex memory structure in which to embed microstructural ele-

ments.

Complex memory structures are built by the generation of local inferences

to connect stated elements. For example we easily see that the following two

sentences are related:

John was hungry.

John went to a diner.

We would say that the two concepts are associated in memory by a “‘Reason”’

relation since we know that John is going to the diner because he is hungry.

Macrostructure relations can also be represented propositionally. The fol-

lowing proposition serves as a frame into which other propositions can be

embedded if proposition-1 is a reason for proposition-2:

(REASON, proposition-1, proposition-2).

Thus the two sentences about John can be turned into propositions and embed-

ded in the following instantiation of the REASON frame:

(REASON, (HAVE, John, hunger), (GO, John, diner)).

124 SCOTT P. ROBERTSON

Several theories of natural language comprehension hold that representa-

tions of connected prose can be thought of as hierarchical structures consisting

of related macrostructure elements at superordinate nodes with microstructure

elements at the terminal nodes (Graesser, Robertson, Lovelace, & Swinehart,

1980; Kintsch, 1976; Kintsch & Keenan, 1973; vanDijk & Kintsch, 1983).

Recall experiments have tended to support this hypothesis, showing for ex-

ample that memory is better for superordinate information than for subor-

dinate information (Graesser, et al., 1980).

Complex propositions at the macrostructure level capture inferences about

the relations among stated propositions. The source of such inferences has

been of considerable interest to researchers in natural language comprehen-

sion. Scripts, plans, and goals (Schank & Abelson, 1977; Seifert, Robertson,

& Black, 1985; Warren, Nicholas, & Trabasso, 1979) are considered to be

common sources of pragmatic inferences.°

A script is a knowledge structure that contains a sequence of the actions

that constitute a common activity. People acquire many scripts and then use

this knowledge to guide their actions and comprehension of actions when they

are in script-like situations.

Bower, Black, & Turner (1979) asked subjects to describe what typically

happens when the subjects performed several different activities like going to

a restaurant or going to the doctor. For each of these activities there was a

core set of actions that virtually every subject mentioned. Subsequent recall

and recognition experiments showed that these core activities were present in

subjects’ memory representations of script-based stories even when they had

been left out of the stories. A reading-tme study, also by Bower, et al. , showed

increased reading time for script actions when a prior action was missing. This

was taken as evidence that inferences about script-based actions are generated

when necessary during comprehension. These inferences connect the micro-

structure of a story in the final cognitive representation.

Reading-time data from Seifert, Robertson, & Black (1985) provided evi-

dence that inferences about goals and plans are also routinely made during

comprehension. Recognition data from the same study showed that these

inferences become part of the memory representation.

In programs there are also links between statements, and in fact the local

inferences about how statements are connected are often more reliable in

programming languages than in natural languages. For example, a BEGIN

statement in Pascal will always be accompanied by an END statement, a FOR

statement in Basic will always be accompanied by a NEXT statement, and so

on. Compilers make use of these mandatory contingencies to recognize simple

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS _ 125

syntax errors. Also, some program editors make use of these rules by providing

the programmer with the complementary statements. The boundaries of code

searching episodes found by Robertson & Davis (1990) very often consisted

of these procedurally related statement pairs.

What is interesting about connections among lines of code is that they

achieve code functions. The FOR-NEXT construction in Basic or the DO-

UNTIL construction in FORTRAN achieve the iteration function, and this

function must be accompanied by loop control processes. All programming

languages contain constructs for iteration and all must have a way of indicating

the scope of the iteration and controlling the number of iterations. This general

knowledge about how programs do things has come to be called ‘“‘plan” knowl-

edge (Ehrlich & Soloway, 1984; Rist, 1986; Robertson & Yu, 1990). Several

researchers suggest that good programmers acquire a repertoire of plans that

they can access to comprehend and design code (Adelson, 1981; Guindon,

1990).

Rist (1989) characterizes programming plan knowledge by identifying the

focus calculation, goal output, and extension initialization of common plans.

For example, the running total plan has an accumulation operation as its focus

calculation

(e.g., count: =count+ 1, in Pascal)

the value of the accumulating variable as its goal output (e.g., the value of

count), and an assignment operation as its extension initialization (e.g., count: = 0,

in Pascal). Rist (1989) presents verbal protocol data in support of the view

that programmers use plans in the design of code.

In a recent study (Robertson & Yu, 1990) we attempted to show that

programming plans were abstract knowledge structures that were not specific

to a language. We asked programmers to read several programs, divide them

into meaningful ‘“‘chunks,” and provide a verbal label for the chunks. Then

we asked them to sort the programs into groups—placing those that seemed

to work the same way into the same group. We had written the programs to

do many different things, from simulating a calculator to running a psychology

experiment, but we used three distinct program schemas—or “‘plans.’’ One

plan, for example, was to show a menu, wait for a selection, act on the

selection, and display a result.

Our subjects chunked the programs which were in the same plan groups in

the same way, provided similar labels within those groups, and sorted the

programs into groups according to the plans. Interestingly, the experiment

was run using both FORTRAN and Pascal code, and the results were the

126 SCOTT P. ROBERTSON

same regardless of the programming language. We interpreted this as evidence

that plan knowledge for programming is more abstract than knowledge about

the programming language (Adelson, 1981).

A second group of subjects was given the verbal labels that the first group

of subjects had provided (both FORTRAN and Pascal labels mixed together)

and was asked to sort them into similar categories. A clustering analysis of

their sorting data is shown in Figure 1. Programs that were in the same plan

group are indicated by groupings within parentheses just below the Figure’s

abscissa (there were three plan groups). “‘F” programs represent FORTRAN

programs while “‘P” programs represent Pascal programs. It is evident that

the subjects perceived the plan groupings across both tasks and languages.

Only two programs, F2 and P8, out of eighteen were out of place.

To summarize, program representations include the relations between state-

ments. These relations encode procedural associations between lines of code

and are generated by inferences based on knowledge of common programming

2 | _ Dimension 1

ch D

o c| =

o 5

° e 0 Ig o

Dimension 2 al lg he

-1

2 1 0 \ 2

Fe P8

(F4 PS F6 FS P6 P4) (F7 F8 F9P9P7) (P3_ Fi F3 Pl P2)

RCP RTRDP MGOM

PLAN PLAN PLAN

Fig. 1. Clustering analysis of sorting data from the verbal labels given to Pascal and FORTRAN programs

in three plan groups (after Robertson & Yu, 1990).

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 127

constructs like plans. These procedural constructs are more abstract than

knowledge about a specific programming language and develop as program-

ming expertise increases.

Task and Function Representations

Another aspect of representation of both natural and programming lan-

guages concerns information that is entirely distinct from the text. For natural

language this includes world knowledge, or context, which can have a signif-

icant influence on the perceived meaning of sentences. For programming

languages, this includes the task domain or the problem that the program is

designed to solve.

Several lines of research illustrate this point for natural language compre-

hension. These include research on ‘“‘advance organizers” (Barnes & Clawson,

1975; Mayer, 1976, 1979b), the role of titles or context statements (Bransford

& Johnson, 1973), and comprehension of conventional language usages like

indirect requests and idioms (Clark, 1979; Gibbs, 1984, 1986; Gibbs & Mueller,

1990).

An advance organizer is a text item, like a title, outline, or diagram, that

is presented to a comprehender before a text is read. Advance organizers

have considerable influence on the interpretation that readers have of sub-

sequent text. In an extreme example, Bransford and Johnson (1973) con-

structed stories that were incomprehensible to subjects when advance organ-

izers were not presented, but which seemed mundane and easy to understand

when they followed a clarifying picture or title. In other studies, the use of

outlines, summaries, questions or other extraneous material as accompani-

ments to text enhanced memory for important text information (Anderson,

1980; Anderson & Biddle, 1975) and increased problem-solving ability (Brans-

ford & Franks, 1976).

It is possible that very basic comprehension mechanisms may be affected

by prior information. Psychologists have long been interested in the processes

underlying indirect language usage (e.g., ‘Do you have a watch?” as a request

for the time of day). Reading-time studies for indirect requests have shown

that they take longer to understand than direct requests and a two-step com-

prehension model has been proposed to account for this result (Clark, 1979).

In this model the literal meaning of a sentence is first determined. If the literal

meaning does not make sense or violates what Grice (1975) called a conver-

sational postulate, then the comprehender attempts to determine a possible

non-literal meaning.

128 SCOTT P. ROBERTSON

Research by Gibbs (1984, 1986), however, has shown that reading-time

increases for indirect language usages disappear when the context suggests a

conventional, but non-literal interpretation. For example, if the sentence ‘“‘Do

you have a watch?” is preceded by a sentence like “‘Mary didn’t know what

time it was so she stopped a friend,” it is read quickly, interpreted non-literally

(i.e., as a request for the time), and no apparent ambiguity is noticed. Thus

it appears that the basic sentence understanding mechanism can be affected

by prior context.

Advance organizers have an effect on comprehension because they allow

comprehenders to activate relevant knowledge structures that can then be

used to represent and elaborate the text information most effectively. This

occurs because top-down processes play an important role in comprehension.

There is every reason to believe that this is true of program comprehension

as well. The problem is to determine which types of advance organizers will

be most useful.

Mayer (1979a, 1981) has studied the role of advance organizers in the

acquisition of programming knowledge. Again, the procedural nature of code

changes the form of what makes an effective advanced organizer. Mayer

reasoned that subjects would learn a programming language more quickly if

they had a ‘“‘mental model” of the device on which the code would run. Subjects

who studied such a model (a metaphorical description of a computer system

as a combination blackboard and filing system) were able to pick up the Basic

programming language more quickly and use it with fewer errors.

Fitter & Green (1979) have explored the types of diagrams and illustrative

materials that are useful to programmers. They suggest that auxiliary material

should help programmers focus on the information that is relevant to their

needs and, most interestingly, that the perceptual code of the material should

match the representational code that will be used by the programmer to solve

problems.

Several researchers have focussed on the form of the information given to

programmers (Brooke & Duncan, 1980; Cunniff & Taylor, 1987; Kammann,

1975; Ramsey, Atwood, & VanDoren, 1983; Sheppard, Kruesi, & Bailey,

1982; Shneiderman, 1982; Shneiderman, Mayer, McKay, & Heller, 1977),

with the most common form being some type of graphical representation like

a flowchart. In general, this material is helpful to programmers when they

use it, but a recent study in my laboratory (Koenemann, 1990) suggests that

programmers spend as little time as possible with material extraneous to the

code itself.

A twist on the notion of using graphical representations as aids to code

comprehension is to represent the program itself graphically instead of tex-

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 129

tually. Cunniff & Taylor (1987) report that programs written in a graphical

programming language called FPL are comprehended more quickly and ac-

curately than programs written in Pascal. The significance of work of this

nature will increase as object-oriented programming becomes more common

and as programmers begin to appear who have never had experience with

text-based programming languages.

In an interesting verbal protocol study of program comprehension, Pen-

nington (1987a) observed that programmers who attained a high level of

comprehension were concerned both with the program structures and with

the application domain and that they mixed study of both aspects of the

program. In an explicit attempt to relate code comprehension to the prose

comprehension studies of Kintsch (1986) and VanDijk & Kintsch (1983),

Pennington suggested that programmers build two distinct models. One, the

program model, consists of code microstructure and macrostructure and rep-

resents the procedural detail of the program. A second, the situation model,

consists of information about the real-world objects that the program manip-

ulates, the real-world consequences of program actions, and the functional

properties of the program in the task domain. The two models must be “‘cross-

referenced” so that reasoning can occur easily about the correspondence of

parts of both models. Later Pennington (1987b) elaborated her model and

suggested that programmers first build the procedural representation from

their knowledge of programming conventions. They then use this represen-

tation to help them understand the functional characteristics of the code and

its relation to the task domain.

To summarize, programmers represent more than the microstructure and

program-based macrostructure of code. Their conceptual representations of

programs also include knowledge about the task domain and other functional

characteristics of the code. A detailed model of the processes that generate

this knowledge is lacking, although it is clear that presenting programmers

with information that helps them conceptualize these aspects of code func-

tionality is worthwhile.

Goals of the Programmer

A final issue that is important in understanding code comprehension is the

nature of the programmer’s goals. A programmer may be inspecting code for

bugs, reading code to get an idea, searching code in order to make a modi-

fication, and so on. Each of these goals leads to a different strategy for code

comprehension. Again, research in text comprehension has shown that sub-

130 SCOTT P. ROBERTSON

jects make different kinds of inferences, and hence end up with a different

representation of a text, when their reading goals differ (Frederiksen, 1975).

Some researchers (Jeffries, 1982; Nanja & Cook, 1987) have suggested that

programmers who have a specific goal in mind, modification for example,

read the text of the code for comprehension first. Under this view a more-

or-less complete representation of the program is built before problem-solving

on the programming task begins. Others, however, have noticed that pro-

grammers read code strategically when they have specific goals in mind and

this has led to the view that they may build only partial representations as

necessary for achieving their goals.

In a recent study Koenemann (1990) observed the behavior of programmers

as they sought to make several modifications in a very long Pascal program.

He found that programmers only looked at between 12% and 43% of the

lines of code when their task was to modify the code. Further, the particular

lines studied varied considerably within the same program depending on the

modification task. He proposed that programmers follow an opportunistic

relevance strategy, trying to determine which parts of the code they need to

understand in order to make their modification and then only studying those

parts. Littman, Pinto, Letovsky, & Soloway, (1986) made a similar obser-

vation, noting that good programmers pursue what they called an as-needed

strategy in comprehension. Interestingly, novice programmers in that study

utilized a more comprehensive strategy than experienced programmers did to

understand the code.

Several other researchers have noted the strategic nature of code inspection

and comprehension when the programmers have clear goals (Letovsky, Pinto,

Lampert, & Soloway, 1987; Littman, Pinto, Letovsky, & Soloway, 1986; Myers,

1978; Weiser, 1982). This has led to a general consensus that we must begin

to understand the problem-solving goals of programmers in order to charac-

terize the processes that guide their actions (Gray & Anderson, 1987; Guin-

don, 1990; Letovsky, 1986).

As the result of several of our own studies of programming knowledge and

programmer behavior we have begun to view program comprehension, and

especially modification, from a problem-solving perspective. This perspective

stresses the importance of programmers’ goals, prior knowledge, and strategic

decision-making processes. Programmers seldom, if ever, just “read for com-

prehension.”

In order to help us understand programmers’ strategies more clearly, we

asked a group of programmers to repeat the experiment reported above (Rob-

ertson & Davis, 1990) in which a 135-line Pascal program was inspected. We

again collected data on where subjects looked in the code and how they

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 131

searched, but we asked them also to stop periodically (at times that they

determined) and explain what they were doing.

We categorized the programmers’ movements according to the scheme de-

scribed in Robertson & Davis (1990) and Table 4 shows the proportion of

movements in each category (a comparison with Table 1 shows that the dis-

tribution of movement types was similar in this study to the earlier study).

What is of interest is the distribution of comments across these movement

categories.

The five programmers in this study provided us with 182 comments on their

activities. By assuming that the comments should be evenly distributed across

the movement types by chance, we determined the expected frequencies of

comments in each of the movement categories based on the proportion of

movements in each category (from Table 4). Table 5 shows the expected

frequencies and the observed frequencies of comments that occurred in each

movement category. There were many more comments than expected (by

chance) that occurred in conjunction with switches in direction. Together with

the observation that line reading times tend to be very high at these positions

(Table 2), a consistent explanation is that considerable problem-solving activity

is associated with changes in direction in the search sequences.

We were able to categorize the programmers’ verbal comments into six

groups: Analyze, Assume, Question, Answer, Function, and Strategy. An

analyze comment was one in which the programmer offered an explanation

of a code segment. An assume comment was one in which the programmer

offered a prediction about what was coming up. A question was a query about

the code. An answer was a statement that could be clearly linked to an earlier

question. A function comment was a statement about what the code did

functionally. A strategy comment was a statement about what the programmer

planned to do next, usually where they wanted to go in the program or what

kind of information they wanted to find out.

Table 6 shows the proportion of comments of each type that occurred in

the movement categories defined for this experiment. The two most frequent

Table 4.—Mean number of moves per category and proportion of total moves in each category for

subjects who made comments

Movement Category

Total

Forward-Backward Backward-Forward moves

Forward Backward Switch Switch per subject

Mean 774.8 307.6 78.0 78.0 1238.4

Proportion .626 .248 .063 .063 1.0

132 SCOTT P. ROBERTSON

Table 5.—Expected versus observed frequencies of comments in each movement category. Expected

frequencies reflect a chance distribution. A chi-square test suggests that comments were not distributed

among the movement categories by chance, x? (3) = 138.28, p < .001.

Movement Type

Forward-Backward Backward-Forward

Forward Backward Switch Switch

Expected Frequency 114.7 45.5 lO 10.9

Observed Frequency 102.0 1330 23.0 44.0

comment types in each movement category are marked by asterisks. Note that

the functionality of the code was the topic of most of the programmers’ com-

ments. This was true for each type of movement except backward movements.

Apparently subjects do not discover as much about code functionality when

they are moving backwards. Inspection of the backward movement category

shows that questions and strategy are the primary concerns when programmers

regress through the code. Strategy comments are also prevalent when pro-

grammers switched from the forward to the backward direction.

The unequal distribution of comment types across movement categories

shows that programmers have qualitatively different things in mind as they

move around in code. We are working now on a model of programmers’ goals

and comprehension strategies that would account for the differences in reading

times and comment types that co-occur with changes in reading direction.

Comments on Programming Tools

Designers of programming tools who are interested in supporting the cog-

nitive processes and problem-solving strategies of programmers will have sev-

eral issues to take into account. First, the actual lines of code are the least

important aspects of a program from the point of view of long-term memory

Table 6.—Proportions of comments of each type within the movement categories. The two most frequent

comment types within each category are marked by asterisks.

Movement Type

Forward-Backward Backward-Forward

Comment Category Forward Backward Switch Switch

Analyze .078 053 0 .074

Assume DLO= .158 .167 220"

Question .078 263° 125 118

Answer .094 .105 .083 118

Function ADL .158 we .368*

Strategy .109 263" Dba .103

1.0 1.0 1.0 1.0

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 133

and problem solving. Information about the surface details of program state-

ments is probably lost soon after a statement is read. Only the meaning of a

statement, which corresponds in programming to its functional significance,

is retained. Even at the statement-level meaning representation is only useful

in recognizing larger functional units like loops or conditional clauses.

The fleeting significance of statement-level representations suggests that

improvements to programming interfaces which enhance readability or oth-

erwise aid in the recognition of lines of code will have limited influence.

Rather, enhancements that improve the recognition of meaningful code seg-

ments would be more useful. Designers of programming tools should focus

on presenting information about the functionality of each line to programmers

since this seems to be their final goal anyway when they read a line of code.

Second, pragmatic inference generation is an important part of program

comprehension. Inferences provide the links between program statements

which eventually form the macrostructure representation of the code. Inter-

estingly these types of connections are often not explicitly made in the code

(they may be present in comments, but commenting is an unreliable art form).

Programming tools that explicitly mark high-level connections between code

segments would be useful. One possibility is a hypertext environment in which

the hierarchical structure of code elements is represented and can be used to

maneuver through the program.

If good programmers learn a set of plans that they use over and over, and

if these plans have typical instantiations within a given language, then it should

be possible to construct a programming tool that generates the code for given

plans directly. For example, when a programmer wishes to create a loop that

will keep a running total, he or she might instruct the programming tool to

create the code for a running total in Pascal with the accumulation variable

named var]. Once created, the tool should keep track of the various pieces

of code that belong to the plan (e.g., an initialization of varl, the accumulation

statement, the loop control statements, the variable that holds the final output

value, etc.) so that changes in one part of the plan are propagated throughout

the plan—even if it is widely distributed spatially in the program.

A third issue relevant to programming tool design is related to the multiple

representations utilized by programmers. We saw that programmers not only

represent the procedural components of code, but also the functional relations

of code segments to each other based on the real-world semantics of the

problem that the program handles. Tools should help programmers encode

(in the design phase) and understand (in modification or comprehension tasks)

the semantics of the domain and the constraints it places on the code itself.

A useful, though difficult, tool would provide programmers with a “‘semantic

134 SCOTT P. ROBERTSON

compiler” that works the way syntactic compilers now operate. A semantic

compiler would recognize (or at least represent) functional relations and trap

violations of the semantics of the code. For example, such a tool might rec-

ognize errors in the ordering of function calls if the effects of one function

enable the operation of another in the real world application (e.g., in an

elevator control program a function that initiates the movement of one of the

elevators should be called after a function that assesses all waiting requests

for the elevator). A modification to the code that resulted in an inadvertent

violation of this semantic constraint should be trapped. Such a tool might go

further in suggesting how to structure a program to help avoid inadvertent

semantic errors (e.g., the tool might suggest to the programmer of the elevator

program that placing the call to the request assessment module within the

movement module would minimize the chance of later violating the constraint

on their ordering).

Finally the issue of programmers’ goals might influence the design of tools.

A debugging or modification tool should be different from a design tool, and

these in turn should be different from a maintenance tool. A programmers’

goal first affects the manner in which he or she wishes to search through the

code. In modification and debugging, programmers want to minimize the

amount of code that they look at, focussing only on those parts that are relevant

to the required changes. Their search strategies emphasize finding functionally

related components of the code and checking blocks of code for syntactic

consistency. Designers and maintainers might need to search code according

to the requirements of the task domain. Their search strategies might involve

finding all function calls that affect a certain action in the real world—and a

tool should help them do this.

In addition to search strategies, different goals affect the problem-solving

processes of, and hence the information required by, programmers. A pro-

grammer making a fix in a function wishes to reason locally, within the scope

of the function and perhaps those functions called by it. On the other hand,

a programmer or a designer making a major change in the functionality of a

whole program wishes to reason more globally. The information that these

two individuals require and the types of decisions they will make are quite

different. Designers of programming tools should have these considerations

in mind.

Conclusion

Research on the cognitive processes and knowledge structures underlying

program comprehension and modification have been discussed. Four major

issues were identified as important in understanding comprehension: micro-

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 135

structure representation, macrostructure representation, task and domain

knowledge, and programmers’ goals. While there are many similarities be-

tween theories in natural language comprehension and program comprehen-

sion, the procedural nature of programs and their role in controlling real-

world processes makes program comprehension an especially complicated

matter.

In theorizing about the processes involved in program comprehension it

may be most fruitful to think in terms of problem solving. This view increases

the importance of programmers’ goals and makes comprehension a less ho-

mogeneous phenomenon across tasks and programmers. It makes search strat-

egy and selective representation central issues as well. These changes in focus

in understanding program comprehension have implications for the design of

programming tools.

Programming tools should support high-level reasoning in several different

domains. The constraints on program structure due to the semantics of abstract

programming constructs and a program’s functionality in the task domain are

the primary concerns of programmers. These relations are not systematically

present in the code itself or even in the supporting documents. Tools that

make these constraints easier to understand and that support reasoning and

modification at these levels of representation will be the most useful. While

the technology is probably there to implement such programming aids, our

understanding of the structure of these representations and how they are used

in problem solving limits our ability to take advantage of it.

FOOTNOTES

In this paper I discuss work conducted in my laboratory by Erle Davis, Doug Fitz-Randolf, Jurgen

Koenemann, Kyoko Okabe, and Chiung Chen Yu. The work was supported by the Office of Naval Research

under contract number N00014-86-K-0876.

*Mayer does not report the proportion of variance for this test. Rather he reports a correlation of .75

between number of transactions and reading time. By squaring this coefficient we derive the proportion

of variance for which the predictor accounted.

>The equation is actually predicting log(reading time). Log transformations are commonly performed

on skewed data, especially time data which has a minimum of zero but no maximum. The transformation

creates a more normal distribution and does not affect ordinal relations. Interval relations are affected

only in the extremes.

‘This omnibus figure glosses many details. Below we discuss a breakdown of this analysis.

‘Dividing by the number of syllables controls for line length and makes the reading times directly

comparable.

‘Pragmatic inferences are distinguished from logical inferences in that they are not necessarily true.

Bibliography

Adelson, B. 1981. Problem solving and the development of abstract categories in programming languages.

Memory and Cognition, 9, 422-433.

Anderson, J. R. 1974. Verbatim and propositional representation of sentences in immediate and long term

memory. Journal of Verbal Learning and Verbal Behavior, 13, 149-162.

Anderson, R. C., and Biddle, W. B. 1975. On asking people questions about what they are reading. In G.

Bower (Ed.), The psychology of learning and motivation, Vol. 9, New York: Academic Press.

136 SCOTT P. ROBERTSON

Anderson, T. H. 1980. Study strategies and adjunct aids. In R. J. Spiro, B. C. Bruce, and W. F. Brewer

(Eds.), Theoretical issues in reading comprehension: Perspectives from cognitive psychology, linguistics,

artificial intelligence, and education, Hillsdale, NJ: Erlbaum.

Barnes, B. R., and Clawson, E. U. 1975. Do advance organizers facilitate learning? Recomendations for

further research based on an analysis of 32 studies. Review of Educational Research, 45, 637-659.

Bower, G. H., Black, J. B., and Turner, T. J. 1979. Scripts in memory for text. Cognitive Psychology,

11, 177-220.

Bransford, J. D., and Franks, J. J. 1976. Toward a framework for understanding learning. In G. Bower

(Ed.), The psychology of learning and motivation, Vol. 10, New York: Academic Press.

Bransford, J. D., and Johnson, M. K. 1973. Contextual prerequisites for understanding: Some investigations

of comprehension and recall. Journal of Verbal Learning and Verbal Behavior, 11, 717-726.

Brooke, J. B., and Duncan, K. D. 1980. Experimental studies of flowchart use at different stages of

program debugging. Ergonomics, 23, 1057-1091.

Clark, H. H. 1979. Responding to indirect speech acts. Cognitive Psychology, 11, 430-477.

Cuniff, N., and Taylor, R. P. 1987. Graphical versus textual representation: An empirical study of novices’

program comprehension. In G. M. Olson, S. Sheppard, and E. Soloway (Eds.), Empirical studies of

programmers: Second workshop, Norwood, NJ: Ablex.

Dyck, J., and Auernheimer, B. 1989. Comprehension of Pascal statements by novice and expert program-

mers. Poster presented at CHI’89: Human Factors in Computing Systems, Austin, Texas.

Ehrlich, K., and Soloway, E. 1984. An empirical investigation of tacit plan knowledge in programming.

In J. C. Thomas and M. L. Schneider (Eds.), Human factors in computer systems, Norwood, NJ: Ablex.

Fitter, M., and Green, T. R. G. 1979. When do diagrams make good computer languages? International

Journal of Man-Machine Studies, 11, 235-261.

Flores d’Arcais, G. B. 1974. Is there memory for sentences? Acta Psychologica, 38, 33-58.

Frederiksen, C. H. 1975. Effects of context-induced processing operations on semantic information acquired

from discourse. Cognitive Psychology, 7, 139-166.

Gibbs, R. W. 1984. Literal meaning and psychological theory. Cognitive Science, 9, 275-304.

Gibbs, R. W. 1986. What makes some indirect speech acts conventional? Journal of Memory and Language,

25, 181-196.

Gibbs, R. W., and Mueller, R. A. G. 1990. Conversation as coordinated, cooperative interaction. In

S. P. Robertson, W. Zachary, and J. B. Black (Eds.), Cognition, computing, and cooperation. Norwood,

NJ: Ablex.

Graesser, A. C., Robertson, S. P., Lovelace, E., and Swinehart, D. 1980. Answers to why-questions expose

the organization of story plot and predict recall of actions. Journal of Verbal Learning and Verbal

Behavior, 19, 110-119.

Gray, W. D., and Anderson, J. R. 1987. Change-episodes in coding: When and how do programmers

change their code? In G. M. Olson, S. Sheppard, and E. Soloway (Eds.), Empirical studies of program-

mers: Second workshop, Norwood, NJ: Ablex.

Grice, H. P. 1975. Logic and conversation. In P. Cole and J. Morgan (Eds.), Syntax and semantics 3:

Speech Acts. New York: Academic Press.

Guindon, R. 1990. Knowledge exploited by experts during software system design. International Journal

of Man-Machine Studies, 33, 279-304.

Jeffries, R. A. 1982. Comparison of debugging behavior of novice and expert programmers. Paper presented

at the Annual meeting of the American Educational Research Association.

Kammann, R. 1975. The comprehensibility of printed instructions and the flowchart alternative. Human

Factors, 17, 183-191.

Kintsch, W. 1974. The representation of meaning in memory. Hillsdale, NJ: Erlbaum.

Kintsch, W. 1976. Memory for prose. In C. N. Cofer (Ed.), The structure of human memory. San Francisco:

Freeman.

Kintsch, W. 1986. Learning from text. Cognition and Instruction, 3, 87-108.

Kintsch, W., and Keenan, J. 1973. Reading rate and retention as a function of the number of propositions

in the base structure of sentences. Cognitive Psychology, 5, 257-274.

Koenemann, J. 1990. Expert problem solving strategies for program comprehension. Masters thesis, Psy-

chology Department, Rutgers University.

Letovsky, S. 1986. Cognitive processes in program comprehension. In E. Soloway and S. Iyengar (Eds.),

Empirical studies of programmers, Norwood, NJ: Ablex.

Letovsky, S., Pinto, G., Lampert, R., and Soloway, E. 1987. In G. M. Olson, S. Sheppard, and E. Soloway

(Eds.), Empirical studies of programmers: Second workshop, Norwood, NJ: Ablex.

Littman, D. C., Pinto, J., Letovsky, S., and Soloway, E. 1986. Mental models and software maintenence.

In E. Soloway and S. Iyengar (Eds.), Empirical studies of programmers, Norwood, NJ: Ablex.

Mayer, R. E. 1976. Comprehension as affected by structure of problem representation. Memory and

Cognition, 4, 249-255.

KNOWLEDGE REPRESENTATIONS USED BY COMPUTER PROGRAMMERS 137

Mayer, R. E. 1979. A psychology of learning BASIC. Communications of the ACM, 22, 589-593. (a)

Mayer, R. E. 1979. Can advance organizers influence meaningful learning? Review of Educational Research,

49, 371-383. (b)

Mayer, R. E. 1981. The psychology of how novices learn computer programming. Computer Surveys, 13,

121-141.

Mayer, R. E. 1987. Cognitive aspects of learning and using a programming language. In J. M. Carroll

(Ed.), Interfacing thought: Cognitive aspects of human-computer interaction. Cambridge, MA: MIT Press.

Myers, G. J. 1978. A controlled experiment in program testing and code walkthroughs/inspections. Com-

munications of the ACM, 21, 760-768.

Nanja, M., and Cook C. R. 1987. An analysis of the on-line debugging process. In G. M. Olson, S.

Sheppard, and E. Soloway (Eds.), Empirical studies of programmers: Second workshop, Norwood, NJ:

Ablex.

Pennington, N. 1987. Comprehension strategies in programming. In G. M. Olson, S. Sheppard, and E.

Soloway (Eds.), Empirical studies of programmers: Second workshop, Norwood, NJ: Ablex. (a)

Pennington, N. 1987. Stimulus structures and mental representations in expert comprehension of computer

programs. Cognitive Psychology, 19, 295-341. (b)

Ramsey, H. R., Atwood, M. E., and VanDoren, J. R. 1983. Flowcharts versus program design languages:

An experimental comparison. Communications of the ACM, 26, 445-449.

Rist, R. 1986. Plans in programming: Definition, demonstration, and development. In E. Soloway and S.

Iyengar (Eds.), Empirical studies of programmers, Norwood, NJ: Ablex.

Rist, R. 1989. Schema creation in programming. Cognitive Science, 13, 389-414.

Robertson, S. P., and Davis, E. 1990. A problem-solving analysis of code comprehension. Invited poster

at CHI’90: Human Factors in Computing Systems, Seattle, WA.

Robertson, S. P., and Yu, C. C. 1990. Common cognitive representations of program code across tasks

and languages. International Journal of Man-Machine Studies, 33, 343-360.

Sachs, J. S. 1967. Recognition memory for syntactic and semantic aspects of connected discourse. Perception

& Psychophysics, 2, 437-442.

Schank, R. C. 1972. Conceptual dependency: A theory of natural language understanding. Cognitive

Psychology, 3, 552-631.

Schank, R. C., and Abelson, R. P. 1977. Scripts, plans, goals, and understanding: An inquiry into human

knowledge structures. Hillsdale, NJ: Erlbaum.

Seifert, C. M., Robertson, S. P., and Black, J. B. 1985. Types of inferences generated during reading.

Journal of Memory and Language, 24, 405-422.

Sheppard, S. B., Kruesi, E., and Bailey, J. W. 1982. An empirical study of software documentation formats.

Proceedings of CHI-82: Human Factors in Computing Systems (pp. 121-124). New York: Association

for Computing Machinery.

Shneiderman, B. 1982. Control flow and data structure documentation: Two experiments. Communications

of the ACM, 25, 55-63.

Shneiderman, B., Mayer, R., McKay, D., and Heller, P. 1977. Experimental investigations of the utility

of detailed flowcharts in programmming. Communications of the ACM, 20, 373-381.

VanDijk, T. A., and Kintsch, W. 1983. Strategies of discourse comprehension. New York: Academic Press.

Warren, W. H., Nicholas, D. W., and Trabasso, T. 1979. Event chains and inferences in understanding

narratives. In R. O. Freedle (Ed.), New directions in discourse processing. Norwood, NJ: Ablex.

Journal of the Washington Academy of Sciences,

Volume 80, Number 3, Pages 138-151

The Human Factors of

Voice Interfaces

John C. Thomas

NYNEX Artificial Intelligence Laboratory,

White Plains, New York 10604

ABSTRACT

Speech recognition and synthesis devices are examined for their present-day capability

to function as effective interfaces to computer-based systems, and their limitations are

identified. Some guidelines on the utility of those devices are offered but the practitioner

is cautioned that the adoption of those systems will be highly dependent on the specific

user-population, task, and context.

The technology for recognizing speech, synthesizing speech from text, ver-

ifying the identity of speakers, and digitally storing and manipulating human

speech has seen tremendous advances over the past few decades. In order to

help design efficient human-machine systems, the human factors specialist

must be aware of not only the uses of such technologies and the growing

literature on how to use these technologies most effectively, but also under-

stand something of the underlying technologies. I will begin by describing the

state-of-the-art in automatic speech recognition. I will also say something

about the underlying approaches and where the technology is likely to go. I

will then discuss the kinds of situations for which speech recognition as an input

medium to a machine is particularly appropriate. Then, some tentative guide-

lines will be presented based on the growing experimental literature in this

field. As always however, it will be necessary to test the details of any particular

system for the users, task, context, and specific system that is of interest to

the particular human factors investigator.

Another technology which has seen substantial progress is the area of text-

to-speech-conversion, otherwise known as speech synthesis. This latter term

however, is also sometimes used to refer to systems that digitize human voice

and play it back later (speech encoding and decoding). The quality of digitally

138

THE HUMAN FACTORS OF VOICE INTERFACES 139

encoding, storing, and later reconstructing human speech is only limited by

the money that one is willing to spend for storage. In contrast, unlimited text

to speech conversion, i.e., speech synthesis, is not yet as intelligible as human

speech and not nearly so aesthetically pleasing. I will describe in more detail

the state-of-the-art in speech synthesis and briefly outline some of the ap-

proaches. This will be followed by a discussion of the factors that would lead

one to consider using speech synthesis in an application. Of necessity, this

discussion will also touch upon speech encoding and decoding as an alternative.

I will present some guidelines concerning the use of speech synthesis.

Finally, I will briefly describe some additional voice technologies that can

be used today or in the near future. These include speaker verification, word

spotting, language identification, affect detection, and speech detection.

Automatic Speech Recognition

The state-of-the-art

The state-of-the-art in automatic speech recognition, as in many other new

technologies, is a rapidly moving target. As we will see, both the overly op-

timistic view that speech recognition is a panacea and already exists in such

a form as to provide an automatic unlimited speech conversion system, and

the pessimistic view that speech recognition is still far off in the future, are

equally fallacious. The truth of the matter is that for certain applications

(spelled out in more detail below), speech recognition already provides a

reasonable interface for human input. In order to understand the current state-

of-the-art, it will be necessary to review some of the dimensions on which

speech recognition systems vary.

One important dimension is the degree of independence of the system to

the various accents, speaking styles, and vocal tract shapes of individual speak-

ers. Systems which are geared to and must be trained on specific individual

speakers are referred to as “‘speaker-dependent” systems. At the other ex-

treme are systems which are called “speaker-independent”’ systems. Other

things being equal, speaker-independent systems are typically much more

limited in their capabilities than speaker-dependent systems. One reason for

this may be that, although it seems clear that human listeners quickly adapt

to individual speakers, automatic speech recognition devices have not typically

made use of this facility.

Another important characteristic to understand about speech recognition

devices is the acoustic environment in which they operate. To take one com-

mon and important example, the telephone network typically band-passes

140 JOHN C. THOMAS

speech between three hundred Hertz and thirty three hundred Hertz. In ad-

dition, some noise and other types of distortion may be introduced. This makes

speech recognition more difficult. Similarly, a speech recognition system that

must operate in the presence of large amounts of environmental noise, par-

ticularly if this noise includes human speech, will be at a disadvantage relative

to a speech recognition device operating in a noise free environment. Speech

recognition has also been investigated for use in airplane cockpits for fighter

pilots. In such an environment, not only is noise a problem, but also the G-

forces to which the pilot is subjected.

A third dimension for consideration is the degree to which the user’s be-

havior is constrained to be unnatural in a given situation. One important way

to constrain user behavior is to limit the available vocabulary. To take one

extreme example, if a speech recognition system need only discriminate be-

tween the words ‘“‘no”’ and “I here answer in the affirmative,” then it will

probably do quite well. As vocabulary size increases, the chances of two words

being similar enough acoustically to be confused by the recognizer increases.

Not only do most speech recognition systems enforce limitations of vocabulary,

but often on the manner these vocabulary items are put together. For example,

the system may require discrete utterances; that is, the user must pause be-

tween each word. In a continuous speech recognition system, the user speaks

in a more normal, continuous fashion. For the recognizer to deal with this

continuous speech, it must not only recognize words but segment the contin-

uous stream of speech into the individual words as well. In addition, the

continuous speech recognition problem is made more difficult because in

continuous speech the co-articulation effects at the word boundaries can be

quite strong.

A fourth set of considerations has to do with the time to process, the cost

of the device, and other operating constraints. If a system costs a million

dollars, it will probably not find its way into every telephone set, no matter

how good the performance. Whether or not the system needs to recognize

speech in real time will depend upon the application.

The point for the human factors professional in understanding these di-

mensions is that speech recognition accuracy, in and of itself, has no meaning.

Recognizers can only be compared when it is clearly understood what the

conditions of testing were and what the conditions of actual use will be. For

example, if one uses one’s colleagues and has them read telephone numbers

over the phone, the accuracy performance results will probably be quite dif-

ferent than if one requests naive users to speak telephone numbers that they

have just dialed.

As noted above, specific accuracy figures can be very misleading and speech

THE HUMAN FACTORS OF VOICE INTERFACES 141

recognition devices must be tested under conditions of actual use. For this

reason, I intend to give the reader some feel of what is generally possible.

The reader, in turn, must keep in mind that actual results will depend upon

the particular vocabulary items chosen, the range of speakers who will use

the system, the acoustic condition of the system that will be used, and a variety

of other factors. Broadly speaking however, we can say that speech recognition

devices exist today which will operate in a quiet office environment and un-

derstand a large vocabulary (on the order of a few thousand words) provided

the system is trained on an individual speaker and provided the speaker inserts

pauses between each word. Whether or not such a device constitutes an ac-

ceptable substitute for human transcription must be looked at in the context

of a specific application.

Smaller vocabulary speaker-trained devices have been profitably employed

in industry for sorting and inspecting. In such cases, there is typically some

background noise. The systems typically allow more efficient input of data

than do other alternatives. With regards to over-the-telephone speech, both

speaker-independent and speaker-dependent systems suffer somewhat.

Speaker-dependent systems exist which will allow up to twenty to fifty words

discretely uttered, to be recognized at any one point. The total vocabulary

can be larger provided that there is a clear a priori way of knowing at which

point in the dialogue various words can be uttered. Speaker independent

systems are limited at this point to the digits, yes, no, and perhaps a few

control words. The state-of-the-art demonstrates that systems can recognize

a fair proportion of speakers saying continuous digits over the phone.

Applications of Speech Recognition

Speech recognition technology can serve as an alternative input device.

While the advisability of using speech must be examined on an individual

basis, and if it seems like a reasonable input modality, tested under conditions

of use, there are some general guidelines for which applications are likely to

prove amenable to automatic speech recognition.

There are people for whom keyboard entry is not an option; for example,

children learn to talk several years before they learn to read and type. In

many parts of the world, there are also large populations of adults who can

speak in a particular native language quite fluently but who are unable to read

or type. Aside from this, speech recognition is particularly suitable when the

person’s eyes/hands are otherwise occupied and therefore keyboard entry

becomes cumbersome and time consuming. It is largely for this reason that

speech recognition is used in several industrial applications involving inspec-

142 JOHN C. THOMAS

tion and sorting tasks. Because speaker dependent systems typically have

higher accuracy rates and/or larger vocabularies than speaker independent

systems, speech recognition is typically feasible over a wider range of appli-

cations where there are dedicated users involved who will spend the time to

train the speech recognition device on their particular speech pattern.

About 40% to 50% of the phones in the United States are still of the rotary

type. The use of automatic speech recognition (ASR) over the telephone

allows many telephone-related services to be used by all subscribers and not

just those possessing touch-tone service. One may also argue that even for

touch-tone subscribers, speech is a more natural form of input. Factors mit-

igating against the use of automatic speech include a heavy penalty for errors.

Almost no current speech recognition system is 100% or even 99% accurate.

Therefore unless there are additional ways of insuring safety, speech recog-

nition is probably not the input medium of choice where a single error can

have extremely costly results. It is problematic to use with one time customers

since it may be difficult to arrange the situation so they will naturally contain

their speech in the ways that are currently necessary for speaker independent

systems.

Speech recognition systems can thought of, not only as alternatives to man-

ual input, but as supplements. For example, a combination speech recognition

and handwriting recognition system would be much more accurate than either

system by itself. One can also imagine a very natural text-editing system in

which the operands or areas to be operated on were specified manually (for

example by mouse or trackball) and the operators specified by discrete voice

commands.

For speaker-independent systems, there are typically some proportion of

speakers for whom the system works quite well within the constraints of

vocabulary rate etc., and some speakers for whom the system does not work

very well at all. Because of this it will be necessary in the case of speaker

independent applications to have a non-speech backup.

The most important thing to say about guidelines is that they should be

used as tools, not rules. We do not know enough to make rules. In addition,

the technology changes. Finally, systems must ultimately be tested for real

use by real users in a real context doing real tasks. While readers should find

the references at the end of this paper useful, I have refrained from the

practice of relating guidelines to references. I believe correlating specific guide-

lines to specific references grossly overstates the empirical basis for today’s

human factors guidelines. The guidelines are better thought of as personal

best-guesses based on intuition, experience, and a reading of experiments that

are sure to be done under different conditions than the application of interest.

Table 1 shows architectural guidelines. Table 2 lists guidelines for recognition.

THE HUMAN FACTORS OF VOICE INTERFACES 143

Table 1.—Architectural Guidelines

@ Consider ‘‘Wizard-of-Oz” prototyping for capabilities that are difficult to implement and the team

disagrees on value.

@ Use high level prototyping languages initially.

@ Never let a bad feature slip by now to be fixed later. Insist on quality throughout.

@ Circulate to all development team members a description of the product from the end-user’s

perspective.

@ Define the end-user(s), task(s), context(s).

@ Provide a “Home Base”’.

@ Provide “Undo” capability.

@ Use table-driven interface.

@ Put all messages in one place in the code.

@ Use variable length fields for messages, prompts, etc.

Guidelines For Using Speech Recognition

Prior to describing specific guidelines for the application of speech recog-

nition devices, it may be useful to point out several general principles. While

one of the advantages of speech recognition is touted to be its naturalness,

the fact of the matter is, current systems will require some deviations on the

part of the user from their natural speech habits. Some care must be given

therefore to selecting a limited vocabulary which contains items which are

acoustically dissimilar enough to be recognized and yet which also contains

enough items to appropriately cover the domain of interest. It should also be

noted that users may well attribute too much intelligence and human-like

capability to a system that uses speech recognition as an input. For example,

they may expect the system to understand that synonyms refer to the same

item. Commonly, when the speech recognition system does not understand

the user’s first try on a trained word, the user’s natural reaction is to speak

the word more loudly and distinctly than before. This of course, decreases

the chances for speech recognition systems to recognize the word spoken.

Text to Speech Synthesis

The state-of-the-art

Speech synthesis systems convert text in ASCII or EBCDIC into spoken

speech. The process of converting written or printed text into internal com-

Table 2.—Recognition Guidelines

@ Use ASR where errors are not catastrophic.

@ Provide an alternative (back-up) to ASR, especially in speaker independent systems.

@ Use ASR when eyes/hands are busy.

@ Have users participate in vocabulary design.

@ Make sure users are immediately signalled that they are using ASR.

@ Make the “‘training”’ situation like the “use” situation.

e@ Test the ASR in its real environment with the real users.

144 JOHN C. THOMAS

puter codes is a separate process that will not be discussed. In converting text

(inside the computer) into speech, several separate sets of problems must be

addressed. First, special handling may be required for the appropriate reading

of abbreviations, numbers, and special symbols and signs. This task is not as

trivial as it may sound. There are many fairly arbitrary conventions that must

be followed. For example, the capital letters without periods “IBM” are

usually pronounced as separate letters: ““I-B-M” whereas the capital letters

“RBOC” are generally pronounced ‘“Are-Bok’’. “Dr.” can be pronounced

“Doctor” or “Drive” depending upon the context. When the numbers 1492

appear, they are pronounced “‘Fourteen-ninety-two”’ in the context of a date.

In the context of net profit however, it may be pronounced ‘‘One thousand

four hundred and ninety-two.” In pronouncing a transcription of someone’s

notes, an ampersand should be read as ‘“‘and’’, while in context of a computer

language the ampersand should probably be pronounced ‘‘ampersand.”’ Cur-

rent speech synthesis systems do fairly well in such matters. It should be clear

however, that doing perfectly (i.e., as a human being would do it) is equivalent

to having a general purpose natural language understanding system. Instead,

current synthesis systems rely on fairly simple-minded statistical context rules

for making such decisions.

A second challenge facing speech synthesis systems is the translation from

a string of orthographic. characters to units of pronunciation. The so-called

letter-to-sound rules are particularly irregular for English. This is illustrated

by George Bernard Shaw’s famous example of how one might pronounce

‘“GHOTYI’;/ namely, “‘fish.”’ In this example, ““GH” is pronounced as it is in

the word “rough.” The “O” is pronounced as in the word “‘women”’. Finally,

“TT” is pronounced as in the word “‘nation.’’ Generally, common words in

English are likely to be more irregular. For this reason, most speech synthesis

systems have an exception table. If a word is found in the exception table, an

associated pronunciation is used. If not, letter-to-sound rules are used for

pronunciation. The problem is particularly difficult in dealing with proper

names. Commercially available systems may only pronounce 50% of proper

names correctly. Research systems do substantially better.

Aside from the difficulties and vagaries of English pronunciation, there is

the further problem that many words (over two hundred) in English are

pronounced differently depending upon the context. These non-homophonic

homomorphs are pronounced differently depending on the part of speech or

which of two words is meant. For example, “does” as an auxiliary verb is

pronounced ‘“‘duz”’ while as the plural noun for female deer is pronounced

“doze.” In other cases, even two nouns can be pronounced differently. For

instance “‘bow”’ is pronounced to rhyme with ‘‘bough”’ when it is the front of

THE HUMAN FACTORS OF VOICE INTERFACES 145

a boat but to rhyme with “‘beaux’’ when it is a ribbon configuration. Again,

current synthesizers rely on fairly simple-minded statistical context and sen-

tence-position rules to make a guess about which pronunciation in such cases

is likely to be correct. However, in the absence of an adequate parse, such

systems will always be prone to error.

Given that a sequence of phonemes for individual words is correctly decided

upon, the synthesizer must then do the work of actually pronouncing this

string of phonemes. There are several approaches to this task. Two main

variables are the choice of the unit of pronunciation and the manner in which

these units are conjoined. The units of speech are either segments of actual

human speech or some abstractions from that; commonly, for example, for-

mant positions, some indications of phonetic features such as nasality, and

whether the source of the sound is periodic or gaussian noise. Some synthe-

sizers have a relatively small vocabulary of fundamental units; for example,

the phonemes. Although notionally the forty or so phonemes of English are

the fundamental units of pronunciation, the same strong co-articulatory effects

that make speech recognition difficult make intelligible, natural sounding speech

difficult. If one uses a small number of fundamental units such as phonemes,

complex rules must be applied so that they may be appropriately modified

depending upon context. Another common choice is to store all the major

phonetic variants as fundamentals units. In this scheme, there are usually

contextually defined variants of particular phonemes. A still larger set of

fundamental building blocks is required in the diphone approach which stores

transition between adjacent phonemes thus requiring only slightly less than

forty times forty fundamentals units (Thomas, et al., 1984). A somewhat

similar approach is to use demi-syllables.

In addition to the requirements for a synthesizer to produce the right se-

quence of sound units, the English language is also subject to a large number

of variations in fundamental frequency, amplitude, speaking rate, spectral tilt,

and degree of articulateness depending upon meaning and intent. These vari-

ations are known collectively as prosody.

The state-of-the-art in speech synthesis is that on a word by word basis,

common words of English are generally pronounced fairly intelligibly. Very

similar sounds such as “T’’, ““K’’, ““P”, are more frequently confused than

they are in real speech. However, given some listening experience with a high

quality synthesizer and the context provided by narrative text, synthetic speech

is intelligible enough so that most listeners can answer comprehension ques-

tions after listening to synthetic speech as well as after listening to real speech

(Pisoni, 1982). Speech synthesizers today, however, do not sound natural and

they make many errors when it comes to prosody, the pronunciation of proper

146 JOHN C. THOMAS

names, and non-homophonic homomorphs. Thus, whether or not today’s

speech synthesis systems are “good enough”’ for an application depends upon

the application. Even subtle task variations within what is labelled as an

application may provide quite different acceptability ratings. For example,

one application of interest to phone companies is automated customer name

and address (ACNA). In this application, a user specifies a phone number

and receives name and address information, i.e., a reverse directory. In some

cases, users wish to verify credit information and may have, for instance, a

check with the customer’s name and address written on it. In such cases,

today’s synthesizers, provided one obtains one of the highest quality, are

sufficient. In other cases, however, sales people may be trying to map out a

territory and need to find the customer’s name and address ‘‘cold.” For this

purpose, today’s commercially available synthesizers are probably not suffi-

ciently accurate.

Guidelines for Using Speech Synthesis

There are several types of applications where speech synthesis provides a

reasonable medium for computer output to a person. Speech output provides

a way of understanding textual material when that is not possible by sight. It

is an alternative for blind people, children too young to read, and the over

one billion illiterate people in the world. The only real alternative in such

cases is live or recorded human voice. For many applications this is prohibi-

tively expensive. Speech synthesis is particularly appropriate when the ma-

terials must be flexibly presented or the textual basis changes frequently.

Speech synthesis is also particularly useful when eyes and hands are busy.

Thus, for instance, it is both more efficient and safer for a machine tool

operator to listen to instructions while focusing their visual system on the task

at hand. A related useful aspect of speech as an output medium is its omni-

directionality. In a sense, it serves as its own orienting signal. So for instance,

on a large factory floor, a combination of alarms with a central visual moni-

toring station can be replaced with verbal warnings. This means that someone

may now hear the warning and go directly to the area where their involvement

is needed rather than having first to go to the visual monitoring station.

Similarly, if someone is making an adjustment in an awkward position inside

a partially completed piece of machinery, that person can get feedback about

the adjustment without moving out of position.

Speech, of course, also offers an alternative means of presenting information

and as such can be a useful adjunct in educational settings or situations where

the human operator is already overloaded. According to Wickens (1984), our

ability to deal with a visually presented spatial task, and listening and com-

THE HUMAN FACTORS OF VOICE INTERFACES 147

prehending verbal materials, is much more independent than trying to deal

with two visually presented tasks, for instance. One can easily imagine the

advantages of presenting problems requiring three dimensional visualization

on a visual display unit while simultaneously giving auditory commentary.

Research is underway to determine the most effective way of using auditory

output in the information overload situation in which fighter pilots often

find themselves.

Another aspect of speech synthesis which opens up a wide number of ap-

plications is the simple ubiquity of the telephone. The telephone offers a cheap,

sturdy, potentially lightweight, and already deployed alternative to visual

display units for the presentation of information. Unlike E-mail transmissions,

the telephone offers point-to-point instant two-way connectivity. Thus, via a

combination of using either touch-tone input or limited speech recognition

and speech synthesis, one can gain access to data that is stored anywhere in

the world from nearly anywhere in the world. Again, it should be noted that

recorded human speech is an alternative in many of these situations but is

generally prohibitively expensive. This is particularly true for information

which changes rapidly; for instance, ‘“‘news’’, or current weather and driving

conditions. Tables 3 and 4 summarize some considerations for recorded speech

vs. synthesis.

Recorded Speech vs. Synthesis

Even if one deploys speech synthesis in an appropriate situation, care must

still be taken to insure that the technology is deployed appropriately within

the application. It should be first noted that since intelligibility is not as high

with speech synthesis as it is with human speech, it is useful to ensure that

there is some redundancy in the messages. Telegraphic short cuts (which were

Table 3.— Guidelines for Use

@ Totally recorded messages are

—natural

—intelligible

—hard to edit

—very hard to update

@ Recombinant recorded messages are

—almost natural

—intelligible

—limited in scope

Synthesis (text-to-speech) is

—flexible (prototype messages)

—easy to edit

—cheap to update

—fairly intelligible

—not natural

—more attention demanding

148 JOHN C. THOMAS

Table 4.— Guidelines for Use

@ Provide information in topic-comment order

@ Provide high level maps of system

@ Use medium length words

@ Avoid words with multiple parts of speech (“‘run time buffer’’?)

e@ Differentiate warnings, prompts, and feedbacks

@ Adopt frequently used words

@ Avoid negations .

e Avoid long strings of nouns (also known as the “Long noun string confusion avoidance principle’’)

@ Verb-object statements are easier than conditionals (‘Enter first name”’ is better than “If first name,

then enter’’)

@ Use consistent command patterns (Forward/background”,, not ‘“‘Next/back’’)

e@ Avoid ambiguous words (‘“‘Last”? = “Final” or ‘“‘Previous’’)

e@ Avoid ‘“‘Telegraphese’’ (Memory cheap keep short unneeded)

originally designed in the 1950’s to save the internal storage space of digital

computers) have continued to be in vogue in instruction manuals and computer

error messages in the much the same way that our appendix survives. Today,

in our anatomy, it only provides grief and serves no useful function. In giving

instructions to be followed, it is generally a good idea to repeat sections in

this manner. “‘Proceed south on interstate 684 and take exit 8. Exit 8 will put

you on Smith Road. You will take Smith Road, north-bound to the third light.

At the third light you will turn right off of Smith Road onto. . . .” In addition,

words should be chosen for messages which are fairly common words and yet

not overly short. Short words tend to be more easily acoustically confused

and are often ambiguous as well. Sentence structures should generally be

active and fairly simple.

One of the things that is difficult about auditory interfaces is keeping track

of the menu structures and where one “is” within that structure. To some

degree these difficulties can be mitigated by using “‘maps”’ where some brief

attention to visual stimuli is possible, by using distinctive voices at various

levels or within groups in a mnemonic way, and by limiting menu choices to

three or four at a time.

Many commercially available synthesis systems have a number of suitable

parameters and defaults for those parameters. It is important not to assume

that the default values are optimal for intelligibility. In many cases, other

values of rate, fundamental frequency, and loudness may improve intelligi-

bility. It may be worthwhile, if you are contemplating using a particular speech

synthesis device, to do preliminary testing to determine a confusion matrix

for the various phonemes. You may find particular confusions that you may

want to avoid in your messages.

Listeners differ quite a bit in their ability to understand synthetic speech

and this ability changes over time. This provides a challenge for the human

factors professional since both within-subject and between-subjects designs

THE HUMAN FACTORS OF VOICE INTERFACES 149

are problematic. It also implies that it is worthwhile to test out speech synthesis

on populations that are truly representative of your users and to realize that

there will be some learning involved. The positive side is that laboratory results

may understate the performance that repeat listeners will eventually achieve

in the field.

In speech synthesis applications, because listeners will improve over time,

and there are large individual differences, it is probably highly desirable to

allow user control of rate and to allow interruptions and replay. Adding a

warning tone or beginning to a speech synthesis message, according to Simpson

and Williams (1980), does not seem to help intelligibility. Since speech syn-

thesis seems to require greater attentional demands than natural speech, this

must be also taken into account.

Additional Speech Technologies

While speech synthesis and speech recognition are the two main voice tech-

nologies, there are also a number of others which have applications in human

computer interactions. One such technology is language detection. In other

words, the computer automatically determines which language is being spo-

ken. This can be useful for directing users to the appropriate person or sub-

system. It can also be used to scan large amounts of auditory data which would

be of interest to a particular person.

Similar applications can be used for speaker identification. At this time,

speaker identification only works within a fairly limited set of people. Speaker

verification can be used in conjunction with speech recognition, but also in

other applications as an added measure of security. Even the simple function

of speech detection can be useful in certain contexts. For example, we could

imagine a phone call that would be forwarded to a secretary if a conversation

were taking place within an office. Similarly, we could alert potential visitors

to the presence or absence of conversation in an office before they entered

for a visit. We could also search phone lines sending data or voice, for the

voice portions to be segregated and processed differently.

While analog techniques for storing and recording and playing human voice

have existed for many decades, there are definitely advantages to storing,

recording, and playing back human voice digitally. In some cases, it allows

cheaper storage, i.e., less physical space. In addition, various error-correction

and noise-elimination algorithms can be brought to bear more easily on digital

speech. Digital speech may also be varied as to rate and fundamental frequency

according to some schemes of digitization. Digitization also allows a number

of security measures to be imposed through encoding-decoding schemes. Dig-

150 JOHN C. THOMAS

ital transmission saves bandwidth in much the same way that digital-speech

storage saves space. Most of the comments in the section above on the pre-

sentation of speech synthesis apply equally well to the presentation of digitized

human speech (but see Table 2).

Summary and Conclusions

Speech technology offers partial solutions for a number of application prob-

lems. The human factors professional should gain an understanding of the

underlying technologies beyond what can be presented here. They should also

familiarize themselves with the empirical work (see references). The main

value, however, of such reading is to get a sense of what is possible and how

to do your own evaluations. You cannot rely on any specific values for rec-

ognition accuracy or synthesis intelligibility. Your own users, tasks, and con-

text will exert too big an influence on what will happen in your application.

References

Ballas, J. A., Dick, K. N., and Groshek, M. R. Failure to Identify “Identifiable” Sounds, Proceedings of

the Human Factors Society 31st Annual Meeting, 144-146, 1987.

Basson, S., Christie, O., Levas, S., and Spitz, J. Evaluating Speech Recognition Potential In Automating

Directory Assistance Call Completion, NYNEX Science and Technology, White Plains, NY.

Basson, S., Lauritzen, N., Lea, W., Levas, S., Loring, A., Ormazabal, G., Ortel, W., Spitz, J., and

Yashchin, D. Automation Potential of Operator Number Identification (ONI) Intercept Services, Tech-

nical Memorandum, NYNEX, 88-0006, White Plains, NY.

Basson, S., Lauritzen, N., Levas, S., Loring, A., Rubin-Spitz, J., and Yashchin, D. Performance of Speech

Recognition Devices: Evaluating Speech Produced over the Telephone Network, 1989 ICASSP Pro-

ceedings.

Boies, S. J., and Gould, J. D. Human Factors of the 1984 Olympic Message System. Proceedings of the

Human Factors Society—28th Annual Meeting, 547-551, 1984.

Boies, S. J., Gould, J. D., Levy, S., Richards, J. T., and Schoonard, J. The 1984 Olympic Message

System—A case study in system design. Res. Rep. RC-11138, IBM, Yorktown Heights, NY, 1985.

Edward, Alistair D. N. Soundtrack: An Auditory Interface for Blind Users. Human Computer Interaction,

1989, 4 (1), 45-66.

Gardner-Bonneau, D. J., Bonneau, J. R., and Ahlstrom, B. M. Voice Recognition Technology Potential

for a Complex Map Query System. 956-960.

Gaver, W. W. The SonicFinder: An Interface That Uses Auditory Icons. Human Computer Interaction,

1989, 4 (1), 67-94.

Halstead-Nussloch, R. The Design of Phone-Based Interfaces for Consumers. Proceedings of the ACM

SIGCHI Meeting, 347-352, 1989.

Holmgren, J. E. Toward Bell System Applications of Automatic Speech Recognition. The Bell System

Technical Journal, 1983, 62, (6), 1865-1880.

Kondziela, J. M.. (September 1988) Rapid Prototyping and Evaluation of User Interfaces for Voice

Messaging and Other Telephony Services. In Proceedings of The Telephone Revolution: Its Impact on

the Business Market for Speech Technology (pp. 167-171). New York: Media Dimensions, Inc.

Kondziela, J. M. (1989). Radio-analogue Method of Audio Presentation and Control for Audiotex Services.

Washington, DC: U.S. Patent and Trademark Office. Patent No. 4899369. To be issued 2/6/90.

Kondziela, J. M. (1989). Voice Messaging at NYNEX Corporation: Multiple User Perspectives of the

Service. In Proceedings of the American Voice Input/Output Society ’89. San Jose, CA: AVIOS.

MacDonald, J., and McGurk, H. Hearing Lips and Seeing Voices, Nature, 1976, 264, 746-748.

Martin, G. L. The Utility of Speech Input in User-Computer Interfaces. International Journal of Man-

Machine Studies, 1989, 30, 355-375.

THE HUMAN FACTORS OF VOICE INTERFACES 151

Pisoni, D. B. Perceptual Evaluation of Synthetic Speech: Perception of Speech: The Human Listener as

a Cognitive Interface. Speech Technology, 1982, 1 (2), 10-23.

Rosson, M. B. Using Synthetic Speech for Remote Access to Information. IBM Thomas J. Watson Research

Center, Yorktown Heights, NY, 1985.

Rubin-Spitz, J., and Yashchin, D. Effects of Dialogue Design on Customer Responses in Automated

Operator Services. NYNEX Science and Technology, White Plains, NY.

Simpson, C. A., and Williams, D. H. Response Time Effects of Alerting Tone and Semantic Context for

Synthesized Voice Cockpit Warnings. Human Factors, 1980, 22 (3), 319-330.

Thomas, J. C. A Method of Studying Natural Language Dialogue. IBM Research Report, RC-5882,

Yorktown Heights, NY, 1976.

Thomas, J.C. The Computer as an Active Communications Medium. Proceedings of the 18th Annual

Meeting of the Association for Computational Linguistics, pp. 83-86, ACL, 1980.

Thomas, J. C. Studies In Office Systems: The Effect of Communications Medium on Person Perception.

Office Systems Research Journal, 1983, 1, (2), 75-88.

Thomas, J. C. Problem Solving By Human-Machine Interaction. In Human and Machine Problem Solving,

K. J. Gilhooly (Ed.), London: Plenum, 1989

Thomas, J. C., and Gould, J. D. A Psychological Study of Query By Example, AFIPS Conference

Proceedings, 1975, 44, 439-445.

Thomas, J.C., Klavans, J., Nartey, J., Pickover, C., Reich, D., and Rosson, M. B. WALRUS: A De-

velopment System for Speech Synthesis, IBM T. J. Watson Research Report RC-10626. Yorktown

Heights, NY, 1984.

Wickens, C. D. Engineering Psychology and Human Performance. Charles Merrill: Columbus, OH, 1984.

Williges, B. H., Schurick, J. M., Spine, T. M., and Hakkimen, M. T. Speech In Human-Computer

Interaction. In R. C. Williges and R. W. Ehrich (Eds.), Human-Computer Dialogue Design. Amsterdam:

Elsevier, 1985.

Witten, I. H. Principles of Computer Speech. Academic Press: London, 1982.

i he debt oF iat nek "I Q ol

re 7 j

be HR SA OG sitar Dg

| Om beta ay yf ane f wathons a6

fi Ree arith? A, eft HK aqracitey, ee tae. “ AG A aidoteat &

| ia sae atiatettess 3: ens eon herald Bie gowie YAY

ye 6 Wie es etre bes arutye, genre Py ys et omets aT IC, Oa kT <semalttte? bain

uli ee 8 Pie 40) ME ROT. nia, + ARE grin cra fel ale? x

Rae EN thetngeart MalD gts ge genet wee aeons te bas thd:

a Winey ns ay AY Loe Bae.’ a 0

Hitec WA aah is Sagi nvlkaal “eyes Le porate nee

PORER ACCS ath Gi ns Ar Pi | paotesayih, aiectenes mar

Ct ee Re aT 0 Ba iy aru Ha rome, a3. wae) 36) eawiete pets

pe il ar ‘ i *) sae i 2 . ae: we oxy at re oa eee eth oy thee

Serene aiyhes ie share sid DAP Sie ular hen iegieaeiodl jee hares

AA Hotkey ELMS oh, es hace mi Abn thy haga Tab ggrsbFHREAT “_ Spe soe

a8) Full nel A cone 4 ae

eth ao came ica ol 1ay

ote MA has Oh ek ee Oy ee ieee | Aus

%, 4 7 A) i 2 e. _ 4 a fe r i >

i hb) ar ‘1 uM Pe aan ey oS : oi Te i rue Pa. a)

yet A ei I — Dit Ie

ere sae) aie pee Lk tone dah dee tek i: aah Va hath, Ge 2 ois AS

hai +: ome? 2) eo sg were * se He | cota

isda A: tealest) gaya a The D mee Coad) fobs A tie: eg iW a

: a pr

i i iad

MO) teehee. eee ARNEIA.. eves? toleoress) oe sl

> 7 tr » rh e z

They i, tree, I ® j hs Bh Tae 4

K ;

A : ci on,

a Marg Aue ’ f wee. ye

aa, Ls b . ay e na. * sf. - i MD a

a es - ‘Ave ae S . ~~e

, p _ S

a Y ners : ns ee ” -

Vie St ertiovts, ' Te ™ , .

it i} ' 14 1, oy ie i i ‘ Mew 8

' r i

s, : ; ¥ ' <a ae &

tnd pall wry ; j a the Kae ie) i eat teat

Ly ; F oi § } Ne ' i

. ? * i & F = a 4 ¥ a :

= 2

bie > nes ie ; hy \ } y

‘ M pa? = wit

pli é * 4 “= 2 r pe

f arinds Se tie, be F PN Es Ae wu Pats ge

vi j " ; te f - "

rie wey % ‘ ce | \ » 4

* J

| 4 oa

ey F i a i 1 fF yes a) By i

* * " 1 pas

a TACKS B

} i 4

¢ =

ee 4 * at 7. ae y At

' ae ¥ ' , ‘ ke y

i ae c ithe " “A Ce '

tare 3 WAUDY Olek 1 hh > ere ‘yt hey 5a ‘ ee |

ss i

i E M4 J » r Bi f P Li “i i >

MECsaneere Oat Atte yhtacpehy | Tea yr ig, Oe Meneaundyy ay 4 iwae Ane

” ‘ : 4 \ i nl ’ ny 4

By i é was a, | j _ wr oe tei Sark Ltn)? F oy 4

RPE RY PS COE PLRE | eRe re ae ie wh Ry Mie! PRES. Ty et: RARE

- ecitieboke, 5: ATOR Radi ae dens : ake tecatd ans Pree ia not Copenh Oat

ier Ve . dD. At. ahd Tray ee A it ta r pot he Me te inp ean

Kardishole, 4M (iP Wn) AON SI ot SVG otal Btole Tiner Fee

‘Served (ae. a eres Ot pie ‘iy yr ly ning, tiny biteh an la! ‘Set a ‘KO. ‘, le Jime,

Piet, 0. eed Weedamyts, 1e tinertery Livi el Savebryy: Hilicws. Medeor, LUNG, Rody 76

par Per ie hee Chae LR ik OORT, Hagia (recs sr, aienastenies 8

foigectowe Steadies, ry, Wik | .

aie?

pir tS ’ Xt ; iy ¥

‘= bl f

Donations of back issues of Journal

From time to time, the Academy receives and furnishes gratis back issues of

the Journal of the Washington Academy of Sciences that are requested by

reference libraries in the US and abroad to fill gaps in their collection of this

journal.

The Academy’s inventory of back issues is adequate except for a few issues

that are depleted. It is requested that members who have no further need for

the following issues forward them to the Academy office for re-distribution

to libraries.

Volume 77, number 4

Volume 78, numbers 2, 3, 4

Volume 79, numbers 1, 2

Please address your donation to the following:

Washington Academy of Sciences

Attn: Dr. Jean Boek

1101 N. Highland Street

Arlington, VA 22201

A * *§ ( Piles

ty asniedt toed Rist me aaietinnihh bad ait abla vrmobeo. ort ain

pi haji 291199 TR. ted, edonstd Ts) ynobsaA roteqimes W A aah

aint. t0 Houdallod tigdt ni jeg LE at bsoida bas e F aut tri Bok

ouaal Wel e IOl Kyssxe srstepabit g a busca Aosirl Yo Ning |

ei (0 son todhet on sved ordw avodimact fei i} hoiaoupal es iT.

- pottediaib-o1 10) soo yeh eet of mol! biavwriel eave

“i | » yedomara .T\ amiuioV

bf SC eroiieim cv omuloy

m he vrodinun Wy SrnuloV

“Oh AROOL od? of nol} woe

geonsio? lo peaebo A. natgnidedy

HeOR nee wl atta.

Tyee 4 iia g be 0

LOSE ANY. dnotil ne

DELEGATES TO THE WASHINGTON ACADEMY OF SCIENCES,

REPRESENTING THE LOCAL AFFILIATED SOCIETIES

Em OcOPhICAl SOCIELY Ob, WaSHIMPUON 6 )...03. ccs cet ec edbeecascwenseseseetesanes James F. Goff

Eaunropological Society Of Washineton............0004. 00s eeececsmes neces Edward J. Lehman

Pemeeieadl SOCIELY Ok VWASHINIBION 206.5 y eee ae eee esd wee econ eb pape alee « Austin B. Williams

Senae AN SOCICEY OF WaSMUOlOU. o.oo s.c6 csc esd eo te wed noe ne lep roe aeboes Jo-Anne A. Jackson

Emeimolapical Society of Washinston 2... ........ 6.6.6 ce cee cw teen ecerees Manya B. Stoetzel

mma COPLAMIMIC SOGICLY ie hielo Pie sdse ee bu eRe e dae cee ie one blela wie we Stanley G. Leftwich

Peel SOCICtY OL WaSMINCtOM 2... ec. eed ce cede ecee ede bebelncees James V. O’Connor

cee society Of the District Of Columbia :. 2... 05 eee eee ced ene ee John P. Utz

PMB MAI AU EAISEOTICAl SOGCICUY) 5 Dac). ).bh ces «yee eb dee baleen o bso ned oad oe eo ue Paul H. Oehser

ee ea SOCICLY Ol WWVaSIMIMOION, 25.552. .0i ee ese eee le clewele tole a emis oulleeiee edoes Conrad B. Link

: Society of American Foresters, Washington Section ................... Forrest Fenstermaker

PETAR SOCICLY Of PNGINCEIS... 0... eee ne de ecw ese chee Bbecmwcuessueses Alvin Reiner

Institute of Electrical and Electronics Engineers, Washington Section...... George Abraham

American Society of Mechanical Engineers, Washington Section................ Michael Chi

Pemuntnolosical Society of Washington.................. 20. ee pee eeeecees Kendall G. Powers

American Society for Microbiology, Washington Branch................... To be determined

Society of American Military Engineers, Washington Post............. Charles A. Burroughs

American Society of Civil Engineers, National Capital Section.......... Herbert A. Pennock

Society for Experimental Biology and Medicine, DC Section ............ Cyrus R. Creveling

minetmcan Society for Metals, Washington Chapter ....................200- Pamela S. Patrick

American Association of Dental Research, Washington Section............ J. Terrell Hoffeld

American Institute of Aeronautics and Astronautics, National Capital

eae ae ete amare | ee ele. dS Pee ee ees Reginald C. Smith

imencam wMetcorolopical Society, DC Chapter.............6.0.00.c0ee dees A. James Wagner

oemeeiomeS SOCICty OF WashiMgtON, . 2.6.60... sce eee eee c nec e ese nacneanes Ralph Webb

Acoustical Society of America, Washington Chapter....................06. K. Richard Cook

icmedm Nuclear Society, Washington Section........... 0.00.00 .05.cce seen eee es Kamal Araj

iusitute ot Food Technologists, Washington Section...................-2-.00-- Elvira L. Paz

American Ceramic Society, Baltimore-Washington Section .............. Joseph H. Simmons

ee ORMDAIE MMIC AN SOCICIN, ee ceil cele Wie scence ie olay es ose aeaieel ena weed mane Alayne W. Adams

Pare Eastory oF Science CUD. 3)... 6. cee e's eee eee tee eee ene Albert G. Gluckman

American Association of Physics Teachers, Chesapeake Section ............ Peggy A. Dixon

Optical Society of America, National Capital Section..................... William C. Graver

American Society of Plant Physiologists, Washington Area Section ... Walter Shropshire, Jr.

Washington Operations Research/Management Science Council .............. John G. Honig

fistrument Society of America, Washington Section.................0ccceeeeenees Carl Zeller

American Institute of Mining, Metallurgical and Petroleum Engineers,

Se SCHON Ee ne) None th cigs aes dnbina ne ow denn deddemtaciadedes oe Ronald Munson

MMU AIG AL PASELOMOMEDS 22). tas af ose le so oe hee deen weal dale ee Robert H. McCracken

Mathematics Association of America, MD-DC-VA Section ................ Alfred B. Willcox

Sinict ot Columbia Institute of Chemists. ............060.2. 60064005 os Miloslav Rechcigl, Jr.

Metrice of Columbia Psychological Association... ...2.......-0..0pseceone sees een Jane Flinn

Pasmneton, Paint Nechmical Group... 0... ...<..0c0c.0ksseeeheanceces Robert F. Brady, Jr.

American Phytopathological Society, Potomac Division................... Deborah R. Fravel

Society for General Systems Research, Metropolitan Washington

(LEIS STEG SetnN 5 w/a 3 Cs Ree EP RY Ce eS A Ronald W. Manderscheid

miunnian Factors Society, Potomac Chapter...............6..2ce0ececeaee Thomas B. Malone

Pilicitcal Pishercs society, Potomac Chapter... 0.6.6 ..0s26e cee sen tans ens Robert J. Sousa

Association for Science, Technology and Innovation...................-...005- Ralph I. Cole

Pein SOCIO PICA SOCIELY 0405 bs J eighnaek Sec edls heey eed ew oe cence Ronald W. Manderscheid

Institute of Electrical and Electronics Engineers, Northern Virginia Section. ... Ralph I. Cole

Association for Computing Machinery, Washington Chapter............. Charles E. Youman

Re Aste: Statistical “SOGIGLY hk. 2a.) fede des cen cece eeeaieancs onebaneass cs Robert Jernigan

Society of Manufacturing Engineers, Washington, DC Chapter ............. James E. Spates

Nocieny on tidustial Pneineers, Chapter 14). 0... .....0 6600 cscs cc cae b een caes John Larry Baer

Delegates continue to represent their societies until new appointments are made.

Washington Academy of Sciences 2nd Class Postage Paid

1101 N. Highland St.

Arlington, Va. 22201

Return Requested with Form 3579

at Arlington, Va.

and additional mailing offices.

J ff VOLUME 80

Number 4

7 J Our nal of the December, 1990

WASHINGTON

ACADEMY ..SCIENCES

ISSN 0043-0439

Issued Quarterly

at Washington, D.C.

CONTENTS

Articles:

WALTER E. BOEK, “A Comparison of Early With Late Respondents” ... 153

C. R. CREVELING, “The Washington Academy of Sciences Awards Program

il SClIoMilie wae VeUIe MIMI TIOO il cok we veh a ele ay alee eel es 8

R. H. MCCRACKEN, “‘President’s Report to the Membership for the Year

De are RS OR Gi alas ee iae Stay aula d has ae ame ahs SO

“1990 Washington Academy of Sciences Membership Directory” ..........

Smee Mel SCCM IPRS NLS EU dN, is 2.2890 sic uous Gy ch ne lean whadu ue oom rue en amen

Washington Academy of Sciences

Founded in 1898

EXECUTIVE COMMITTEE

President

Armand B. Weiss

President-Elect

Walter E. Boek

Secretary

F. K. Mostofi

Treasurer

Norman Doctor

Past President |

Robert H. McCracken

Vice President, Membership Affairs

Marie Bourgeois

Vice President, Administrative Affairs

Grover C. Sherlin

Vice President, Junior Academy Affairs

Marylin F. Krupsaw

Vice President, Affiliate Affairs

Edith L. R. Corliss

Board of Managers

R. Clifton Bailey

Jean K. Boek

James W. Harr

Betty Jane Long

Thomas N. Pyke

T. Dale Stewart

REPRESENTATIVES FROM

AFFILIATED SOCIETIES

Delegates are listed on inside rear

cover of each Journal.

ACADEMY OFFICE

1101 N. Highland Street

Arlington, VA 22201

Phone: (703) 527-4800

EDITORIAL BOARD

Editor:

John J. O'Hare, CAE-Link Corpora-

tion :

Associate Editors:

Ibert G. Gluckman, University of

Maryland

Joseph H. Neale, Georgetown Uni-

versity

Marc Rothenberg, Smithsonian Insti-

tution

Marc M. Sebrechts, Catholic Univer-

sity of America

Edward J. Wegman, George Mason

University

The Journal

This journal, the official organ of the Wash-

ington Academy of Sciences, publishes orig-

inal scientific research,

historical articles, proceedings of scholarly

meetings of its affiliated societies, reports of

the Academy, and other items of interest to

Academy members. The Journal appears

four times a year (March, June, September,

and December). The December issue con-

tains a directory of the current membership

of the Academy.

Subscription Rates

Members, fellows, and life members in good

standing receive the Journal without charge.

Subscriptions are available on a calendar year

basis, payable in advance. Payment must be

made in U.S. currency at the following rates:

U.S, and Canada ......... 32 $25.00 4

Other countries... ........5 353g

Single copies, when available.........

Claims for Missing Issues

Claims will not be allowed if received more

than 60 days after the day of mailing plus time

normally required for postal delivery and

claim. No claims will be allowed because of

failure to notify the Academy of a change of

address.

Notification of Change of Address

Address changes should be sent promptly to

the Academy office. Such notifications

should show both old and new addresses and

zip-code numbers, where applicable.

Published quarterly in March, June, September, and December of each year by the

Washington Academy of Sciences, 1101 N. Highland Street, Arlington, VA 22201.

Second-class postage paid at Arlington, VA, and additional mailing offices.

critical reviews, —

Journal of the Washington Academy of Sciences,

Volume 80, Number 4, Pages 153-160, December 1990

A Comparison of Early with Late

Respondents to a Mailed

Questionnaire

Walter E. Boek

College of Democracy

Arlington, VA 22201

ABSTRACT

Differences between early and late respondents to an opinion questionnaire survey,

in which the total response was 98 percent of the universe, are analyzed. Those tending

to respond early were those who were fellows of their Association, attended more meet-

ings, knew more about their Association, felt that their Association’s support of legislation

was important, were better informed about their Association and were in accord with its

policies.

The problem of non-respondents or late respondents is a vexing one for

users of mail questionnaires. The literature contains many hypotheses about

who does not answer questionnaires, whether there is a difference between

respondents and non-respondents and if there is a difference, whether or not

it influences the findings.

Concern about non-respondents has changed during the last 40 years from

not being regarded as posing a threat to inferences from sample surveys,

according to Frankel and Frankel,' to where it was considered so serious that

the American Statistical Association had a National Science Foundation spon-

sored workshop on the non-response issue in 1973.

George Gallup also called attention to this research problem in his article

“Opinion Polling in a Democracy”’.* The most significant work, however, was

a three-volume report® published by a Panel on Incomplete Data established

by the Committee on National Statistics of the Commission on Behavioral

and Social Sciences and Education of the National Academy of Sciences. The

primary focus in this report was on statistical means of controlling errors with

some attention given to methods of increasing response to questionnaires and

interviews.

153

154 W. E. BOEK

Return rates for questionnaire studies vary from a low of about 5% for

those conducted by survey firms doing studies of physician readership of

medical journals to above 90% for some research by sociologists. From a

sample of 1,769 teachers in 62 schools, 983 responded.? In a study of 850 cable

television subscribers, 54% responded after three follow-up letters.” When

part of the universe was given monetary rewards of 25 cents to $2 for returning

the questionnaires, responses increased to 88% but those responding had lower

incomes than the 54% in the first group not given money. A survey of 215

public administrators resulted in an 80.5% return.® A survey of 1,000 members

of a professional executive woman’s organization produced 545 returned forms.’

The questions in these four surveys all seemed to be innocuous as were those

utilized in comparing early with late respondents in the present study.

A compilation and evaluative analysis of the literature on factors affecting

rates of returns of questionnaires was published in the American Sociological

Review.® Response bias of respondents and non-respondents has been dis-

cussed by a number of survey specialists,”-'° Question wording, sponsorship,

readability, layout, and size in relation to rate of return are the subjects of

other authors.!77!

Four factors are mentioned throughout the literature as affecting returns

to mail questionnaires. They are: failures, involvement with and loyalty to

the sponsor, interest in the topic, and education.

Failures are, in this context, those who have not done something considered

‘‘sood”’ by the sponsors of the questionnaire. For example, unemployed al-

umni may consider themselves failures when surveyed by their colleges. That

this type of individual responds later is shown in the studies by Shuttleworth”

and Barnette.”

In another study,” teachers who did not have up-to-date instructional equip-

ment in their classrooms tended not to respond to the complete questionnaire.

They may have felt that they had “failed” to keep up with progressive teaching

methods. Data relating to the failure factor, per se, were not obtained in the

survey reported in this study.

Loyalty to the questionnaire sponsor, may be separated in this study, to

only a small degree, from interest in the topic. That interest in the organization

Or institution is important in the early return of questionnaires is shown in

studies by Phillips, Franzen and Lazarsfeld,*° and Britten and Britten.*’

Method

To meet planning information needs of officers of the American Public

Health Association, a questionnaire was sent to all APHA members in one

state. Questions were designed to seek information such as whether or not

A COMPARISON OF EARLY WITH LATE RESPONDENTS 155

they knew the functions of the sections of their Association, qualifications for

becoming a fellow, legislation the Association had supported, and research

projects it was carrying out.

Other questions were included to determine if the respondents felt that

membership dues they paid to APHA were justified, if they preferred a dif-

ferent method of voting for officers, if one profession should have more to

say in the Association than others, and if they were satisfied with the section

of which they were members.

The questionnaires were all mailed on the same day with a cover letter from

the state health commissioner. A follow-up letter was sent to all non-respon-

dents to the first mailing and a second one to those not responding to the

second letter.

As the questionnaires were returned, each was stamped with the date re-

ceived. A return of 98% of the universe was achieved, 216 out of 220. The

‘Post Card” technique was utilized to obtain the high return in this confidential

survey.*> This compares to the lower return rates of surveys cited earlier in

which most questions solicited information on non-controversial issues.

Respondents were divided into these two groups, “‘early respondents’, who

answered previous to the first follow-up letter, and “‘late respondents”? who

returned their questionnaires after the first follow-up letter. There were so

few who waited until after the second follow-up letter that they were combined

with the late group. As shown in Table 1, there were 143 in the early group

and 73 in the late group.

The objective of the analyses in this study was to determine whether those

responding late to the questionnaire used are significantly different from those

sending them back early. The data do not permit generalizations relating to

questionnaire studies of other topics or of other populations.

In the analyses reported here four variables were selected: background data

on respondents, their knowledge about activities and functions of the Asso-

ciation, their participation in the Association, and their degree of satisfaction

with the Association’s programs for members.

Table 1.—Early and Late Respondents Compared by Profession

Profession

San. Statis- Health

Physician Nurse Eng’r tician Educator Others

When Returned N % N % N % N % N % N %

Early 62 70 29 58 21 62 5 45 10 100 16 70

Late 26 30 asi 42 13 38 6 55 0 0 7 30

Total 88 100 50 100 34 100 1 100 10 100 Zo 100

156 W. E. BOEK

Table 2.—Early and Late Respondents Compared by Affiliation with the Association

Affiliation

Fellow Member Other

When Returned N % N % N %

Early 68 78 70 61 a 36

Late 19 22 45 39 9 64

Total 87 100 115 100 14 100

In relation to background, tabulations were made on their status in the

APHA, that is whether they were a member or fellow, the section of the

Association of which they were affiliated, and the specific profession of each

respondent. Since office holding and chairmanships are confined to fellows,

only a limited idea of participation of respondents could be obtained.

Differences between professions shown in Table 1 are not statistically sig-

nificant at the conventional level (p = .05) by the Chi Square test even though

a larger proportion of health educators and physicians than the other profes-

sions responded early.

Another variable that was examined was whether or not a respondent was

a fellow. Fellows responded earlier as is indicated in Table 2. Although almost

any member working in the health field who wants to apply can become a

fellow, it is likely that those who decide to become fellows have greater interest

in the Association. To become a fellow, however, would be attractive to those

concerned about the activities of the Association, because only fellows are

allowed to hold office and serve on committees. This identification with the

objectives of the APHA might also account for the earlier responses of the

fellows and be congruent with the conclusions reached by Sirles,”” Clausen

and Ford,” and Phillips.*°

At first glance, data on attendance at annual meetings of the Association

(Table 3) would seem to support the role of interest. Early respondents at-

tended more meetings than the late respondents. By the Chi Square test, this

was significant at the p = .O5 level.

Table 3.—Early and Late Respondents Compared by Attendance at the Annual Meeting of their As-

sociation

When Returned

Early Late

Number of Meetings SEE ee

Attended N % N %

None 15 10 17. 23

One or More 128 90 56 a7

Total 143 100 73 100

A COMPARISON OF EARLY WITH LATE RESPONDENTS 157

Table 4.—Members and Others of the Association and their Attendance at Annual Meetings

When Returned

Early Late

Number of Meetings SUUEnEEIIEEEIEEEIEEEEEEEEEEREEIEEE ra

Attended N % N %

None 14 19 i7 31

One or More 61 81 37 69

Total fie) 100 54 100

When the factor of affiliation was taken into account though, the interpre-

tation changed. Of all the fellows, only one had not attended any meetings;

however, a tabulation (Table 4) of the members and others, in relation to

their attendance at meetings, and whether they responded early or late, showed

early respondents attended more meetings than the late ones.

Of interest was whether or not early respondents would be better informed

about activities and functions of their Association than the late respondents.

The APHA adopts positions on federal and state legislation, and supports

federal programs of interest to its members.

In response to the question “Do you know whether or not the APHA

actively supported or opposed any federal legislation this year?” a greater

proportion of the late respondents did not know that their Association sup-

ported federal legislation (Table 5). When tested by the Chi Square test, this

difference was found to be significant at the p = .02 level.

A similar relationship was found (Table 6) when they were asked: ‘‘What

is your feeling about supporting federal agencies in such a way as to protect

them from budgetary cuts and other abuses in any way possible?” A larger

proportion of early respondents reported that this support of federal agencies

was important to them while a higher proportion of the late ones said that

they did not know about this function of their Association. The Chi Square

test indicated that the probability of this happening by chance was low (p =

02).

Because the APHA carries on various studies, respondents were asked to

Table 5.—Respondent’s Knowledge of Their Association’s Support of Federal Legislation

When Questionnaire Was Returned

Early Late

Response N % N %

aes 40 28 9 12

No 11 8 6 8

Don’t Know 92 64 58 80

Total 143 100 fis 100

158 W. E. BOEK

Table 6.—Attitude Toward Support of Relevant Federal Agencies

When Returned

Early Late

Response N % N %

Important to Me 68 48 BH 37

Not Important to Me 14 10 0 0

Don’t Know 12 8 10 14

Didn’t Know About This Function 49 34 36 49

Total 143 100 73 100

list one or two studies they knew about (Table 7). Sixty-six percent of the late

respondents did not name any, while 53% of the early ones did list such

research. This statistically significant difference (p = <.05) indicated that

early respondents were better informed.

In an attempt to learn whether respondents knew the difference between

members and fellows, they were asked this question: “Do you know the

difference between a fellow and a member? If you do, what is it?” In Table

8 it is shown that early respondents were better able to do this than late ones

(p = .05 with the Chi Square test). The three categories of ‘“‘yes”’ answers

have been ranked according to the correctness of the response. Therefore,

“Yes I” includes those individuals giving the correct answer which is those

who have responsible positions in the health field who wish to apply and are

recommended by people who already are fellows. ““Yes II” included those

respondents leaving out either the recommendation requirement or that fel-

lows needed to have a responsible role in public health, while included in

“Yes III” are those stating that to become a fellow a member needed to be

a public health worker.

These questions dealt with qualifications for being a fellow, and, in general,

fellows replied earlier than others, so it was necessary to control the affiliation

variable in Table 8. When this was done, the relationship was not significant.

The difference was larger between the two groups. The fellows were, as was

to be expected, better informed on this topic.

Finally, it was found that early respondents were more satisfied with their

Society’s policy of only fellows being allowed to hold offices and serve on

Table 7.—Whether or Not Respondents Listed Studies They Knew Their Association Had Conducted

When Returned

Early Late

Response N % N %

Yes 76 53 25 34

No 67 47 48 66

Total 143 100 gf) 100

A COMPARISON OF EARLY WITH LATE RESPONDENTS 159

Table 8.—Knowledge of the Difference Between Fellows and Members

When Returned

Early Late

Response N % N %

Wes il 24 17 3 +

Yes II 43 30 20 27

Yes III Oi. 40 32 44

No ie, 13 18 25

Total 143 100 163 100

the Executive Council than those taking longer to return the questionnaires

(p = <.05) (Table 9). Because this question dealt with the privileges of the

fellows, and a greater proportion of fellows answered earlier than members,

it was advisable to control again on the affiliation variable. When this was

done, the difference between early response and high satisfaction with the

Association disappeared. The fellows were satisfied and the members were

dissatisfied. ,

Another question involved physicians working in public health who were

members and fellows of the APHA. The question was: “Should physicians

have more power than members of other professions?’’ Analysis showed that

physicians felt that they should have that power while the other respondents

did not (p = <.05).

Conclusions

Data from this questionnaire survey provided a unique opportunity to eval-

uate demographic bias in research with these instruments, especially because

its completion rate was 98%. The tabulations presented herein show that

misleading conclusions would result from data furnished by respondents to

questionnaire surveys to which a significant proportion did not respond.

In this study interest in the survey sponsor and the topic of the questionnaire

seemed to elicit an early response to it. The members of the Association who

responded early had attended more meetings and were better informed about

functions and activities of their Association.

Table 9.—Attitude Toward Only Fellows Holding Office and Being on The Executive Council

When Returned

Early Late

Response N % N %

Yes 91 64 25 34

No 49 34 35 48

Other 3 w 13 18

Total 143 100 DB 100

160

W. E. BOEK

References

. Frankel, M. R., & L. R. Frankel. 1987. Fifty Years of Survey Sampling in the United States. Public

Opinion Quarterly, 51(4), Part 2, 127-128.

. Gallup, G. 1972. Opinion Polling in a Democracy. In J. Tamur, Statistics: A Guide to the Unknown.

(pp. 146-157). San Francisco: Holden Day.

. Madow, W. G. (Ed.). 1983. Incomplete Data in Sample Surveys. Panel on Incomplete Data: Committee

on National Statistics. Vols. 1-3. New York: Academic Press.

. Licato, J. W., & C. B. Teddlie. 1990. Principal Vision, Teacher Sense of Autonomy and Environmental

Robustness. Journal of Educational Research, 84(2), 93-99.

. James, J. M., & R. Balstein. 1990. The Effect of Monetary Incentives and Follow-up Mailings on the

Response Rate and Response in Mail Quality. Public Opinion Quarterly, 54(3), 346-361.

. Cleary, R. E. 1990. What do Public Administration Masters Programs Look Like. Do they do What

is Needed? Public Administration Review, 50(6).

. LaRosa, J. H. 1990. Executive Women and Health: Perception and Practices. American Journal of

Public Health, 80(12), 1450-1454.

. Heberlein, T., & A. R. Baumgartner. 1978. Factors Affecting Response Rates to Mailed Question-

naires: A Quantitative Analysis of the Published Literature. American Sociological Review, 43, 447-

462.

. Hammett, W. E., & C. D. McDonald. 1982. Response Bias and the Need for Extensive Mail Ques-

tionnaire Follow-up Among Selected Recreation Samples. Journal Leisure Research, 14(3), 207-216.

. Gershen, J. A., & C. P. McCreary. 1983. Personality Comparisons of Respondents and Non-respon-

dents to a Mailed Personality Inventory. Psychological Reports, 52, 555-562.

. Pearl, D. K., & D. Fairley. 1985. Testing for the Potential for Non-response Bias in Sample Surveys.

Public Opinion Quarterly, 49, 553-560.

. DeMaris, A., & J. K. Jackson. 1986. Non-respondent Characteristics and Bias in a Study of Batterers.

Social Service Review, 60, 460-474.

. Suchman, E. A., & F. McCandless. 1941. Who Answers Questionnaires. Journal of Applied Psychology,

25, 588-591.

. Bloch, T., & I. D. Nebenzahl. 1983. Respondents’ Bias in Mail Surveys. Psychological Reports, 53,

1227-1230.

. Baur, E. J. 1947. Response Bias in a Mail Survey. Public Opinion Quarterly, 11(4), 594-600.

. Egarton, H. A., Britt, S. H., & Norman, R. D. 1947. Objective Differences Among Various Types

of Respondents to a Mailed Questionnaire. American Sociological Review, 12, 435-444.

. Reuss, C. F. 1943. Differences Between People Responding and Not Responding to a Mail Question-

naire. American Sociological Review, 8, 433-438.

. Larsen, J. D., Mascharka, C., & Toronski, C. 1987. Does the Wording of the Questions Change the

Number of Headaches People Report on a Health Questionnaire? Psychological Record, 37, 423-427.

. Harvey, L. 1988. The Effect of Auspices, Style, and Layout on Response Rates to Mailed Question-

naires. Sociology, 22, 129-135.

. Privette, G., & D. Sherry. 1986. Reliability and Readability of Questionnaires: Peak Performance

and Peak Experience. Psychological Reports, 58, 491-494.

. Jansen, J. H. 1985. Effect of Questionnaire Layout and Size and Issue Involvement on Response Rates

in Mail Surveys. Perception and Motivation Skills, 61, 139-142.

. Shuttleworth, F. K. 1941. Sampling Errors Involved in Incomplete Returns to Mail Questionnaires.

Journal of Applied Psychology, 25, 588-591.

. Barnette, W. L. 1950. Report of a Follow-up of Counseled Veterans. Journal of Social Psychology,

32, 129-136.

. Stanton, F. 1939. Notes on the Validity of Mail Questionnaire Returns. Journal of Applied Psychology,

23, 95-107.

. Phillips, W., Jr. 1956. Weaknesses of the Mail Questionnaire. Sociology and Social Research, 32, 129-

156.

. Franzen, M., & P. Lazarsfeld. 1945. Mail Questionnaires on a Research Problem. Journal of Psy-

chology, 20, 293-320.

. Britten, J. H., & J. O. Britten. 1951. Factors in the Returns of Questionnaires Mailed to Older

Persons. Journal of Applied Psychology, 35, 57-60.

. Boek, W. E., & J. Lade. 1963. A Test of the Usefullness of the Postcard Technique in a Mail

Questionnaire Study. Public Opinion Quarterly, 27, Part 2, 303-306.

. Sirles, E. A. 1984. Who Responds to Follow-up Studies? Social Casework, 65, 454-456.

. Clausen, J., & R. Ford. 1947. Controlling Bias in Mail Questionnaires. Journal of the American

Statistical Association, 42, 497-511.

Journal of the Washington Academy of Sciences,

Volume 80, Number 4, Pages 161-165, December 1990

The Washington Academy of

Sciences Awards Program for

Scientific Achievement in 1990

C. R. Creveling

National Institute of Diabetes, Digestive, and Kidney Diseases

Bethesda, MD 20982

One of the many ways by which The Washington Academy of Sciences

contributes to the growth and recognition of scientists in the Washington area

is through the awards program of the Academy. Each year the Academy

recognizes scientists who work in the Washington, DC area for scientific work

of merit and distinction. Awards are made for outstanding contributions in

Mathematics and Computer Sciences, Behavioral and Social Sciences, Engi-

neering Sciences, Biological Sciences and Physical Sciences. In addition the

Academy makes an award designated the “Distinguished Career in Science

Award” to recognize persons who have made distinguished and overall con-

tributions to science.

In keeping with the goals of the Academy which include the promotion of

excellence in the teaching of science the Academy also presents awards for

the Teaching of Science. These awards include the Leo Schubert Award for

excellence in college teaching and the Bernice Lambertson Award for excel-

lence in high school teaching.

The awardees are selected from those persons nominated by either Academy

members or the public, by panels of experts in each of the respective fields.

The decisions of the Award Committee are then considered by the Board of

the Academy for final approval.

In 1990, the Awards were presented at a ceremony, held at the Georgetown

University Conference Center, on Friday, May 25th. Following dinner, the

program was opened by the President of the Academy, Robert M. McCracken.

161

162 C. R. CREVELING

After the award was presented, each person selected made a brief presentation

on their work. The awardees were:

Dr. Jesse Bernard Behavioral and Social Sciences

Dr. Angela M. Gronenborn Biological Sciences

Dr. G. Marius Clore Biological Sciences

Dr. Michael R. Moldover Physical Sciences

Dr. Guillermo C. Gaunaurd Engineering Sciences

Mr. Alan O. Plait Leo Schubert Award

Dr. James Edward Falk Leo Schubert Award

Mr. William John Entley Bernice Lambertson Award

Mr. Grover C. Sherlin Special Recognition for Services to

the Washington Academy of Sci-

ences

Behavioral and Social Sciences

The award in the Behavioral and Social Sciences was granted to DR.

JESSIE BERNARD “For her ground-breaking work on the changing role of

women—for making the invisible visible.”’ Dr. Bernard has carved out for

special study the “‘world of women.” Beginning in 1925 she has worked in

this relatively uncharted terrain of scientific research. The titles of her major

books illustrate the range of her contributions: Academic Women, The Future

of Marriage, The Sex Game, The Future of Motherhood, Self-Portrait of a

Family. Her works have challenged most conventional thinking about women.

Drawing upon her own prodigious store of learning, she has traced the chang-

ing role and status of women in modern history. Her explorations have probed

the changing networks of women by social class, ethnicity, age, power, kin

relationships, and friendships. In addition to using the traditional methodol-

ogies of social scientific research, she has uniquely drawn upon qualitative

and largely fugitive materials: letters, diaries, photographs, arts, and crafts.

In doing so she has opened new paradigms for research. She has brought the

insights and understandings of basic science to the controversies surrounding

social issues and problems. Her scholarship has profoundly influenced the

work of modern sociology and of the surrounding disciplines. Dr. Jessie Ber-

nard is not only one of the outstanding scholars of our times, she has also

materially contributed to one of the major social changes in this period—the

changing place of women in society. Dr. Bernard was nominated by Matilda

White Riley.

AWARDS PROGRAM FOR SCIENTIFIC ACHIEVEMENT IN 1990 163

Biological Sciences

The award in the Biological Sciences was conferred on the husband and

wife team of Drs. ANGELA M. GRONENBORN and G. MARIUS CLORE

for their contributions to the development and application of nuclear magnetic

resonance spectroscopy and computer modeling for the determination of the

three-dimensional structure of proteins in solution. Drs. Gronenborn and

Clore have jointly developed an internationally renowned research program

for the determination of three-dimensional structure of proteins and other

macromolecules in solution. Their innovative research employs an integrated

combination of nuclear magnetic resonance spectroscopy (NMR) and sophis-

ticated computer modeling. This method harnesses the power of high magnetic

field NMR spectrometers and modern computers to examine a very large

number of three-dimensional conformations of the thousands of atoms that

comprise proteins and selects those with favorable low-energy structures. This

information coupled with the ranges of hundreds of interatomic distances

derived from two dimensional NMR permits the determination of proteins,

in aqueous solution, with a precision comparable with that of classical x-ray

crystallography. At present the team of Gronenborn and Clore has produced

40 percent of the known protein structures by this technique. Drs. Gronen-

born and Clore not only have made effective use of methods pioneered by

others but have made unique and innovative contributions to this field. These

awardees were nominated by Dr. Edwin D. Becker.

Physical Sciences

The award in the Physical Sciences was accorded to DR. MICHAEL R.

MOLDOVER for outstanding achievements in the measurement of the ther-

mophysical properties of fluids. In particular, for developing and using spher-

ical acoustic resonators to redetermine the universal gas constant R, for

accurately measuring the exponents and amplitudes characterizing phenomena

near the critical points of fluids, and for demonstrating the ubiquitous nature

of wetting layers and wetting transitions. Dr. Moldover has made outstanding

contributions in many areas of the physical sciences, including critical phe-

nomena, fluid interfacial phenomena, and acoustic metrology. He has per-

formed theoretical work, although he is known best for his outstanding and

creative experimental work. His contributions include models for the ther-

modynamic and interfacial properties of fluids and fluid mixtures, and more

recently his work on the characterization of alternatives to the chlorofluoro-

carbon compounds believed to be depleting the Earth’s ozone layer. Dr.

Moldover was nominated by Dr. Victor Nedzelnitsky.

164 C. R. CREVELING

Engineering Sciences

The award in the Engineering Sciences was granted to DR. GUILLERMO

C. GAUNAURD for his outstanding contributions in inverse scattering, par-

ticularly in acoustic resonance scattering. For over twenty years, Dr.Gaunaurd

has been engaged in individual research on the interaction of acoustic, elastic

and electromagnetic waves with material media. This effort has led to a basic

understanding of the scattering processes occurring when waveforms emerging

from projectors such as sonars or radars are incident on and reflected by

penetrable targets, and of the waveforms radiated by bodies when they are

excited into oscillation by various external agents. In brief, these have been

twenty years devoted to the study of the radiation and scattering of mechanical

and electromagnetic waves. Dr. Gaunaurd was nominated by Dr. Albert G.

Gluckman.

Leo Schubert Award For Teaching of Science in College

The Leo Schubert Award for the Teaching of Science in College was pre-

sented to two persons. An award was made to Mr. Alan O. Plait for his

excellence and innovative methods in the teaching of mathematics. Mr. Plait

has had a long and distinguished career in the teaching of science and, par-

ticularly, in mathematics. He has developed and instructed a wide range of

topics in such areas as electronics, mechanical engineering, reliability engi-

neering, and quality assurance and control. Since 1963 he has taught math-

ematics. He received the USDA Graduate School Faculty Excellence Award

in 1986. Mr. Plait served as Chairman of the Mathematics and Statistics Ad-

visory Board from 1987 to 1989. Under his leadership the school significantly

broadened its offerings in mathematics and advanced statistics. Mr. Plait, by

incorporating anecdotes on the history and development of calculus, reveals

a rarely seen dimension of calculus to his students. Formulas and solutions

take on reality with real people attempting to find the answer to real problems.

Mr. Plait’s dynamic ability in the teaching of mathematics has become leg-

endary throughout the Graduate Program of the USDA. Mr. Plait was nom-

inated by Drs. Philip Hudson and Ronald MacNab.

The Leo Schubert Award was also given to DR. JAMES EDWARD FALK

for his dedicated and enthusiastic teaching of applied mathematics and op-

erations research, and for his sympathetic and valuable counseling of college

students. Dr. Falk is Professor of Operations Research in the Department of

Operations Research of The George Washington University. One phrase that

typified Dr. Falk is ‘‘clarity of exposition” which underlies his outstanding

AWARDS PROGRAM FOR SCIENTIFIC ACHIEVEMENT IN 1990 165

abilities as a teacher. He is able to present areas of applied mathematics so

coherently and lucidly that students are drawn to the beauty of mathematics

and to the value of its application. Dr. Falk’s deep interest in and commitment

to his subjects, both mathematical and human, are manifest in the excellence

of his teaching. Because of his empathy with students at all levels, Dr. Falk

is a much sought after counselor with respect to academic programs, research,

and professional concerns. Dr. Falk was nominated by Dr. Richard M. Soland.

Bernice Lambertson Award For Teaching of Science in High School

The Bernice Lambertson Award for the Teaching of Science in High School

was granted to MR. WILLIAM JOHN ENTLEY for his devotion to the

teaching profession and endless energy directed toward the preparation of

students for higher education and the real world. After twenty years as a

physicist and engineer, Mr. Entley entered the teaching profession as a high-

school physics teacher. Mr.Entley has promoted a thorough understanding of

physics and stressed the relationship of physics and the other sciences to

everyday life. In his teaching, Mr. Entley emphasizes practicality and logical

objective thinking, the need for intellectual honesty, and the necessity to write

and communicate effectively. He has devoted much of his own time after

hours, on weekends, and during the summer months to extracurricular activ-

ities that involve students in dynamic experiences in the mixed disciplines of

present-day science. Mr. Entley was nominated by Maynard J. Pro.

Special Award For Services to the Washington Academy of Sciences

The awards ceremony ended with the granting of a Special Award of Rec-

ognition to Mr. GROVER C. SHERLIN for his many years of devoted service

to the Washington Academy of Sciences. Mr. Sherlin, in many cases single-

handledly, maintained the records, mailing lists, the library of the Journal of

the Washington Academy of Sciences, and the multitude of administrative

details that provide the life blood of the organization.

The Academy thanks the Chairpersons of each of the selection committees

for their diligent efforts in the selection of outstanding candidates in 1990.

Journal of the Washington Academy of Sciences,

Volume 80, Number 4, Pages 166-170, December 1990

President’s Report to the

Membership for the year 1989-90

Robert H. McCracken

Bethesda, MD

Thanks to a group of dedicated, supportive, and active individuals, the

Academy program from June 1989 through May 1990 was unusually useful.

While it was necessary to make adjustments, solve some problems, and cope

with some obstacles, administrative emphasis was placed upon the useful,

external, charter functions of the Academy.

There was a pressing need to bring expenses within income. This was ac-

complished by steps which simultaneously increased the usefulness of the

Academy.

With the help of a select group of fine people, costs were brought well

under control, and a program of activities and lectures was developed which

was highly useful to the educational system, the scientific community, and the

public.

An objective of this administration, in promotion of science education, was

to foster rapport with the University of the District of Columbia and both the

scientific community and the local neighborhood. To this end, and to reduce

high costs and eliminate severe parking problems, we moved the Academy’s

scientific and board meetings to the University of DC, where both a Metrorail

station and a parking garage are available within a few feet, and the Academy

is not charged for meeting space. (The official headquarters remain, at least

until 1992, at 1101 N. Highland Street, Arlington, VA 22201.)

Speakers and their guests were hosted at dinner in a nearby restaurant. As

in the past, a reception preceded each scientific colloquium, but now with

volunteers, chiefly Edith Corliss, expertly providing a buffet of delightful

variety, at only a small cost. The substantial saving (several hundred dollars

per meeting) made possible another objective: to admit students and faculty

166

PRESIDENT’S REPORT 167

at no cost, and others for a very nominal charge. The much lower cost, far

greater convenience, and current relevance encouraged attendance.

Many scientists and others who attended Academy functions at the Uni-

versity for the first time expressed very favorable impressions.

Another cost-cutting measure improved efficiency and accuracy by having

the Journal sent to Academy Headquarters where it was mailed at a $100

saving by volunteers.

Another objective was to employ the monthly scientific colloquia as a forum

for useful, cross-discipline dissemination of the status, problems, and progress

of leading work in the respective fields.

Program Chair William Busch, of NOAA, brought eminent scientists to our

scientific programs, usually co-sponsored by Academy affiliates, to dissemi-

nate the current status of relevant work at the horizons of their fields.

One of the most valuable functions of the Academy, the Washington Junior

Academy of Sciences continued its excellent contributions to the development

of tomorrow’s leadership under the uniquely skillful guidance of Marylin Krup-

saw, who was authorized to extend her fine work to the intermediate school

level.

Another continuing, very important function for education, to which we

also made some valuable new appointments, is the Joint Board on Science

and Engineering Education. The JBSEE is sponsored by the Academy and

the District of Columbia Council of Engineering and Architectural Societies;

under the dedicated leadership of Dr. Donald Roe, it provides resource people

and contact representatives for all the public, private, and parochial schools

in the metropolitan DC area.

Competent, effective, inspired volunteerism is the life blood of non-profit

organizations. At the risk of inadvertently missing some, for which, if so, I

apologize, I thank the following for their dedication and support:

Dr. Jean Boek, who so ably chaired the nominating committee and assisted

with other details; Dr. Philip L. Brach, Dean of the College of Physical

Sciences, Engineering, and Technology, for his kind hospitality in hosting the

Academy programs at the University of DC; Dr. William S. Busch, Program

Manager, NOAA Undersea Research, who, as program chairman provided

a valuable, relevant series of colloquia; Edith Corliss, both for her function

as Vice President for Affiliate Affairs and for providing the excellent reception

buffets; Dr. Cyrus R. Creveling, who, as chairman of the awards committee,

administered an impressive awards program; Norman Doctor, for his generous

help with the membership data-base and related problems; Dr. Thomas W.

Doeppner, both for his service on the bylaws committee and for helpful advice

on other matters; Dr. Alphonse F. Forziati, for chairing the committee of

168 ROBERT H. McCRACKEN

tellers; Dr. William R. Graver and Dr. J. Terrell Hoffeld, for their help with

the bylaws; Mrs. Marylin Krupsaw, for her outstanding work with students,

as Vice President of the Washington Junior Academy of Sciences; Dr. Stanley

G. Leftwich, both for his fine work as chairman of the bylaws committee, and

for dedicated support in other matters; Eric O. Nystrom, for his substantial

contributions in providing audio-visual services for Academy affairs; Dr. Don-

ald W. Roe, for his valuable work as chairman of the Joint Board on Science

and Engineering Education; Grover C. Sherlin, Treasurer, for more dedicated

service in many more areas than could be listed on this page; Dr. Simon W.

Strauss, our Eminent Scholar in Residence, for his wise counsel, guidance,

and help on many matters.

Grover Sherlin, upon his election as treasurer, acquired and expertly over-

came some serious problems. An independent, outside auditor was engaged

(donated, at no cost to the Academy) who initially found that without the

data that Sherlin had no access to, it was impossible to submit an accurate

audit. As information became available, Norman Doctor kindly volunteered

his effective assistance in assembling data and establishing an improved com-

puterized data-base.

Working diligently with records of past years, Mr. Sherlin, with the coop-

eration of the Internal Revenue Service, developed corrections of previous

years’ irregularities, which led to the refunding to the Academy, with interest,

of several thousand dollars of previously imposed penalties.

Mr. Sherlin also helped with the mountainous task of reestablishing and

updating our affiliate society relations. He was also faced with the necessity

of reassembling a valid membership list from incomplete information.

Dr. Stanley Leftwich chaired a committee to correct several inconsistencies

and errors in the bylaws and to add a needed provision regarding conflict of

interest. The rewrite was passed by the Academy membership in an almost

unanimous (approximately 97 percent) vote.

Very dependable, high-quality audio-visual services were regularly provided

at no cost by Eric Nystrom, using overhead projector, slide projector, radio

microphone, amplifier-speaker system, and recorder, lent by National Capital

Astronomers. We opened our program series with a special, free, public event

at the University of the DC, co-sponsored with National Capital Astronomers:

A Voyager-Neptune Fly-by party for the public, students, and others, free of

charge, all costs privately donated, at no cost to the Academy. A large mi-

crowave antenna was rented to receive the Voyager images from the NASA

satellite, which were projected to the full screen in an auditorium. Enthusiastic

participants saw the Voyager images of Neptune and Triton upon arrival of

PRESIDENT’S REPORT 169

the radio signals at the Earth. A continental breakfast (costs donated) was

also provided at no cost to the guests of the Academy.

It was gratifying to see such public enthusiasm at three o’clock in the morn-

ing, and on until about 9:30 AM.

We thank Dr. Philip Brach, Dean of the College of Physical Sciences,

Engineering, and Technology, for hosting the program, as well as other Acad-

emy programs, which we believe were, as intended, of mutual benefit to the

University and the Academy, as well as to the public, in building good neigh-

borhood relations. We also thank Radio Station WGMS for not only repeat-

edly announcing the program, but also for voluntarily emphasizing it by playing

Gustav Hoslst’s Suite, ‘““The Planets”’ for us!

The September lecture, ““Voyager—Neptune—Triton,” by Dr. Michael

Kaiser, NASA Goddard Space Flight Center, reviewed early scientific results

of that mission. In October, we heard “‘New French Marine Technology Ex-

tends Horizons of Undersea Exploration,” by Dr. Guy Imbert, Centre Na-

tionale de la Researche Scientifique, Marseille. He described two remarkable

current developments that substantially extend human diving depth and time

limitations, using new mixtures of breathing gas and diving equipment. In

November, “A Palladium Curtain Descends over Utah”’ discussion by Dr.

Robert L. Park, Physics Department, University of Maryland, put the hot

topic of “cold fusion” to rest, detailing the reasons. In his January 1990

presentation, ‘Quasars and Quakes,” Dr. Thomas A. Clark, a radio astron-

omer with NASA Goddard Space Flight Center, described his work imme-

diately following the Loma Prieta earthquake, and continuing. With an “‘in-

verted” extremely accurate astronomical star-position measuring technique,

very long baseline interferometry, he used the farthest observable objects in

the universe—quasars—to measure the motions and velocities of the Pacific

and North American tectonic plates, to an accuracy of 1 cm. He described

the work and gave results of these important, ongoing measurements at several

sites in California, Alaska, Hawaii, Japan and elsewhere. In his February

lecture, “Bright Light Eye Damage: Infrared, Ultraviolet, and Blue,’ David

Sliney, an ophthalmic health physicist at Edgewood Arsenal, advised of cur-

rent knowledge on the damage mechanisms of various wavelengths of light,

including the dangerous visible blue. Wear your amber ‘“‘blue blockers”? when

exposed to bright white or blue! In March, well-known meteorologist and

Academy member, Glen Brier, NOAA and University of Colorado, reported

his current discovery of ‘Significant Periodicities in El Nino,’ which have

globel climatic implications. April brought Kurt Stehling, NOAA, with an

assessment of the “Future Potential of Lighter-than-air Craft in Military and

170 ROBERT H. McCRACKEN

Civilian Applications.”” An eminent scientist in this field, he was also the

balloon pilot in the popular IMAX film. “To Fly.”

In May, the annual Academy Awards dinner and ceremony were impres-

sively arranged and conducted by Dr. Cyrus R. Creveling, Chair of the Awards

Committee, at Georgetown University. For the list of awards and awardees,

see Dr. Creveling’s report elsewhere in this issue. I am privileged to add my

hearty congratulations to the winners of these awards.

Other important contacts were made for the Academy for future devel-

opment of the enormous potential of the more than a half hundred professional

societies in DC. Long-range goals, too long for short tenure, but while it

lasted, June 1989—May 1990 was a very good year. I sincerely thank you for

your support.

Journal of the Washington Academy of Sciences,

Volume 80, Number 4, Pages 171-186, December 1990

1990 Washington Academy of Sciences Membership Directory

M = Member; F = Fellow; LM = Life Member; LF = Life Fellow; EM = Emeritus

Member; EF = Emeritus Fellow; NRF = Non-Resident Fellow.

ABDULNUR, SUHEIL F. (Dr) 5715 Glenwood Rd., Bethesda, MD 20817 (F)

ABELSON, PHILIP H. (Dr) 4244 50th St., NW, Washington, DC 20016 (F)

ABRAHAM, GEORGE (Dr) 3107 Westover Dr., SE, Washington, DC 20020 (F)

ABSOLON, KAREL B. (Dr) 11225 Huntover Dr., Rockville, MD 20852 (F)

ACHTER, MEYER R. (Dr) 417 Sth St., SE, Washington, DC 20003 (EF)

ADAMS, CAROLINE L. (Dr) 242 N. Granada St., Arlington, VA 22203 (EM)

ADLER, VICTOR E. (Mr) 8540 Pineway Court, Laurel, MD 20723 (EF)

AFFRONTI, LEWIS F. (Dr) 5003 Woodland Way, Annandale, VA 22003 (F)

ALDRIDGE, MARY H. (Dr) 7904 Hackamore Dr., Potomac, MD 20854-3825 (EF)

ALEXANDER, BENJAMIN H. (Dr) P.O. Box 41126 NE, Washington, DC 20018 (EF)

ALICATA, J. E. (Dr) 1434 Punahou St., Apt. 736, Honolulu, HI 96822 (EF)

ALLEN, J. FRANCES (Dr) P.O. Box 284 (Meeker Hollow Rd.), Roxbury, NY 12474-0284 (NRF)

ANDRUS, EDWARD D. (Mr) 2497 Patricia Court, Falls Church, VA 22043 (M)

ARGAUER, ROBERT J. (Dr) 4208 Everett St., Kensington, MD 20895 (F)

ARONSON, CASPER J. (Mr) 3401 Oberon St., Kensington, MD 20895 (EM)

ARSEM, COLLINS (Mr) 10821 Admirals Way, Potomac, MD 20854 (M)

ARVESON, PAUL T. (Mr) 10205 Folk St., Silver Spring, MD 20902 (F)

AXELROD, JULIUS (Dr) 10401 Grosvenor Place, Apt. 1209, Bethesda, MD 20852 (EF)

AXILROD, BENJAMIN M. (Dr) 9216 Edgewood Dr., Gaithersburg, MD 20877 (EF)

BAILEY, R. CLIFTON (Dr) 6507 Divine St., McLean, VA 22101 (LF)

BAKER, ARTHUR A. (Dr) 5201 Westwood Dr., Bethesda, MD 20816 (EF)

BAKER, LEONARD (Dr) 4720 Sentinel Dr., Bethesda, MD 20816 (M)

BAKER, LOUIS C. W. (Dr) Dept. of Chemistry, Georgetown University, Washington, DC 20057 (F)

BALLARD, LOWELL D. (Mr) 7823 Mineral Springs Dr., Gaithersburg, MD 20877 (F)

BARBOUR, LARRY L. (Mr) Rural Route 1, Box 492, Great Meadows, NJ 07838 (M)

BARTFELD, CHARLES I. (Dr) 6007 Kirby Rd., Bethesda, MD 20817-6247 (M)

BATAVIA, ANDREW I. (Mr) 700 Seventh St., SW, Apt. 813, Washington, DC 20024 (LF)

BAUMANN, ROBERT C. (Mr) 9308 Woodberry St., Seabrook, MD 20706 (F)

BEACH, LOUIS A. (Dr) 1200 Waynewood Blvd., Alexandria, VA 22308-1842 (F)

BECKER, DONALD A. (Mr) 13115 Dauphine St., Silver Spring, MD 20906 (F)

BECKER, EDWIN D. (Dr) N.I.H., Bldg 2, Room 122, Bethesda, MD 20892 (F)

BECKMANN, ROBERT B. (Dr) 10218 Democracy Lane, Potomac, MD 20854 (F)

BEKEY, IVAN (Mr) 4624 Quarter Charge Dr., Annandale, VA 22003 (F)

BENDER, MAURICE (Dr) 16518 N.E. Second Place, Bellevue, WA 98008 (EF)

BENESCH, WILLIAM M. (Dr) 4444 Linnean Ave., NW, Washington, DC 20008 (LF)

BENJAMIN, CHESTER R. (Dr) 315 Timberwood Ave., Silver Spring, MD 20901 (EF)

BENNETT, JOHN A. (Mr) 7405 Denton Rd., Bethesda, MD 20814 (F)

171

172 1990 MEMBERSHIP DIRECTORY

BENSON, WILLIAM M. (Dr) 636 Massachusetts Ave., NE, Washington, DC 20002 (F)

BERGMANN, OTTO (Dr) Dept. of Physics, George Washington University, Washington, DC 20052 (F)

BERKSON, HAROLD (Dr) 12001 Whippoorwill Lane, Rockville, MD 20852 (EM)

BERNETT, MARIANNE K. (Mrs) 5337 Taney Ave., Alexandria, VA 22304 (EM)

BERNSTEIN, BERNARD (Mr) 7420 Westlake Terrace, Apt. 608, Bethesda, MD 20817 (M)

BESTUL, ALDEN B. (Dr) 9400 Overlea Dr., Rockville, MD 20850 (F)

BETTS, ALLEN W. (Mr) 2510 South Ivanhoe Place, Denver, CO 80222 (M)

BHAGAT, SATINDAR M. (Prof) 112 Marine Terrace, Silver Spring, MD 20904 (F)

BICKLEY, WILLIAM E. (Dr) 6516 Fortieth Ave., University Park, Hyattsville, MD 20782 (EF)

BIRKS, LAVERNE S. (Mr) 306 Lands End Dr., Chapin, NC 29036 (EF)

BISHOP, WILLIAM P. (Dr) Desert Research Institute, 2505 Chandler Dr., Suite 1, Las Vegas, NV 89120

(NRF)

BLACKMON, RICHARD F. (Mr) Quintilian Institute, P.O. Box 9351, Arlington, VA 22209-0351 (M)

BLACKSTEN, HARRY RIC (Mr) 4413 N. 18th St., Arlington, VA 22207 (M)

BLANCHARD, DAVID L. (Dr) 1015 McCeney Ave., Silver Spring, MD 20901 (LF)

BLANK, CHARLES A. (Dr) 7085 46th Ave. West., Apt. 173, Bradenton, FL 34210 (NRF)

BLOCH, CAROLYN C. (Mrs) P.O. Box 1889, Rockville, MD 20849 (M)

BLUNT, ROBERT F. (Dr) 5411 Moorland Lane, Bethesda, MD 20814 (F)

BOEK, HEATHER (Dr) SP-FR-5-1, Corning Glass Works, Corning, NY 14831 (M)

BOEK, JEAN K. (Dr) National Graduate University, 1101 N. Highland St., Arlington, VA 22201 (LF)

BOEK, WALTER E. (Dr) 5011 Lowell St., NW, Washington, DC 20016 (F)

BOGNER, M. SUE (Dr) 9322 Friars Rd., Bethesda, MD 20817 (LF)

BONEAU, C. ALAN (Dr) 5305 Waneta Rd., Bethesda, MD 20816 (F)

BORIS, JAY PAUL (Dr) 3516 Duff Dr., Falls Church, VA 22041 (F)

BOTBOL, JOSEPH MOSES (Dr) 60 Curtis St., Falmouth, MA, 02540 (F)

BOURGEOIS, LOUIS D. (Dr) 8701 Bradmoor Dr., Bethesda, MD 20817 (F)

BOURGEOIS, MARIE J. (Dr) 8701 Bradmoor Dr., Bethesda, MD 20817 (F)

BOWMAN, THOMAS E. (Dr) Smithsonian Institution, Invertebrate Zoology, NHB Mail Stop 163,

Washington, DC 20560 (F)

BOYD, WENDELL J. (Mr) 6307 Balfour Dr., Hyattsville, MD 20782 (M)

BRADY, ROBERT F., JR. (Dr) 706 Hope Lane, Gaithersburg, MD 20878 (F)

BRANCATO, EMANUEL L. (Dr) 7370 Hallmark Rd., Clarksville, MD 21029 (F)

BRANDEWIE, DONALD F. (Mr) 6811 Field Master Dr., Springfield, VA 22153 (F)

BRENNER, ABNER (Dr) 7204 Pomander Lane, Chevy Chase, MD 20815 (F)

BRIER, GLENN W. (Mr) 1729 N. Harrison St., Arlington, VA 22205 (LF)

BRISKMAN, ROBERT D. (Mr) 6728 Newbold Dr., Bethesda, MD 20817 (F)

BROADHURST, MARTIN G. (Dr) 116 Ridge Rd., Box 163, Washington Grove, MD 20880 (F)

BROWN, BRICKMAN (Mr) 6811 Nesbitt Place, McLean, VA 22101 (M)

BROWN, ELISE A. B. (Dr) 6811 Nesbitt Place, McLean, VA 22101 (LF)

BRUCK, STEPHEN D. (Dr) 3247 St. Augustine Court, Olney, MD 20832 (F)

BRYAN, MILTON M. (Mr) 3322 N. Glebe Rd., Arlington, VA 22207 (M)

BURAS, EDMUND M., JR. (Mr) 824 Burnt Mills Ave., Silver Spring, MD 20901 (EF)

BUSCH, WILLIAM S. (Mr) 1035 Sun Valley Dr., Annapolis, MD 21401 (M)

BUTTERMORE, DONALD O. (Mr) 1519 N. Utah St., Arlington, VA 22207 (LF)

CACERES, CESAR A. (Dr) 1759 Que St., NW, Washington, DC 20009 (M)

CAHNMAN, HUGO N. (Mr) CASSO-SOLAR Corp., P.O. Box 163, Pomona, NY 10970 (M)

CAMPBELL, LOWELL E. (Mr) 14000 Pond View Rd., Silver Spring, MD 20905 (F)

CANNON, EDWARD W. (Dr) 18023 134th Ave., Sun City West, AZ 85375 (NRF)

CANTELO, WILLIAM W. (Dr) 11702 Wayneridge St., Fulton, MD 20759 (F)

CARROLL, WILLIAM R. (Dr) 4802 Broad Brook Dr., Bethesda, MD 20814 (EF)

1990 MEMBERSHIP DIRECTORY 173

CARTER, HUGH (Dr) 158 N. Harrison St., Princeton, NJ 08540 (EF)

CASH, EDITH K. (Ms) 505 Clubhouse Rd., Binghamton, NY 13903 (EF)

CERRONI, MATTHEW, J. (Mr) 12538 Browns Ferry Rd., Herndon, VA 22070 (M)

CHAMBERS, RANDALL M. (Dr) 2704 Winstead Circle, Wichita, KS 67226 (NRF)

CHAPLIN, HARVEY R., JR. (Dr) 1561 Forest Villa Lane, McLean, VA 22101 (F)

CHAPMAN, ROBERT D. (Dr) 10976 Swansfield Rd., Columbia, MD 21044 (F)

CHEEK, CONRAD H. (Dr) 4334 H St., SE, Washington, DC 20019 (F)

CHEZEM, CURTIS G. (Dr) 3378 Wisteria St., Eugene, OR 97404 (EF)

CHI, MICHAEL (Dr) 2000 N. 14th St., Suite 310, Arlington, VA 22201 (F)

CHRISTIANSEN, MERYL N. (Dr) 4603 Barbara Dr., Beltsville, MD 20705 (EF)

CIVEROLO, EDWIN L. (Dr) 12340 Shadetree Lane, Laurel, MD 20708 (F)

CLAIRE, CHARLES N. (Mr) 4403 14th St., NW, Washington, DC 20011 (F)

CLARK, GEORGE E., JR. (Mr) 4022 N. Stafford St., Arlington, VA 22207 (F)

CLEVEN, GALE W. (Dr) P.O. Box 998, Maggie Valley, NC 28751 (EF)

CLIFF, RODGER A. (Dr) 2331 Cheshire Way, Redwood City, CA 94061 (M)

CLINE, THOMAS LYTTON (Dr) 13708 Sherwood Forest Dr., Silver Spring, MD 20904 (F)

COATES, JOSEPH F. (Mr) 3738 Kanawha St., NW, Washington, DC 20015 (F)

COFFEY, TIMOTHY P. (Dr) Naval Research Laboratory, Code 1001, Washington, DC 20375-5000 (F)

COLE, RALPH I. (Mr) 3705 S. George Mason Dr., Apt. 1515, South Falls Church, VA 22041 (F)

COLWELL, RITA R. (Dr) Dept. Microbiology, University of Maryland, College Park, MD 20742 (LF)

COMPTON, W. DALE (Dr) Ford Motor Company, P.O. Box 1603, Dearborn, MI 48121 (NRF)

CONDELL, WILLIAM J., JR. (Dr) 4511 Gretna Green, Bethesda, MD 20814 (F)

CONNELLY, EDWARD MCD. (Mr) 1625 Autumnwood Dr., Reston, VA 22094 (F)

COOK, RICHARD K. (Dr) 4111 Bel Pre Rd., Rockville, MD 20853 (F)

COOPER, KENNETH W. (Dr) 4497 Picacho Dr., Riverside, CA 92507 (EF)

CORLISS, EDITH L. R. (Mrs) 2955 Albemarle St., NW, Washington, DC 20008 (LF)

CORMACK, JOHN G. (Mr) 10263 Gainsborough Rd., Potomac, MD 20854 (M)

COSTRELL, LOUIS (Mr) 621 Interlachen Dr., Silver Spring, MD 20906 (F)

COTHERN, C. RICHARD (Dr) 4732 Merivale Rd., Chevy Chase, MD 20815 (F)

COTTERILL, CARL H. (Mr) 6030 Corland Court, McLean, VA 22101 (F)

CRAIN, DARRELL C. (M.D.) 6422 Garnett Dr., Chevy Chase, MD 20815 (F)

CREVELING, CYRUS R. (Dr) 4516 Amherst Lane, Bethesda, MD 20814 (F)

CULBERT, DOROTHY K. (Mrs) 109 Calle La Pena, Santa Fe, NM 87501 (EF)

CURRAN, HAROLD R. (Dr) 19310 Club House Rd., Gaithersburg, MD 20879-3829 (EF)

CURRIE, CHARLES L., S. J. (Dr) Georgetown University, Washington, DC 20057 (F)

CURTIS, ROGER W. (Dr) 6308 Valley Rd., Bethesda, MD 20817 (EF)

CUTKOSKY, ROBERT DALE (Mr) 19150 Roman Way, Gaithersburg, MD 20879 (F)

D’ANTONIO, WILLIAM V. (Dr) 3701 Connecticut Ave., NW, Apt. 818, Washington, DC 20008 (F)

DAVIS, ANDREW VIVIAN (Mr) 4201 Mass. Ave., NW, Washington, DC 20016 (M)

DAVIS, CHARLES M., JR. (Dr) 8458 Portland Place, McLean, VA 22102 (M)

DAVIS, MARION MACLEAN (Dr) Crosslands, Apt. 100, Kennett Square, PA 19348 (LF)

DAVIS, MILES (Dr) 1214 Bolton St., Baltimore, MD 21217-4111 (M)

DAVIS, ROBERT E. (Dr) 1793 Rochester St., Crofton, MD 21114 (F)

DAVISON, MARGARET C. (Mrs) 2928 N. 26th St., Arlington, VA 22207 (M)

DAVISSON, JAMES W. (Dr) 400 Cedar Ridge Rd., Oxon Hill, MD 20745 (EF)

DAWSON, ROY C. (Dr) 4019 Beechwood Rd., Hyattsville, MD 20782 (EF)

DAWSON, VICTOR C. D. (Dr) 9406 Curran Rd., Silver Spring, MD 20901 (F)

DEAL, GEORGE E. (Dr) 6245 Park Rd., McLean, VA 22101 (F)

DEBERRY, MARIAN B. (Mrs) 3608 17th St., NE, Washington, DC 20018 (EM)

DEDRICK, ROBERT L. (Dr) 1633 Warner Ave., McLean, VA 22101 (F)

174 19909 MEMBERSHIP DIRECTORY

DELANEY, WAYNE R. (Mr) 602 Oak St., Farmville, VA 23901-1118 (M)

DEMING, W. EDWARDS (Dr) 4924 Butterworth Place, NW, Washington, DC 20016 (F)

DEMUTH, HAL P. (Cdr) 24 S. Washington St., Winchester, VA 22601 (NRF)

DENNIS, BERNARD K. (Mr) 915 Country Club Dr., Vienna, VA 22180 (F)

DESLATTES, RICHARD D. (Dr) 610 Aster Blvd., Rockville, MD 20850 (F)

DEUTSCH, STANLEY (Dr) 7109 Lava Rock Lane, Bethesda, MD 20817 (EF)

DEVEY, GILBERT B. (Mr) 2801 New Mexico Ave., NW, Washington, DC 20007 (F)

DEVIN, CHARLES, JR. (Dr) 629 Blossom Dr., Rockville, MD 20850 (M)

DEVOE, JAMES R. (Mr) 11708 Parkridge Dr., Gaithersburg, MD 20878 (F)

DEWIT, ROLAND (Dr) 11812 Tifton Dr., Rockville, MD 20854 (F)

DICKSON, GEORGE (Mr) 52 Orchard Way North, Rockville, MD 20854 (F)

DIMOCK, DAVID A. (Mr) 4291 Molesworth Terrace, Mt. Airy, MD 21771 (EF)

DOCTOR, NORMAN (Mr) 6 Tegner Court, Rockville, MD 20850 (F)

DOEPPNER, THOMAS W. (Col) 8323 Orange Court, Alexandria, VA 22309 (LF)

DONALDSON, EVA G. (Ms) 3941 Ames St., NE, Washington, DC 20019 (F)

DONALDSON, JOHANNA B. (Mrs) 3020 N. Edison St., Arlington, VA 22207 (F)

DONNERT, HERMANN J. (Dr) Dept. of Nuclear Engineering, Ward Hall, Kansas State University,

Manhattan, KS 66506-7039 (NRF)

DOOLING, ROBERT J. (Dr) 4812 Mori Dr., Rockville, MD 20853 (F)

DOUGLAS, THOMAS B. (Dr) 3031 Sedgwick St., NW, Washington, DC 20008 (EF)

DRAEGER, HAROLD R. (Dr) 1201 N. 4th St., Tucson, AZ 85705 (EF)

DUBEY, SATYA D. (Dr) 7712 Groton Rd., West Bethesda, MD 20817 (EF)

DUFFEY, DICK (Dr) Chem-Nuclear Engineering Dept., University of Maryland, College Park, MD

20742 (LF)

DUKE, JAMES A. (Mr) 8210 Murphy Rd., Fulton, MD 20759 (F)

DUNCOMBE, RAYNOR L. (Dr) 1804 Vance Circle, Austin, TX 78701 (NRF)

DUNKUM, WILLIAM W. (Dr) 1561 Pensacola St., Apt. 2306, Honolulu, HI 96822 (EF)

DUPONT, JOHN ELEUTHERE (Mr) P.O. Box 297, Newtown Square, PA 19073 (NRF)

DURIE, EDYTHE G. (Mrs) 4408 Braeburn Dr., Fairfax, VA 22032-1845 (F)

EDINGER, STANLEY E. (Dr) 5901 Montrose Rd., Apt. 404N, Rockville, MD 20852 (F)

EDMUND, NORMAN W. (Mr) 407 Northeast 3rd Ave., Fort Lauderdale, FL 33301 (M)

EISENHART, CHURCHILL (Dr) 9629 Elrod Rd., Kensington, MD 20895 (EF)

EL-BISI, HAMED M. (Dr) 258 Bishops Forest Dr., Waltham, MA 02154 (M)

ELASSAL, ATEF A. (Dr) 1538 Red Rock Court, Vienna, VA 22182 (F)

ELISBERG, F. MARILYN (Mrs) 4008 Queen Mary Drive, Olney, MD 20832 (F)

ELLIOTT, F. E. (Dr) 7507 Grange Hall Dr., Fort Washington, MD 20744 (EF)

EMERSON, K. C. (Dr) 560 Boulder Dr., Sanibel, FL 33957 (F)

ENDO, BURTON Y. (Dr) 9215 Wofford Lane, College Park, MD 20740 (F)

ENGLAR, ROBERT JOHN (Mr) 3269 Catkin Court, Marietta, GA 30066 (F)

ETTER, PAUL C. (Mr) 16609 Bethayres Rd., Rockville, MD 20855-2043 (F)

ETZIONI, AMITAI (Dr) 2700 Virginia Ave., NW, Apt. 1002, Washington, DC 20037 (F)

EVERSTINE, GORDON C. (Dr) 12020 Golden Twig Court, Gaithersburg, MD 20878 (F)

EWERS, JOHN C. (Mr) 4432 N. 26th Rd., Arlington, VA 22207 (EF)

FARLEE, CORALIE (Dr) 389 O St., SW, Washington, DC 20024 (F)

FARMER, ROBERT F.,, III (Dr) 21954 N. Green Forest Rd., Barrington, IL 60010-2460 (NRF)

FAULKNER, JOSEPH A. (Mr) Route 2, Box 185 C, Lewes, DE 19958 (F)

FAUST, WILLIAM R. (Dr) 5907 Walnut St., Temple Hills, MD 20748 (F)

FEARN, JAMES E. (Dr) 4446 Alabama Ave., SE, Washington, DC 20019 (F)

FEINGOLD, S. NORMAN (Dr) 9707 Singleton Dr., Bethesda, MD 20817 (F)

FERRELL, RICHARD A. (Dr) 6611 Wells Parkway, University Park, MD 20782 (F)

1990 MEMBERSHIP DIRECTORY 175

FILIPESCU, NICOLAE (Dr) 5020 Little Falls Rd., Arlington, VA 22207 (F)

FINKELSTEIN, ROBERT (Mr) Robotic Technology, Inc., 10001 Crestleigh Lane, Potomac, MD 20854

(M)

FINN, EDWARD J. (Dr) 7500 Lynn Dr., Chevy Chase, MD 20815 (F)

FISHER, JOEL L. (Dr) 4033 Olley Lane, Fairfax, VA 22030 (M)

FLINN, DAVID R. (Dr) 9714 Wildflower Circle, Tuscaloosa, AL 35405 (NRF)

FLORIN, ROLAND E. (Dr) 7407 Cedar Ave., Takoma Park, MD 20912 (EF)

FLYNN, JOSEPH H. (Dr) 5309 Iroquois Rd., Bethesda, MD 20816 (F)

FOCKLER, HERBERT H. (Mr) 10710 Lorain Ave., Silver Spring, MD 20901 (EM)

FONER, SAMUEL N. (Dr) Applied Physics Laboratory, JHU, 11100 Johns Hopkins Rd., Laurel, MD

20723 (F)

FOOTE, RICHARD H. (Dr) Box 166, Lake of the Woods, Locust Grove, VA 22508 (F)

FORZIATI, ALPHONSE F. (Dr) 15525 Prince Frederick Way, Silver Spring, MD 20906 (F)

FORZIATI, FLORENCE H. (Dr) 15525 Prince Frederick Way, Silver Spring, MD 20906 (F)

FOSTER, AUREL O. (Dr) 4613 Drexell Rd., College Park, MD 20740 (EF)

FOURNIER, ROBERT O. (Dr) 108 Paloma Rd., Portola Valley, CA 94028 (NRF)

FOWLER, WALTER B. (Mr) 9404 Underwood St., Seabrook, MD 20706 (M)

FOX, DAVID W. (Dr) 136 Lind Hall, University of Minnesota, 207 Church St., S.E., Minneapolis, MN

55455 (F)

FOX, WILLIAM B. (Dr) 1813 Edgehill Dr., Alexandria, VA 22307 (F)

FRANKLIN, JUDE E. (Dr) 7616 Carteret Rd., Bethesda, MD, 20817-2021 (F)

FRAVEL, DEBORAH R. (Dr) Soilborne Diseases Laboratory, Room 275, Bldg, 011A, BARC West,

Beltsville, MD 20705 (F)

FREEMAN, ANDREW F. (Mr) 5012 N. 33rd St., Arlington, VA 22207 (EM)

FRIEDMAN, MOSHE (Dr) 4511 Yuma St., NW, Washington, DC 20016 (F)

FRIESS, SEYMOUR L. (Dr) 6522 Lone Oak Court, Bethesda, MD 20817 (F)

FRUSH, HARRIET L. (Dr) 4912 New Hampshire Ave., NW, Apt. 104, Washington, DC 20011 (F)

FURUKAWA, GEORGE T. (Dr) 1712 Evelyn Dr., Rockville, MD 20852 (F)

GAGE, WILLIAM W. (Dr) 10 Trafalgar St., Rochester, NY 14619 (F)

GALASSO, GEORGE J. (Dr) 636 Crocus Dr., Rockville, MD 20850 (F)

GARVIN, DAVID (Dr) 18700 Walker’s Choice Rd., Apt. 807, Gaithersburg, MD 20879 (F)

GAUNAURD, GUILLERMO C. (Dr) 4807 Macon Rd., Rockville, MD 20852 (F)

GENTRY, JAMES W. (Prof) Chem-Nuclear Engineering Dept., University of Maryland, College Park,

MD 20742 (F)

GHAFFARI, ABOLGHASSEM (Dr) 7532 Royal Dominion Dr., West Bethesda, MD 20817 (LF)

GHOSE, RABINDRA NATH (Dr) 8167 Mulholland Terrace, Los Angeles, CA 90046 (NRF)

GILLASPIE, A. GRAVES, JR. (Dr) 141 Cloister Dr., Peachtree City, GA 30269

GIST, LEWIS A. (Dr) 1336 Locust Rd., NW, Washington, DC 20012 (EF)

GLASER, HAROLD (Dr) 1346 Bonita St., Berkeley, CA 94709 (EF)

GLASGOW, AUGUSTUS R., JR. (Dr) 4116 Hamilton St., Hyattsville, MD 20781 (EF)

GLOVER, ROLFE E., III (Prof) 7006 Forest Hill Dr., Hyattsville, MD 20782 (F)

GLUCKMAN, ALBERT G. (Mr) 11235 Oakleaf Dr., Apt. 1619, Silver Spring, MD 20901 (F)

GLUCKSTERN, ROBERT L. (Dr) 10903 Wickshire Way, Rockville, MD 20852 (F)

GOFF, JAMES F. (Dr) 3405 34th Place, NW, Washington, DC 20016 (F)

GOLDEN, A. MORGAN (Mr) 9110 Drake Place, College Park, MD 20740 (F)

GOLUMBIC, CALVIN (Dr) 6000 Highboro Dr., Bethesda, MD 20817 (EF)

GONET, FRANK (Dr) 4007 N. Woodstock St., Arlington, VA 22207 (EF)

GOODE, ROBERT J. (Mr) 2402 Kegwood Lane, Bowie, MD 20715 (F)

GORDON, RUTH E. (Dr) American Type Culture Collection, 12301 Parklawn Dr., Rockville, MD

20852 (EF)

GRAVER, WILLIAM R. (Dr) 6137 N. 9th Rd., Arlington, VA 22205 (M)

176 19909 MEMBERSHIP DIRECTORY

GRAY, IRVING (Dr) 9215 Quintana Dr., Bethesda, MD 20817 (F)

GREENOUGH, M. L. (Mr) Greenough Data Associates, 616 Aster Blvd., Rockville, MD 20850 (F)

GREER, SANDRA C. (Dr) Chemistry Department, University of Maryland, College Park, MD

20742 (F)

GRISAMORE, NELSON T. (Prof) 9536 E. Bexhill Dr., Kensington, MD 20895 (EF)

GROSS, DONALD (Mr) 3530 N. Rockingham St., Arlington, VA 22213 (F)

GROSS, ROSALIND L. (Dr) 10816 Antigua Terrace, Apt. 202, Rockville, MD 20852-5517 (M)

GROSSLING, BERNARDO F. (Dr) 10903 Amherst Ave., Apt. 241, Silver Spring, MD 20902 (F)

GRUNTFEST, IRVING (Dr) 140 Lake Carol Dr., West Palm Beach, FL 33411-2132 (EF)

GURNEY, ASHLEY B. (Dr) Manor Care Nursing Center, 550 S. Carlin Spring Rd., Arlington, VA

22204 (EF)

HACSKAYLO, EDWARD (Dr) 2169 Bluebell Rd., Port Republic, MD 20676 (F)

HAENNI, EDWARD O. (Dr) 7907 Glenbrook Rd., Bethesda, MD 20814 (F)

HAGN, GEORGE N. (Mr) 4208 Sleepy Hollow Rd., Annandale, VA 22003 (LM)

HAINES, KENNETH (Mr) 3542 N. Delaware St., Arlington, VA 22207 (F)

HAMER, WALTER J. (Dr) 3028 Dogwood St., NW, Washington, DC 20015 (EF)

HAMMER, GUY S., III (Mr) 8902 Ewing Dr., Bethesda, MD 20817 (F)

HAMMER, JEAN H. (Mrs) 8902 Ewing Dr., Bethesda, MD 20817 (M)

HAND, CADET S., JR. (Prof) Star Route, Bodega Bay, CA 94923 (EF)

HANEL, RUDOLPH A. (Dr) 31 Brinkwood Rd., Brookeville, MD 20833 (F)

HANIG, JOSEPH P. (Dr) 822 Eden Court, Alexandria, VA 22308 (F)

HANSEN, LOUIS S. (Dr) Oral Pathology, Room S-524, OM&D, University of California, San Francisco,

CA 94143-0424 (EF)

HANSEN, MORRIS H. (Mr) 13532 Glen Mill Rd., Rockville, MD 20850 (LF)

HARR, JAMES W. (Mr) 9503 Nordic Dr., Lanham, MD 20706 (M)

HARRINGTON, FRANCIS D. (Dr) 4600 Ocean Beach Blvd., Apt. 204, Cocoa Beach, FL 32931 (F)

HARRINGTON, MARSHALL C. (Dr) 10450 Lottsford Rd., Apt. 2207, Mitchellville, MD 20716 (EF)

HARRIS, MILTON (Dr) 4201 Connecticut Ave., NW, Apt. 610, Washington, DC 20008 (F)

HARTLEY, JANET WILSON (Dr) N.I.H., National Institute of Allergy and Infectious Diseases,

Laboratory of Immunopathology, Bethesda, MD 20892 (F)

HARTMANN, GREGORY K. (Dr) 10701 Keswick St., Garrett Park, MD 20896 (EF)

HARTZLER, MARY P. (Ms) 1250 S. Washington St., Apt. 203, Alexandria, VA 22314 (M)

HASKINS, CARYL P. (Dr) 1545 18th St., NW, Suite 810, Washington, DC 20037 (EF)

HASS, GEORG H. (Mr) 7728 Lee Ave., Alexandria, VA 22308-1003 (EF)

HAUPTMAN, HERBERT A. (Dr) The Medical Foundation of Buffalo, Inc., 33 High St., Buffalo, NY

14203-1196 (F)

HAYDEN, GEORGE A. (Dr) 1312 Juniper St., NW, Washington, DC 20012 (EF)

HAYNES, ELIZABETH D. (Mrs) 4149 N. 25th St., Arlington, VA 22207 (M)

HEADLEY, ANNE RENOUF (PhD, JD) The Metropolitan Square, Suite 330, 655 15th St., NW, Wash-

ington, DC 20005 (F)

HEIFFER, MELVIN H. (Dr) Whitehall, Apt. 701, 4977 Battery Lane, Bethesda, MD 20814 (F)

HENNEBERRY, THOMAS J. (Dr) 1409 E. Northshore Dr., Tempe, AZ 85283 (F)

HERMACH, FRANCIS L. (Mr) 2415 Eccleston St., Silver Spring, MD 20902 (F)

HERMAN, ROBERT (Dr) 8434 Antero Dr., Austin, TX 78759 (EF)

HERSEY, JOHN B. (Mr) 923 Harriman St., Great Falls, VA 22066 (M)

HEYDEN, FRANCIS J., S. J. (Dr) Manila Observatory, Solar Optical Div., APO, San Francisco

96528 (EF)

HEYER, W. RONALD (Dr) Amphibian and Reptile, NHB, Mail Stop 162, Smithsonian Institution,

Washington, DC 20560 (F)

HIBBS, EUTHYMIA (Dr) 7302 Durbin Terrace, Bethesda, MD 20817 (M)

1990 MEMBERSHIP DIRECTORY 177

HILLABRANT, WALTER J. (Dr) 1927 38th St., NW, Washington, DC 20007 (M)

HILSENRATH, JOSEPH (Mr) 9603 Brunett Ave., Silver Spring, MD 20901 (F)

HOBBS, ROBERT B. (Dr) 7715 Old Chester Rd., Bethesda, MD 20817 (F)

HOFFELD, J. TERRELL (Dr) 11307 Ashley Dr., Rockville, MD 20852-2403 (M)

HOGE, HAROLD J. (Dr) 65 Grove St., Apt. 148, Wellesley, MA 02181 (EF)

HOLLINGSHEAD, ARIEL (Dr) 3637 Van Ness St., NW, Washington, DC 20008 (EF)

HOLSHOUSER, WILLIAM L. (Mr) P.O. Box 1475, Banner Elk, NC 28604 (F)

HONIG, JOHN G. (Dr) 7701 Glenmore Spring Way, Bethesda, MD 20817 (F)

HOOVER, LARRY A. (Mr) Gaston Co., P.O. Box 1578, 212 West Main Ave., Gastonia, NC 28053-

1578 (M)

HOPP, HENRY (Dr) 6604 Michaels Dr., Bethesda, MD 20817 (EF)

HOPP, THEODORE H. (Dr) 11911 Hitching Post Lane, Rockville, MD 20852-4489 (M)

HOPPS, HOPE E. (Mrs) 1762 Overlook Dr., Silver Spring, MD 20903 (EF)

HORNSTEIN, IRWIN (Dr) 5920 Bryn Mawr Rd., College Park, MD 20740 (EF)

HOROWITZ, EMANUEL (Dr) 14100 Northgate Dr., Silver Spring, MD 20906 (F)

HOWARD, DARLENE V. (Dr) Department of Psychology, Georgetown University, Washington, DC

20057 (F)

HOWARD, JAMES H., JR. (Dr) 3701 Cumberland St., NW, Washington, DC 20016 (F)

HOWELL, BARBARA F. (Dr) 206 Baybourne Dr., Arnold, MD 21012 (F)

‘HOYT, JAMES A., JR. (Mr) 3717 Thoroughbred Lane, Owings Mills, MD 21117 (M)

HUANG, KUN-YEN (Dr) 1445 Laurel Hill Rd., Vienna, VA 22180 (F)

HUDSON, COLIN M. (Dr) 143 S. Wildflower Rd., Asheville, NC 28804 (NRF)

HUGH, RUDOLPH (Dr) Microbiology Department, George Washington University Medical School, 2300

Eye St., NW, Washington, DC 20037 (F)

HUHEEY, JAMES E. (Dr) Department of Zoology, Southern Illinois University, Carbondale, IL

62901 (LF)

HUMMEL, JOHN M. (Mr) 200 Harry S. Truman Pkwy., 2nd Floor, Annapolis, MD 21401 (M)

HUMMEL, LANI S. (Ms) 9312 Fairhaven Ave., Upper Marlboro, MD 20772 (M)

HUNTER, RICHARD S. (Mr) 1703 Briar Ridge Rd., McLean, VA 22101 (EF)

HUNTER, WILLIAM R. (Mr) 6705 Caneel Court, Springfield, VA 22152 (F)

HURDLE, BURTON G. (Mr) 6222 Berkley Rd., Alexandria, VA 22307 (F)

HURTT, WOODLAND (Dr) Dynamac Corporation, 11140 Rockville Pike, Rockville, MD 20852 (M)

HUTTON, GEORGE L. (Mr) 1086 Continental Ave., Melbourne, FL 23940 (EF)

IRVING, GEORGE W., JR. (Dr) 4601 North Park Ave., Chevy Chase, MD 20815 (LF)

IRWIN, GEORGE R. (Dr) 7306 Edmonston Rd., College Park, MD 20740 (F)

ISBELL, HORACE S. (Dr) 3401 38th St., NW, Apt. 216, Washington, DC 20016 (F)

ISENSTEIN, ROBERT S. (Dr) 11710 Caverly Ave., Beltsville, MD 20705 (M)

JACKSON, DAVID J. (Dr) 13711 S.W. 90th Ave., M 111, Miami, FL 33176-6921 (F)

JACKSON, JO-ANNE A. (Dr) 14711 Myer Terrace, Rockville, MD 20853 (LF)

JACOX, MARILYN E. (Dr) 10203 Kindly Court, Gaithersburg, MD 20879 (F)

JAMES, HENRY M. (M) 6707 Norview Court, Springfield, VA 22152 (M)

JEN, CHIH K. (Dr) 10203 Lariston Lane, Silver Spring, MD 20903 (EF)

JENSEN, ARTHUR S. (Dr) Westinghouse D & E Center, Box 1521, Baltimore, MD 21203 (LF)

JERNIGAN, ROBERT W. (Dr) 14805 Clavel St., Rockville, MD 20853 (F)

JESSUP, STUART D. (Dr) 746 N. Emerson St., Arlington, VA 22203 (F)

JOHNSON, DANIEL P. (Dr) P.O. Box 359, Folly Beach, SC 29439 (EF)

JOHNSON, EDGAR M. (Dr) 5315 Renaissance Court, Burke, VA 22015 (LF)

JOHNSON, PHYLLIS T. (Dr) 4721 East Harbor Dr., Friday Harbor, WA 98250 (EF)

JONES, HOWARD S., JR. (Dr) 3001 Veazey Terrace, NW, Apt. 1310, Washington, DC 20008 (LF)

178 1990 MEMBERSHIP DIRECTORY

JONG, SHUNG-CHANG (Dr) American Type Culture Collection, 12301 Parklawn Dr., Rockville, MD

20852 (LF)

JORDAN, GARY BLAKE (Dr) 13392 Fallen Leaf Rd., Poway, CA 92064 (LM)

JOYCE, PRISCILLA G. (Ms) 605 N. Emerson St., Arlington, VA 22203 (M)

KAISER, HANS E. (Dr) 433 Southwest Dr., Silver Spring, MD 20901 (M)

KANTOR, GIDEON (Mr) 10702 Kenilworth Ave., Garrett Park, MD 20896-0553 (M)

KAPER, JACOBUS M. (Dr) 115 Hedgewood Dr., Greenbelt, MD 20770 (F)

KAPETANAKOS, C. A. (Dr) 4601 North Park Ave., Apt. 921, Chevy Chase, MD 20815 (F)

KARP, SHERMAN (Dr) 10205 Counselman Rd., Potomac, MD 20854 (F)

KARR, PHILLIP R. (Dr) 5507 Calle De Arboles, Torrance, CA 90505 (EF)

KAUFMAN, H. PAUL (Lt. Col) 137 Golden Isle Dr., Apt. 302A, Hallandale, FL 33009 (EF)

KAZYAK, KRISTIN R. (Ms) 2145 Hilltop Place, Falls Church, VA 22043 (M)

KEARNEY, PHILIP C. (Dr) 8416 Shears Court, Laurel, MD 20707 (F)

KEISER, BERNHARD E. (Dr) 2046 Carrhill Rd., Vienna, VA 22180 (F)

KESSLER, KARL G. (Dr) National Institute of Standards and Technology, A505, Bldg 101, Gaithersburg,

MD 20899 (F)

KIRK, KENNETH L. (Dr) National Institutes of Health, Bldg. 8A, B1A02, Bethesda, MD 20892 (F)

KLEBANOFF, PHILIP S. (Mr) 6412 Tone Dr., Bethesda, MD 20817 (EF)

KLINGSBERG, CYRUS (Dr) 1318 Deerfield Dr., State College, PA 16803 (NRF)

KLINMAN, DENNIS MARC (Dr) 10401 Grosvenor Place, Suite 725, Rockville, MD 20852 (F)

KNOX, ARTHUR S. (Mr) 2008 Columbia Rd., NW, Washington, DC 20009 (M)

KNUTSON, LLOYD V. (Dr) Agricultural Research Center, Room 001, Bldg. 003, Beltsville, MD

20705 (F)

KRAMER, CAROLYN M. (Dr) B.R.A.D., The Gillette Company, Gillette Park, 5G-2, Boston, MA

02106 (NRF)

KROP, STEPHEN (Dr) 7908 Birnam Wood Dr., McLean, VA 22102 (F)

KRUGER, JEROME (Dr) 619 Warfield Dr., Rockville, MD 20850 (F)

KRUPSAW, MARYLIN (Mrs) 10208 Windsor View Dr., Potomac, MD 20854 (LF)

LANG, MARTHA E. C. (Mrs) Kennedy-Warren, 3133 Connecticut Ave., NW, Apt. 625, Washington,

DC 20008 (EF)

LANG, SCOTT W. (Mr) 462 Severnside Dr., Severna Park, MD 21146-2216 (M)

LANG, TERESA C. H. (Mrs) 462 Severnside Drive, Severna Park, MD 21146-2216 (M)

LANGSTON, JOANN H. (Ms) 14514 Faraday Dr., Rockville, MD 20853 (F)

LAPHAM, EVAN G. (Mr) 2242 S.E. 28th St., Cape Coral, FL 33904 (EF)

LAPLANT, WILLIAM P. (Mr) P.O. Box 2130, Arlington, VA 22202-0130 (M)

LAWSON, ROGER H. (Dr) 4912 Ridgeview Lane, Bowie, MD 20715 (F)

LEE, MARK A. (Mr) 5539 Columbia Pike, Apt. 407, Arlington, VA 22204 (M)

LEE, RICHARD H. (Dr) 5 Angola By The Bay, Lewes, DE 19958 (EF)

LEFTWICH, STANLEY G., P. E. (Dr) 3909 Belle Rive Terrace, Alexandria, VA 22309 (LF)

LEIBOWITZ, LAWRENCE M. (Dr) 3903 Laro Court, Fairfax, VA 22031 (F)

LEINER, ALAN L. (Mr) 850 Webster St., Apt. 635, Palo Alto, CA 94301 (EF)

LEJINS, PETER P. (Dr) 7114 Eversfield Dr., College Heights Estates, Hyattsville, MD 20782 (F)

LENTZ, PAUL LEWIS (Dr) 5 Orange Court, Greenbelt, MD 20770 (F)

LESSOFF, HOWARD (Mr) O.N.R. EUROPE, Box 39, FPO, New York 09510-0700 (F)

LETTIERI, THOMAS R. (Mr) 10705 Hunters Chase Lane, Damascus, MD 20872 (M)

LEVIN, RONALD L. (Dr) 5012 Continental Dr., Olney, MD 20832 (F)

LEVINSON, NANETTE S. (Dr) CTA-Hurst 206, American University, Washington, DC 20016 (M)

LEVY, SAMUEL (Mr) 2279 Preisman Dr., Schenectady, NY 12309 (EF)

LEWIS, A. D., P. E. (Mr) 3476 Mount Burnside Way, Woodbridge, VA 22192 (M)

1990 MEMBERSHIP DIRECTORY 179

LEY, HERBERT L. (M.D.) 1160 Rockville Pike, Suite 208, P.O. Box 2047, Rockville, MD 20847-

2047 (F)

LIBELO, LOUIS F. (Mr) 9413 Bulls Run Parkway, Bethesda, MD 20817 (LF)

LIEBLEIN, JULIUS (Dr) 1621 East Jefferson St., Rockville, MD 20852 (EF)

LIEBOWITZ, HAROLD (Dr) School of Engineering and Applied Science, George Washington Univer-

sity, Washington, DC 20052 (F)

LINDSEY, IRVING (Mr) 202 E. Alexandria Ave., Alexandria, VA 22302 (EF)

LING, LEE (Mr) 1608 Belvoir Dr., Los Altos, CA 94022 (EF)

LINK, CONRAD B. (Dr) 6812 Pineway St., Hyattsville, MD 20782 (F)

LIST, ROBERT J. (Mr) 1123 Francis Hammond Parkway, Alexandria, VA 22302 (EF)

LOCKARD, J. DAVID (Dr) Botany Department, University of Maryland, College Park, MD 20742 (F)

LOEBENSTEIN, W. V. (Dr) 8501 Sundale Dr., Silver Spring, MD 20910 (LF)

LONG, BETTY JANE (Mrs) 416 Riverbend Rd., Ft. Washington, MD 20744 (F)

LORING, BLAKE M. (Dr) 26889 Lancia St., Moreno Valley, CA 92388-4843 (EF)

LUSTIG, ERNEST (Dr) Rossittenweg 10, D-3340 Wolfenbuttel, Federal Republic of Germany, (F)

LYNN, JEFFREY W. (Prof) 1902 Alabaster Dr., Silver Spring, MD 20904 (F)

LYONS, JOHN W. (Dr) 7430 Woodville Rd., Mt. Airy, MD 21771 (F)

MACDONELL, MICHAEL T. (Dr) 3939 Ruffin Rd., San Diego, CA 92123 (F)

MADDEN, JEREMIAH P. (Mr) Goddard Space Flight Center, Code 403, Greenbelt, MD 20771 (F)

MADDEN, ROBERT P. (Dr) National Institute of Standards and Technology, A 251, Physics Bldg.,

Gaithersburg, MD 20899 (F)

MAIENTHAL, MILLARD (Dr) 10116 Bevern Lane, Potomac, MD 20854 (F)

MALONE, THOMAS B. (Dr) 6633 Kennedy Lane, Falls Church, VA 22042 (F)

MANDERSCHEID, RONALD W. (Dr) 10837 Admirals Way, Potomac, MD 20854-1232 (LF)

MARCUS, MARVIN (Dr) 2937 Kenmore Place, Santa Barbara, CA 93105 (F)

MARTIN, EDWARD J., P. E. (Dr) 7721 Dew Wood Dr., Derwood, MD 20855 (F)

MARTIN, JOHN H. (Dr) 440 N.W. Elks Dr., Apt. 205, Corvallis, OR 97330-3749 (EF)

MARTIN, ROBERT H. (Mr) 2257 N. Nottingham St., Arlington, VA 22205 (EM)

MARTIN, ROY E. (Mr) National Fisheries Institute, Suite 580, 2000 M St., NW, Washington, DC

20036 (F)

MASON, HENRY LEA (Dr) 7008 Meadow Lane, Chevy Chase, MD 20815 (F)

MATLACK, MARION B. (Dr) 4318 N. Pershing Dr., Apt. 2, Arlington, VA 22203 (EF)

MAYOR, JOHN R. (Dr) 3308 Solomons Court, Silver Spring, MD 20906 (F)

McAVOY, THOMAS J. (Dr) Chem-Nuclear Engineering Department, University of Maryland, College

Park, MD 20742 (F)

McBRIDE, GORDON W. (Mr) 3323 Stuyvesant Place, NW, Washington, DC 20015-2454 (EF)

McCARRICK, ANNE K. (Dr) 2939 Blue Spruce Circle, Thousand Oaks, CA 91360 (F)

McCONNELL, DUDLEY G. (Dr) 926 Clintwood Dr., Silver Spring, MD 20902 (F)

McCRACKEN, ROBERT H. (Mr) 5120 Newport Ave., Bethesda, MD 20816 (LF)

McKENZIE, LAWSON M. (Mr) 1719 N. Troy St., Apt. 394, Arlington, VA 22201 (F)

McKINSTRY, PATRICIA A. (Ms) 11671 Gilman Lane, Herndon, VA 22070-2420 (M)

McNESBY, JAMES R. (Dr) 13308 Valley Dr., Rockville, MD 20850 (EF)

MEADE, BUFORD K. (Mr) 5903 Mt. Eagle Dr., Apt. 404, Alexandria, VA 22303-2523 (F)

MEARS, FLORENCE M. (Dr) 8004 Hampden Lane, Bethesda, MD 20814 (EF)

MEARS, THOMAS W. (Mr) 2809 Hathaway Terrace, Wheaton, MD 20906 (F)

MEBS, RUSSELL W. (Dr) 6620 N. 32nd St., Arlington, VA 22213 (F)

MELLINGER, JOHN J. (Dr) 7531 Woodberry Lane, Falls Church, VA 22042 (M)

MELMED, ALLEN J. (Dr) 732 Tiffany Court, Gaithersburg, MD 20878 (F)

MENZER, ROBERT E. (Dr) 612 Silverthorn Rd., Gulf Breeze, FL 32561-4626 (F)

MESSINA, CARLA G. (Mrs) 9800 Marquette Dr., Bethesda, MD 20817 (F)

180 19909 MEMBERSHIP DIRECTORY

MEYERSON, MELVIN R. (Dr) 611 Goldsborough Dr., Rockville, MD 20850 (F)

MILLAR, DAVID B. (Dr) 1716 Mark Lane, Rockville, MD 20852 (F)

MILLER, CARL F. (Dr) P.O. Box 127, Gretna, VA 24557 (EF)

MITTLEMAN, DON (Dr) 80 Parkwood Lane, Oberlin, OH 44074 (F)

MIZELL, LOUIS R. (Mr) 8122 Misty Oaks Blvd., Sarasota, FL 34243 (EF)

MOORE, GEORGE A. (Dr) 1108 Agnew Dr., Rockville, MD 20851-1601 (EF)

MOORE, JAMES G. (Mr) CRS, Library of Congress, Washington, DC 20540 (M)

MORGAN, HARRY D. (Dr) 410 Taylor St., NE, Apt B 414, Washington, DC 20017 (F)

MORRIS, ALAN (Dr) 5817 Plainview Rd., Bethesda, MD 20817 (F)

MORRIS, J. ANTHONY (Dr) 23-E Ridge Rd., Greenbelt, MD 20770 (M)

MORRIS, JOSEPH BURTON (Mr) Chemistry Department, Howard University, Washington, DC 20059

(F)

MORRIS, MARLENE C. (Mrs) 6001 Eighth St., NW, Washington, DC 20011 (F)

MORRISS, DONALD J. (Mr) 102 Baldwin Court, SE, Point Charlotte, FL 33952 (EF)

MOSTOFI, F. K. (M.D.) Armed Forces Institute of Pathology, Washington, DC 20306 (F)

MOUNTAIN, RAYMOND D. (Dr) 5 Monument Court, Rockville, MD 20850 (F)

MUEHLHAUSE, C. O. (Dr) 112 Accomac St., Chincoteague, VA 23336-1401 (EF)

MUESEBECK, CARL F. W. (Mr) 18 N. Main St., Elba, NY 14058 (EF)

MULLIGAN, JAMES H., JR. (Dr) 12121 Sky Lane, Santa Ana, CA 92705 (F)

MUMMA, MICHAEL J., (Dr) 210 Glen Oban Dr., Arnold, MD 21012 (F)

MURDAY, JAMES S. (Dr) 7116 Red Horse Tavern Lane, West Springfield, VA 22153 (F)

MURDOCH, WALLACE P. (Dr) 2264 Emmitsburg Rd., Gettysburg, PA 17325 (EF)

MURRAY, T. H. (Dr) (LtC. Ret) 2915 N. 27th St., Arlington, VA 22207 (M)

MURRAY, WILLIAM S. (Dr) 1281 Bartonshire Way, Rockville, MD 20854 (F)

NAESER, CHARLES R. (Dr) 6654 Van Winkle Dr., Falls Church, VA 22044 (EF)

NAMIAS, JEROME (Mr) Scripps Institute of Oceanography, University of California, Room A 024, La

Jolla, CA 92093 (NRF)

NEF, EVELYN S. (Mrs) 2726 N St., NW, Washington, DC 20007 (M)

NELSON, R. H. (Mr) Bethany Village, 512 Albright Dr., Mechanicsburg, PA 17055 (EF)

NEUBAUER, WERNER G. (Dr) 4603 Quarter Charge Dr., Annandale, VA 22003 (F)

NEUENDORFFER, J. A. (Dr) 911 Allison St., Alexandria, VA 22302 (EF)

NEUPERT, WERNER M. (Dr) Goddard Space Flight Center, Code 680, Greenbelt, MD 20771 (F)

NEUSCHEL, SHERMAN K. (Dr) 7501 Democracy Blvd., Bethesda, MD 20817 (F)

NEWMAN, MORRIS (Dr) 1050 Las Alturas Rd., Santa Barbara, CA 93103 (F)

NICKUM, MARY J. (Mrs) 12174 Island View Circle, Germantown, MD 20874 (M)

NOFFSINGER, TERRELL L. (Dr) Route 1, Box 305, Auburn, KY 42206 (EF)

NORWOOD, JANET L. (Dr) Bureau of Labor Statistics, 200 Constitution Ave., NW, Washington, DC

20214 (F)

OBERLE, E. MARILYN (Ms) 2801 Quebec St., NW, Apt. 622, Washington, DC 20008 (M)

OEHSER, PAUL H. (Mr) Regency at McLean, 1800 Old Meadow Rd., McLean, VA 22102 (EF)

O’HARE, JOHN J. (Dr) 4601 O’Connor Court, Irving, TX 75062 (F)

O’HERN, ELIZABETH M. (Dr) 633 G St., SW, Washington, DC 20024 (F)

OKABE, HIDEO (Dr) 6700 Old Stage Rd., Rockville, MD 20852 (F)

O’KEEFE, JOHN A. (Dr) Goddard Space Flight Center, Code 681, Greenbelt, MD 20771 (F)

OLIPHANT, MALCOLM W. (Dr) 1606 Ulupii St., Kailua, HI 96734 (EF)

OLIPHANT, V. SUSIE (Ms) 910 Luray Place, Hyattsville, MD 20783 (M)

ORDWAY, FRED (Dr) 5205 Elsmere Ave., Bethesda, MD 20814 (F)

OSER, HANS J. (Dr) 8810 Quiet Stream Court, Potomac, MD 20854 (F)

OSTAFF, WILLIAM ALLEN (Mr) 10208 Drumm Ave., Kensington, MD 20895 (M)

1990 MEMBERSHIP DIRECTORY 181

PANCELLA, JOHN R. (Dr) 1209 Viers Mill Rd., Rockville, MD 20851 (F)

PARASURAMAN, RAJA (Dr) 1852 Ingleside Terrace, NW, Washington, DC 20010 (F)

PARMAN, GEORGE K. (Mr) 4255 Donald St., Eugene, OR 97405-3427 (F)

PARRY-HILL, JEAN (Ms) 3803 Military Rd., NW, Washington, DC 20015 (M)

PARSONS, H. McILVANE (Dr) Human Resources Research Organization, 1100 S. Washington St.,

Alexandria, VA 22314 (F)

PAZ, ELVIRA L. (Dr) 172 Cook Hill Rd., Wallingford, CT 06492 (F)

PELCZAR, MICHAEL J. (Dr) Avalon Farm, P.O. Box 133, Chester, MD 21619 (EF)

PELLERIN, CHARLES J. (Dr) National Aeronautics and Space Administration, Code EZ-7, 600 In-

dependence Ave., SW, Washington, DC 20546 (F)

PERKINS, LOUIS R. (Mr) 1234 Massachusetts Ave., NW, Apt. 709, Washington, DC 20005 (M)

PERROS, THEODORE P. (Dr) Chemistry Department, George Washington University, Washington,

DC 20052 (F)

PICKETT, WARREN E. (Dr) 8406 Echo Lane, Clinton, MD 20735 (F)

PICKHOLZ, RAYMOND (Dr) 3613 Glenbrook Rd., Fairfax, VA 22031 (F)

PIEPER, GEORGE F. (Dr) 3155 Rolling Rd., Edgewater, MD 21037 (F)

PIERCHALA, CARL E. (Dr) 2400 16th St., NW, Apt. 537, Washington, DC 20009-6629 (M)

PIKL, JOSEF M. (Dr) Meadowbrook Rd., Lincoln, MA 01773 (EF)

PITTMAN, MARGARET (Dr) 3133 Connecticut Ave., NW, Apt. 912, Washington, DC 20008 (EF)

PLAIT, ALAN O. (Mr) 5402 Yorkshire St., Springfield, VA 22151 (F)

POLACHEK, HARRY (Dr) 11801 Rockville Pike, Apt. 1211, Rockville, MD 20852 (EF)

PONNAMPERUMA, CYRIL (Dr) Chemistry Department, University of Maryland, College Park, MD

20742 (F)

POST, MILDRED A. (Miss) 8928 Bradmoor Dr., Bethesda, MD 20817 (F)

POTTMYER, JAMES J. (Mr) 5540 N. 32nd St., Arlington, VA 22207-1535 (M)

POWELL, JAMES S. (Mr), 7873 Godolphin Dr., Springfield, VA 22153 (M)

POWERS, KENDALL G. (Dr) 6311 Alcott Rd., Bethesda, MD 20817 (F)

PRESTON, MALCOLM S. (Dr) 10 Kilkea Court, Baltimore, MD 21236 (M)

PRINCE, JULIUS S. (M.D.) 7103 Pinehurst Parkway, Chevy Chase, MD 20815 (F)

PRINZ, DIANNE K. (Dr) Naval Research Laboratory, Code 4142, Washington, DC 20375-5000 (F)

PRO, MAYNARD J. (Mr) 7904 Falstaff Rd., McLean, VA 22102 (F)

PROCTOR, JOHN H. (Mr) 308 East St., NE, Vienna, VA 22180 (F)

PRYOR, C. NICHOLAS (Dr) 3715 Prosperity Ave., Fairfax, VA 22031 (F)

PURCELL, ROBERT H. (Dr) 17517 White Grounds Rd., Boyds, MD 20841 (F)

PYKE, THOMAS N., JR. (Mr) National Oceanic and Atmospheric Administration, Code E, FB #4, Rm.

2069, Washington, DC 20057 (F)

QUIROS, RODERICK S. (Mr) 4520 Yuma St., NW, Washington, DC 20016 (F)

RABINOW, JACOB (Mr) 6920 Selkirk Dr., Bethesda, MD 20817 (F)

RADER, CHARLES A. (Mr) Gillette Research Institute, 401 Professional Dr., Gaithersburg, MD 20879

(F)

RADO, GEORGE T. (Dr) 818 Carrie Court, McLean, VA 22101 (F)

RAINWATER, IVAN H. (Dr) 2805 Liberty Place, Bowie, MD 20715 (EF)

RAMSAY, MAYNARD J. (Dr) 3806 Viser Court, Bowie, MD 20715 (F)

RANSOM, JAMES R. (Mr) 107 E. Susquehanna Ave., Towson, MD 21204 (M)

RASKIN, ALLEN (Dr) 7658 Water Oak Point Rd., Pasadena, MD 21122 (F)

RATH, BHAKTA B. (Dr) 10908 Timbermill Court, Oakton, VA 22124 (F)

RAUSCH, ROBERT L. (Dr) P.O. Box 85447, University Station, Seattle, WA 98145-1447 (NRF)

RAVECHE, ELIZABETH S. (Dr) 27 24th St., Troy, NY 12180-1914 (F)

RAVITSKY, CHARLES (Mr) 1505 Drexel St., Takoma Park, MD 20912 (EF)

182 1990 MEMBERSHIP DIRECTORY

RAY, JOSEPH W. (Dr) 2740 Vassar Place, Columbus, OH 43221 (F)

REDISH, EDWARD F. (Prof) 6820 Winterberry Lane, Bethesda, MD 20817 (F)

REED, WILLIAM DOYLE (Mr) Thomas House, 1330 Massachusetts Ave., NW, Apt. 624, Washington,

DC 20005 (EF)

REHDER, HARALD H. (Dr) 3900 Watson Place, NW, Washington, DC 20016 (F)

REINER, ALVIN (Mr) 11243 Bybee St., Silver Spring, MD 20902 (F)

REINHART, FRANK W. (Dr) 9918 Sutherland Rd., Silver Spring, MD 20901 (F)

REMMERS, GENE M. (Mr) 6928 Hector Rd., McLean, VA 22101 (M)

RESWICK, JAMES S. (Dr) 1003 Dead Run Dr., McLean, VA 22101 (F)

REYNOLDS, HORACE N., JR. (Dr) 9223 Woodland Dr., Silver Spring, MD 20910 (F)

RHYNE, JAMES J. (Dr) 14521 Pebble Hill Lane, Gaithersburg, MD 20878 (F)

RICE, ROBERT L. (Mr) 15504 Fellowship Way, Gaithersburg, MD 20878 (M)

RICE, SUE ANN (Ms) 6728 Fern Lane, Annandale, VA 22003 (M)

RICHMOND, ANNE T. (Mrs) 8833 Cold Spring Rd., Potomac, MD 20854 (F)

RIEL, GORDON K. (Dr) Naval Surface Weapons Center, White Oak Laboratory, Code R-41, Silver

Spring, MD 20903-5000 (LF)

RITT, PAUL E. (Dr) 36 Sylvan Lane, Weston, MA 02193 (F)

RIVERA, ALVIN D. (Dr) 4302 Star Lane, Rockville, MD 20852 (M)

ROBBINS, MARY LOUISE (Dr) Tatsuno House, A-23, 2-1-8 Ogikubo, Suginami-Ku, Tokyo 167, Japan

(EF)

ROBERTSON, A. F. (Dr) 4228 Butterworth Pl., NW, Washington, DC 20016 (EF)

ROBERTSON, RANDALL M. (Dr) 1404 Highland Circle, S.E., Blacksburg, VA 24060 (EF)

ROBSON, CLAYTON W. (Mr) 13307 Warburton Dr., Ft. Washington, MD 20744 (M)

RODNEY, WILLIAM S. (Dr) 10707 Muirfield Dr., Potomac, MD 20854 (F)

ROE, DONALD W. (Dr) 17316 Chiswell Rd., Poolesville, MD 20837 (M)

ROLLER, PAUL S. (Dr) 4201 Butterworth Place, NW, Apt. 451, Washington, DC 20016 (EF)

ROSADO, JOHN A. (Mr) 8821 Cardinal Court, Laurel, MD 20707 (LF)

ROSCHER, NINA M. (Dr) 10400 Hunter Ridge Dr., Oakton, VA 22124 (F)

ROSE, WILLIAM K. (Dr) Astronomy Program, University of Maryland, College Park, MD 20742 (F)

ROSENBLATT, DAVID (Prof) 2939 Van Ness St., NW, Apt. 702, Washington, DC 20008 (F)

ROSENBLATT, JOAN R. (Dr) 2939 Van Ness St., NW, Apt. 702, Washington, DC 20008 (F)

ROSENFELD, AZRIEL (Dr) 847 Loxford Terrace, Silver Spring, MD 20910 (F)

ROSENTHAL, SANFORD M. (Dr) 12601 Greenbrier Rd., Potomac, MD 20854 (EF)

ROSS, FRANKLIN J. (Mr) 3830 N. Stafford St., Arlington, VA 22207-4513 (F)

ROSS, SHERMAN (Dr) 19715 Greenside Terrace, Gaithersburg, MD 20879 (EF)

ROSSI, PETER H. (Prof) 34 Stage Coach Rd., Amherst, MA 01003 (NRF)

ROSSINI, FREDERICK D. (Dr) 605 South U.S. Highway #1, Apt. T-900, Juno Beach, FL 33408 (EF)

ROTHMAN, RICHARD B. (Dr) 1510 Flora Court, Silver Spring, MD 20910 (F)

ROTKIN, ISRAEL (Mr) 11504 Regnid Dr., Wheaton, MD 20902 (EF)

RUTNER, EMILE (Dr) 34 Columbia Ave., Takoma Park, MD 20912 (M)

SAENZ, ALBERT W. (Dr) Naval Research Laboratory, Code 6603-S, Washington, DC 20375-5000 (F)

SALISBURY, LLOYD L. (Mr) 10138 Crestwood Rd., Kensington, MD 20895 (M)

SALLET, DIRSE W. (Dr) 4205 Tuckerman St., University Park, MD 20782-2144 (M)

SAMUELSON, DOUGLAS A. (Mr) 1910 Wintergreen Court, Reston, VA 22091 (F)

SANDERSON, JOHN A. (Dr) B-206 Clemson Downs, 150 Downs Blvd., Clemson, SC 29631 (EF)

SANK, VICTOR J. (Dr) 5 Bunker Court, Rockville, MD 20854 (F)

SARMIENTO, RAFAEL A. (Dr) Bldg. 306, Rm. 101, BARC-East, Beltsville, MD 20705 (F)

SASMOR, ROBERT M. (Dr) 4408 N. 20th Rd., Arlington, VA 22207 (F)

SASS, ARTHUR H., USNR (Capt) RFD 6, Box 176, Warrenton, VA 22186 (M)

SAVILLE, THORNDIKE, JR. (Mr) 5601 Albia Rd., Bethesda, MD 20816 (LF)

19909 MEMBERSHIP DIRECTORY 183

SCHALK, JAMES M. (Dr) P.O. Box 441, Isle of Palms, SC 29451 (NRF)

SCHECHTER, MILTON S. (Mr) 10909 Hannes Court, Silver Spring, MD 20901 (F)

SCHINDLER, ALBERT I. (Dr) 6615 Sulky Lane, Rockville, MD 20852 (F)

SCHLAIN, DAVID (Dr) 2-A Gardenway, Greenbelt, MD 20770 (EF)

SCHMIDT, CLAUDE H. (Dr) 1827 Third St. North, Fargo, ND 58102 (EF)

SCHNEIDER, JEFFREY M. (Dr) 5238 Richardson Dr., Fairfax, VA 22032 (F)

SCHNEIDER, SIDNEY (Mr) 239 N. Granada St., Arlington, VA 22203 (EM)

SCHNEPFE, MARIAN M. (Dr) Potomac Towers, 2001 N. Adams St., Apt. 640, Arlington, VA 22201

(EF)

SCHOOLEY, JAMES F. (Dr) 13700 Darnestown Rd., Gaithersburg, MD 20878 (EF)

SCHUBAUER, GALEN B. (Dr) 10450 Lottsford Rd., Unit 1211, Mitchellville, MD 20716 (F)

SCHULMAN, FRED (Dr) 11115 Markwood Dr., Silver Spring, MD 20902 (F)

SCHULMAN, JAMES H. (Dr) 4615 North Park Ave., Apt. 1519, Chevy Chase, MD 20815 (EF)

SCHULTZ, WARREN W. (Dr) 4056 Cadle Creek Rd., Edgewater, MD 21037 (LF)

SCOTT, DAVID B. (Dr) 10448 Wheatridge Dr., Sun City, AZ 85373 (EF)

SCRIBNER, BOURDON F. (Mr) 123 Peppercorn Place, Edgewater, MD 21037 (EF)

SEABORG, GLENN T. (Dr) 1154 Glen Rd., Lafayette, CA 94549 (NRF)

SEEGER, RAYMOND J. (Dr) 4507 Wetherill Rd., Bethesda, MD 20816 (EF)

SEITZ, FREDERICK (Dr) Rockefeller University, 1230 York Ave., New York, NY 10021 (NRF)

SHAFRIN, ELAINE G. (Mrs) 800 Fourth St., SW, Apt. N-702, Washington, DC 20024 (F)

SHAPIRO, GUSTAVE (Mr) 3704 Munsey St., Silver Spring, MD 20906 (F)

SHEAR, RALPH E. (Mr) 1916 Bayberry Rd., Edgewood, MD, 21040 (M)

SHEPARD, HAROLD H. (Dr) 2701 S. June St., Arlington, VA 22202 (EF)

SHERESHEFSKY, J. LEON (Dr) 4530 Connecticut Ave., NW, Apt. 400, Washington, DC 20008 (EF)

SHERLIN, GROVER C. (Mr) 4024 Hamilton St., Hyattsville, MD 20781 (LF)

SHIER, DOUGLAS R. (Dr) 515 Spring Trace, Williamsburg, VA 23185 (NRF)

SHOTLAND, EDWIN (Dr) 418 E. Indian Spring Dr., Silver Spring, MD 20901 (M)

SHRIER, STEFAN (Dr) P.O. Box 19139, Alexandria, VA 22320-0139 (F)

SHROPSHIRE, W., JR. (Rev) (Dr) 2 Castlehill Court, Timonium, MD 21093-1930 (LF)

SILVER, DAVID M. (Dr) Applied Physics Laboratory, JHU, 11100 Johns Hopkins Rd., Laurel, MD

20707 (M)

SILVERMAN, BARRY G. (Dr) 9653 Reach Rd., Potomac, MD 20854 (F)

SIMHA, ROBERT (Dr) Department of Macromolecular Science, Case-Western Reserve University,

Cleveland, OH 44106 (NRF)

SIMPSON, MICHAEL M. (Mr) Congressional Research Service SPR, LM413, Washington, DC 20540

(LM)

SLACK, LEWIS (Dr) 27 Meadow Bank Rd., Old Greenwich, CT 06870 (EF)

SLAWSKY, MILTON M. (Dr) 8803 Lanier Dr., Silver Spring, MD 20910 (EF)

SLAWSKY, ZAKA I. (Dr) 4701 Willard Ave., Apt. 318, Chevy Chase, MD 20815 (EF)

SMITH, BLANCHARD D., JR. (Mr) 2509 Ryegate Lane, Alexandria, VA 22308 (F)

SMITH, MARCIA S. (Ms) 6015 N. Ninth St., Arlington, VA 22205 (LM)

SMITH, REGINALD C. (Mr) 7731 Tauxemont Rd., Alexandria, VA 22308 (M)

SMITH, ROBERT C., JR. (Mr) 9000 Little Stone Dr., Ft. Washington, MD 20744 (F)

SNYDER, HERBERT N. (Dr) P.O. Box 1494, Tappahannock, VA 22560 (NRF)

SOLAND, RICHARD M. (Dr) SEAS, George Washington University, Washington, DC 20052 (LF)

SOLOMON, EDWIN M. (Mr) 15107 Interlachen Dr., Apt. 521, Silver Spring, MD 20906 (M)

SOMMER, HELMUT (Mr) 9502 Hollins Court, Bethesda, MD 20817 (EF)

SORROWS, HOWARD EARLE (Dr) 8820 Maxwell Dr., Potomac, MD 20854 (F)

SOUSA, ROBERT J. (Dr) 56 Wendell Rd., Shutesbury, MA 01072 (NRF)

SPATES, JAMES E. (Mr) 8609 Irvington Ave., Bethesda, MD 20817 (LF)

SPECHT, HEINZ (Dr) Fairhaven, C 135, 7200 3rd Ave., Sykesville, MD 21784 (EF)

184 19909 MEMBERSHIP DIRECTORY

SPENCER, LEWIS V. (Dr) P.O. Box 87, Hopkinsville, KY 42240 (NRF)

SPERLING, FREDERICK (Dr) 1110 Fidler Lane, Silver Spring, MD 20910 (EF)

SPIES, JOSEPH R. (Dr) 507 N. Monroe St., Arlington, VA 22201 (EF)

SPILHAUS, A. F., JR. (Dr) 10900 Picasso Lane, Potomac, MD 20854 (F)

SPRAGUE, G. F. (Dr) 2212 South Lynn St., Urbana, IL, 61801 (EF)

SPROULL, JAMES D. (Mr) 416 Blair Rd., Vienna, VA 22180 (F)

STAUSS, HENRY E. (Dr) 8005 Washington Ave., Alexandria, VA 22308 (F)

STEERE, RUSSELL L. (Dr) 6207 Carrollton Terrace, Hyattsville, MD 20781 (EF)

STEGUN, IRENE A. (Miss) 62 Leighton Ave., Yonkers, NY 10705 (EF)

STEINBERG, ALFRED D. (M.D.) 8814 Bells Mill Rd., Potomac, MD 20854 (F)

STEINER, ROBERT F. (Dr) 2609 Turf Valley Rd., Ellicott City, MD 21043 (F)

STEPHENS, ROBERT E. (Dr) 4301 39th St., NW, Washington, DC 20016 (EF)

STERN, KURT H. (Dr) Naval Research Laboratory, Code 6170, Washington, DC 20375 (F)

STEWART, T. DALE (Dr) 1191 Crest Lane, McLean, VA 22101 (EF)

STIEF, LOUIS J. (Dr) Goddard Space Flight Center, Code 691, Greenbelt, MD 20771 (F)

STIEHLER, ROBERT D. (Dr) 3234 Quesada St., NW, Washington, DC 20015-1663 (EF)

STILL, JOSEPH W. (Dr) 1408 Edgecliff Lane, Pasadena, CA 91107 (EF)

STOETZEL, MANYA B. (Dr) Systematic Entomology Laboratory, Rm. 6, Bldg. 004, BARC WEST,

Beltsville, MD 20705 (F)

STRAUSS, SIMON W. (Dr) 4506 Cedell Place, Camp Springs, MD 20748 (LF)

STRIMPLE, HARRELL L. (Mr) 904 Bowery, Iowa City, [A 52240 (EF)

SVOBODA, JAMES A. (Mr) 13301 Overbrook Lane, Bowie, MD 20715 (M)

SWEZEY, ROBERT W. (Dr) 142 Clarks Ridge Rd., Route 3, Box 142, Leesburg, VA 22075 (F)

SYKES, ALAN O. (Dr) 304 Mashie Dr., SE, Vienna, VA 22180 (M)

TALBERT, PRESTON T. (Dr) Chemistry Department, Howard University, Washington, DC 20059 (F)

TASAKI, ICHIJI (Dr) 5604 Alta Vista Rd., Bethesda, MD 20817 (F)

TATE, DOUGLAS R. (Mr) 11415 Farmland Dr., Rockville, MD 20852 (F)

TAYLOR, BARRY N. (Dr) 11908 Tallwood Court, Potomac, MD 20854 (F)

TAYLOR, JOHN KEENAN (Dr) 12816 Tern Dr., Gaithersburg, MD 20878 (F)

TAYLOR, LAURISTON S. (Dr) 7407 Denton Rd., Bethesda, MD 20814 (EF)

TAYLOR, WILLIAM B., P. E. (Mr) 4001 Belle Rive Terrace, Alexandria, VA 22309 (M)

TEAL, GORDON K. (Dr) 5222 Park Lane, Dallas, TX, 75220 (NRF)

TERMAN, MAURICE J. (Mr) 616 Poplar Dr., Falls Church, VA 22046 (EM)

THOMPSON, F. CHRISTIAN (Dr) 4255 S. 35th St., Arlington, VA 22206 (LF)

TOLL, JOHN S. (Dr) Universities Research Association, 1111 19th St., NW, Washington, DC, 20036 (F)

TOUSEY, RICHARD (Dr) 7725 Oxon Hill Rd., Oxon Hill, MD 20745 (EF)

TOUSIMIS, A. J. (Dr) Tousimis Research Corp., P.O. Box 2189, Rockville, MD 20847-2189 (M)

TOWNSEND, CHARLES E. (M.D.) 3529 Tilden St., NW, Washington, DC 20008-3194 (F)

TOWNSEND, LEWIS RHODES (M.D.) 8906 Liberty Lane, Potomac, MD 20854 (M)

TOWNSEND, MARJORIE R. (Mrs) 3529 Tilden St., NW, Washington, DC 20008-3194 (LF)

TRAUB, ROBERT (Col. Ret.) 5702 Bradley Blvd., Bethesda, MD 20814 (EF)

TUNNELL, GEORGE (Dr) Casa Dorinda 124, 300 Hot Springs Rd., Montecito, Santa Barbara, CA

93108 (EF)

TURNER, JAMES H. (Dr) 509 South Pinehurst Ave., Salisbury, MD 21801-6122 (EF)

TYLER, PAUL E. (M.D.) 12604 Stablehouse Court North, Potomac, MD 20854 (F)

UBERALL, HERBERT M. (Dr) Kenwood, 5101 River Rd., Apt. 1417, Bethesda, MD 20816 (F)

UHLANER, J. E. (Dr) 4258 Bonavita Dr., Encino, CA 91426 (EF)

USDIN, VERA R. (Dr) 6 Stevens Court, Rockville, MD 20850 (F)

19909 MEMBERSHIP DIRECTORY 185

VAISHNAV, MARIANNE P. (Ms) P.O. Box 2129, Gaithersburg, MD 20879 (LF)

VAN COTT, HAROLD P. (Dr) 8300 Still Spring Court, Bethesda, MD 20817 (F)

VAN DERSAL, WILLIAM R. (Dr) 8101 Greenspring Ave., Baltimore, MD 21208 (EF)

VAN TUYL, ANDREW H. (Dr) 1000 West Nolcrest Dr., Silver Spring, MD 20903 (F)

VANARSDEL, WILLIAM C., III (Dr) 1000 Sixth St., SW, Apt. 301, Washington, DC 20024 (M)

VARADI, PETER F. (Dr) 4620 North Park Ave., Apt. 1605-W, Chevy Chase, MD 20815 (F)

VEITCH, FLETCHER P., JR. (Dr) P.O. Box 513, Lexington Park, MD 20653 (F)

VILA, GEORGE J. (Mr) 5517 Westbard Ave., Bethesda, MD 20816 (F)

VINTI, JOHN P. (Dr) 44 Quint Ave., Allston, MA 02134 (NRF)

VITAS, STEPHAN THOMAS (Dr) 2803 Cortland Place, NW, Washington, DC 20008 (M)

VON HIPPEL, ARTHUR (Dr) 265 Glen Rd., Weston, MA 02193 (EF)

WAGNER, A. JAMES (Mr) 7568 Cloud Court, Springfield, VA 22153 (F)

WALDMANN, THOMAS A. (M.D.) 3910 Rickover Rd., Silver Spring, MD 20902 (F)

WALKER, DELORES H. (Mrs) 2521 Branch Ave., SE, Washington, DC 20020 (F)

WALKER, EGBERT H. (Dr) Friends House, 17330 Quaker Lane, Apt. D17, Sandy Spring, MD 20860

(EF)

WALTON, WILLIAM W., SR. (Dr) 1705 Edgewater Parkway, Silver Spring, MD 20903 (F)

WARING, JOHN A. (Dr) 1320 S. George Mason Dr., Apt. 1, Arlington, VA 22204 (M)

WARRICK, EVELYNE J. (Ms) National Color Inc., 2929 Eskridge Rd., Fairfax, VA 22031-2213 (M)

WATERWORTH, HOWARD E. (Dr) 10001 Old Franklin Ave., Seabrook, MD 20706 (F)

WATSON, ROBERT B. (Dr) 1176 Wimbledon Dr., McLean, VA 22101 (EM)

WAYNANT, RONALD W. (Dr) 13101 Claxton Dr., Laurel, MD 20708 (F)

WEBB, RALPH E. (Dr) 21-P Ridge Rd., Greenbelt, MD 20770 (F)

WEBER, ROBERT S. (Dr) 4520 Marissa Dr., El Paso, TX 79924 (EM)

WEGMAN, EDWARD J. (Dr) 10821 Burr Oak Way, Burke, VA 22015 (LF)

WETHE, WERNER K. (Dr) 2103 Bassett St., Alexandria, VA 22308 (EF)

WEINBERG, HAROLD (Mr) 11410 Strand Dr., Bldg. 1-B, Apt. 314, Rockville, MD 20852 (F)

WEINER, JOHN (Dr) 8401 Rhode Island Ave., College Park, MD 20740 (F)

WEINTRAUB, ROBERT L. (Dr) 407 Brooks Ave., Raleigh, NC 27607 (EF)

WEISS, ARMAND B. (Dr) 6516 Truman Lane, Falls Church, VA 22043 (LF)

WEISSLER, ALFRED (Dr) 5510 Uppingham St., Chevy Chase, MD 20815 (F)

WEISSLER, PEARL G. (Mrs) 5510 Uppingham St., Chevy Chase, MD 20815 (EF)

WELLMAN, FREDERICK L. (Dr) 501 E. Whitaker Mill Rd., Whitaker Glen 105-B, Raleigh, NC 27608

(EF)

WENSCH, GLEN W. (Dr) R.R. #1, Box 54, Champaign, IL 61821 (EF)

WERGIN, WILLIAM P. (Dr) 10108 Towhee Ave., Adelphi, MD 20783 (F)

WERTH, MICHAEL W. (Mr) 14 Grafton St., Chevy Chase, MD 20815 (EM)

WESTWOOD, JAMES T. (LCdr) 3156 Cantrell Lane, Fairfax, VA 22031-1989 (M)

WHITE, HOWARD J., JR. (Dr) 8028 Park Overlook Dr., Bethesda, MD 20817 (F)

WHITELOCK, LELAND D. (Mr) 2320 Brisbane St., Apt. 4, Clearwater, FL 34623 (NRF)

WHITTEN, CHARLES A. (Mr) 9606 Sutherland Rd., Silver Spring, MD 20901 (EF)

WIENER, ALFRED A. (Mr) 550 West 25th Place, Eugene, OR 97405 (NRF)

WIGGINS, PETER F. (Dr) 1016 Harbor Dr., Annapolis, MD 21403 (F)

WILHELM, PETER G. (Dr) 206 Gretna Green Court, Alexandria, VA 22304 (F)

WILMOTTE, RAYMOND M. (Dr) 2512 Que St., NW, Apt. 301, Washington, DC 20007 (LM)

WILSON, BRUCE L. (Mr) 423 Valentine St., Highland Park, NJ 08904 (EF)

WILSON, CHARLES L. (Dr) P.O. Box 1194, Shepherdstown, WV 25443 (F)

WILSON, WILLIAM K. (Mr) 1401 Kurtz Rd., McLean, VA 22101 (LF)

WISTORT, ROBERT L. (Mr) 11630 35th Place, Beltsville, MD 20705 (F)

WITTLER, RUTH G. (Dr) 2103 River Crescent Dr., Annapolis, MD 21401-7266 (EF)

186 19909 MEMBERSHIP DIRECTORY

WOLFF, EDWARD A. (Dr) 1021 Cresthaven Dr., Silver Spring, MD 20903 (F)

WOOD, LAWRENCE A. (Dr) National Institute of Standards and Technology, Room A-209, Polymers

Bldg, Gaithersburg, MD 20899 (EF)

WORKMAN, WILLIAM G. (Dr) Washington St., P.O. Box 7, Beallsville, OH 43716 (EF)

WYNN, HARVEY (Mr) 6625 Lee Highway, Arlington, VA 22205 (F)

WULF, OLIVER R. (Dr) 557 Berkeley Ave., San Marino, CA 91108 (EF)

YAPLEE, BENJAMIN S. (Mr) 8 Crestview Court, Rockville, MD 20854 (F)

YEKOVICH, FRANK S. (Dr) School of Education, Catholic University, Washington, DC 20064 (F)

YODER, HATTEN S., JR. (Dr) Geophysical Laboratory, 2801 Upton St., NW, Washington, DC 20008

(EF) |

YOUMAN, CHARLES E. (Mr) 4419 N. 18th St., Arlington, VA 22207 (M)

YOUNG, DAVID A., JR. (Dr) 612 Buck Jones Rd., Raleigh, NC 27606 (EF)

ZELENY, LAWRENCE (Dr) 4312 Van Buren St., University Park, MD 20782 (EF)

ZIEN, TSE-FOU (Dr) Naval Surface Warfare Center, White Oak Laboratory, Code R44, Silver Spring,

MD 20903-5000 (F)

ZOCH, RICHMOND T. (Mr) Route 1, Box 930, Shelby, AL 35143 (NRF)

Necrology

The following members of the Academy deceased since the last publication of the membership directory:

Dr. Arthur J. Ahearn Dr. Norman R. S. Hollies

Dr. Allen L. Alexander Dr. Woodrow C. Jacobs

Mr. Willard H. Bennett Dr. Garbis H. Keulegan

Dr. Herbert R. Bird Dr. Robert Marvin

Dr. Gerhard M. Brauer Dr. Archibald T. McPherson

Dr. F. G. Brickwedde Dr. Margaret D. Miller

Mr. Frank R. Caldwell Dr. Paul R. Miller

Mr. Stanley G. Cawelti Dr. Ralph D. Myers

Dr. Bernice E. Eddy Dr. Anthony M. Schwartz

Mr. Michael Goldberg Mr. Albert Lee Taylor

Mrs. Mary B. Harbeck

Membership Distribution

Member Category Geographic Location

N % N %

Fellow 334 48.3 Maryland 339 48.8

Emeritus Fellow 161 Pozi Virginia 133 19.2

Member 103 14.8 Other states 125 18.0

Life Fellow 44 6.3 District of Columbia 94 (eS)

Non-resident Fellow 34 4.9 Outside USA 3 0.4

Emeritus Member 13 1.9

Life Member 5 On

Totals 694 100.0 — 694 100.0

Journal of the Washington Academy of Sciences,

Volume 80, Number 4, Page 187, December 1990

Corrigenda

Corrigenda should be noted within the article: Gluckman, A. G. (1990). The

discovery of oscillatory current. Journal of the Washington Academy of Sci-

ence, 80(1), 16-25.

(a) pg. 21, Table III, line 1, change to . . .Henry’s 1838 Experimental Re-

sults...

(b) p. 24, line 4, change to . . .translated from his 19th century latinic . . .

Savary sur |’Aimanation” [““ADDITION to the Academy of Sciences in July

of 1826].

187

75 Years of Scientific Thought

The Washington Academy of Sciences, one of the oldest scientific organiza-

tions in the greater Washington, DC area, has published a book entitled “75

years of scientific thought” commemorating the first 75 years of the existence