You are on page 1of 33

Computers in Human Behavior, Vol. 7, pp. 103-135, 1991 Printed in the U.S.A. All rights reserved.

0747-5632/91 $3.00 + .00 Copyright 1991 Pergamon Press pie

Personality Factors in Human-Computer Interaction: A Review of the Literature


Kym E. Pocius
The Ohio State University

A b s t r a c t - - This is a review of studies investigating the relation between personality dimensions and human-computer interaction. The review is divided into three topic areas: programming aptitude and achievement, the programmer personality, and computer-assisted instruction (CAI). It is concluded that: (a) introversion-extraversion and traits characterizing introversion-extraversion are related to many aspects of human-computer interaction, (b) there is little evidence to support a relation between the Jungian personality dimensions and programming aptitude and achievement, (c) further, no correlations are reported between these dimensions and academic achievement in CAl-based courses; however, certain Jungian dimensions are related to attrition rate and speed of completion, (d) although little relation between Jungian dimensions and programming aptitude and achievement is noted, certain personality types select careers that utilize computer programming, (e) anxiety characterized by a threat to self-esteem facilitates or debilitates CAI performance contingent on specific aspects of the learning task, and (f) compulsive and stress-inducing personality traits, internal-external locus of control, and achievement motivation are implicated in human-computer interaction performance.

With the advent of the computer age, the impact of computer technology has seldom been met with a neutral response. Individuals' reactions toward computers range from enthusiastic acceptance to feelings of suspicion, anxiety, and even avoidance behavior. However, computers have become an indispensable aid in both the business and academic world, where they are used as tools for programming, word-processing, database management, statistical analyses, and course instruction. The proliferation of computer use has ushered in an era in which researchers are recognizing the importance of the effects of individual differences in humancomputer interaction (Martin & Fuerst, 1987; Rich, 1983; Van Muylwijk, Van Der Veer, & Waern, 1983; Van Der Veer, Tauber, Waern, & Van Muylwijk, 1985). Human-computer interaction is defined as any process in which "the user and computer engage in a communicative dialogue whose purpose is the accomplishRequests for reprints should be addressed to Kym E. Pocius, Office of Research, OCLC Online ComputerLibrary Center, Inc., 6565 Frantz Rd., Dublin, OH 43017-0702. 103

lO4

P oc ius

ment of some task" (Card, Moran, & Newell, 1983, p. 4). Human-computer interaction is mediated by both the computer system design and by the characteristics of the user. Traditionally, computer systems are designed to enhance user performance by taking advantage of the characteristics of a canonical or "typical" user, and generally to ignore individual differences (Rich, 1983). However, user characteristics influence human-computer interaction by introducing individual differences in performance. For example, ability, academic background, previous computer experience, and age are related to computer course performance (Butcher & Muth, 1985; Campbell & McCabe, 1984; Fowler & Glorfeld, 1981; Konvalina, Stephens, & Wileman, 1983; Petersen & Howe, 1979). Consequently, identifying individual differences that mediate human-computer interaction may help researchers and academicians in designing computer systems to fit a wider range of users. In addition, the ability to isolate individual differences affecting human-computer interaction may help to identify potentially productive programmers, screen students for computer science majors, help individuals make decisions about academic majors and careers, and diagnose employees who might best profit from additional computer training (Evans & Simkin, 1989). Many researchers recognize that personality plays an important role in humancomputer interaction (e.g., Sharma, 1986-1987; Van Muylwijk, Van Der Veer, & Waem, 1983; Van Der Veer, Tauber, Waem, & Van Muylwijk, 1985; Weinberg, 1971). Personality is defined as the relatively stable emotional, motivational, interpersonal, and attitudinal characteristics of the individual, and are distinguished from abilities. In this review, personality is also differentiated from "cognitive style," which is defined as the characteristic way in which the individual organizes and processes information, and which typically does not encompass motivational, emotional, or biological processes (Goldstein & Blackman, 1978). Investigating the effects of personality factors in human-computer interaction has increased in popularity over the last two decades, but currently is not focused in any particular discipline. Publications on this topic are found across a wide breadth of disciplines and journals. Personality factors not only affect human-computer interaction at the task level (e.g., programming), but may determine whether individuals will choose to use the computer to accomplish a task, and even influence whether an individual will be drawn toward a computer-related career. Most of the studies that investigate how personality characteristics of users affect human-computer interaction have been conducted in three general areas. A comprehensive review of the personality variables investigated in these three areas is presented. The first area examines how personality traits are related to programming aptitude and achievement by investigating the relation between personality measures and an individual's performance in an introductory computer programming course or on a computer aptitude test. The second area presents research that describes the "programmer personality" by examining the personality traits of those individuals who currently use programming skills in their profession. The final area reviews studies that investigate the relation between personality measures and computer-assisted instruction performance. Each review area is followed by a summary. General conclusions are presented at the end of this review.
PROGRAMMING APTITUDE AND ACHIEVEMENT

Certain individuals thrive in environments primarily focused on computer technology. Computer programmers, for example, spend a good part of their working

Personality factors in human-computer interaction

105

hours interacting with the computer. Because of the high demand for individuals with programming ability, screening tools have evolved for the purpose of identifying these individuals. Programming aptitude tests such as the IBM Aptitude Test for Programming Personnel, the Computer Programmer Aptitude Battery, and the IBM Programmer Aptitude Test have been available over the years. Although these instruments offer a means of identifying programming aptitude, these tests are not always an adequate predictor of programming ability (e.g,, Alspaugh, 1972; Mazlack, 1980), nor have been found to be related to a programmer's on-the-job performance (Morrison & Noble, 1987, p. 70). Weinberg (1971) suggests that personality is important for programming success. Several studies have examined the relation between personality factors and programming aptitude and achievement. In these investigations performance in programming courses and on computer aptitude tests are used as an indicator of programming aptitude or achievement. These studies have been most prevalent in the disciplines of computer science and education. This literature reviewed is grouped into four areas: (a) introversion-extraversion, (b) Jungian personality dimensions, (c) internal-external locus of control, and (d) compulsiveness and stress-inducing personality traits. A summary of these studies is presented in Table 1.
Introversion.Extraversion

The personality dimension most extensively studied in the human-computer interaction literature is introversion-extraversion. This construct was originally defined by Carl Jung in 1921. Jung regarded introversion-extraversion as "mutually complementary ... [attitudes whose] differences generate the tension that both the individual and society need for the maintenance of life" (1921/1971, p. 160). Research on introversion-extraversion proliferated with the introduction of the Eysenck Personality Questionnaires, followed by the Myers-Briggs Type Indicator (MBTI), instruments that directly assess this construct. Other instruments, such as the Thurstone Temperament Schedule, the California Psychological Inventory, and the Sixteen Personality Factor Questionnaire, assess this construct indirectly by measuring traits that characterize the "introverted" or "extraverted" personality. According to Jung, extraverts are oriented toward the outer world, and therefore tend to focus their perception and judgment on individuals and objects. Extraverts have an interest in events, people, and things, a relation with them, and a dependence on them. They tend to be outgoing, sociable, lively, talkative, expressive, enthusiastic, impulsive, optimistic, understandable, accessible, and have little tolerance for slow or routine tasks. Extraverts often have a wide circle of friends and are both relaxed and skilled in social situations. However, these individuals may appear superficial, too ready to adopt conventional standards, dependent on making a good impression, and disinclined to be alone. In contrast, introverts are oriented toward the inner world, and consequently focus their perception and judgment on concepts and ideas. Introverts enjoy being alone, usually prefer their own thoughts to conversation with others, and may have trouble remembering names and faces. They are quiet, reflective, introspective, reserved, questioning, subtle, and impenetrable. Introverts tend to express more interest in books than in people and have a small but close circle of friends. Given social situations, an introvert may interact clumsily, being either too outspoken or ridiculously polite. These individuals can appear overconscientious, secretive, pessimistic, and critical. Although Jung identifies an individual as either an introvert

106

Pocius

Table 1. Summary of Studies Investigating Relationship Between Personality and Programming Aptitude and Achievement
Measure Study Kagan & Douthat (1985) N 228-302 Personality EPI extraversion
i*

Programming Apt/Ach a

Results

neuroticism SMEB sensitivity to social cues HI irritability Type A Behavior Alspaugh (1972) 50 TTS impulsive sociable reflective impulsive dominant sociable impulsive sociable TTS vigorous impulsive dominant sociable active 16PF MBTI sensing introverted intuitive thinking judging introverted feeling MBTI MBTI MBTI MBTI MBTI Rotter I-E locus of control TCI Order & Regularity DSS overload

FORTRAN exam 1 FORTRAN exam 4 FORTRAN exam 3 FORTRAN exam 1 FORTRAN exam 4 FORTRAN exam 3 FORTRAN exam 3 BAL exam " FORTRAN IV exam " " Comprehensive exam " ICC course " " FORTRAN course Homework BASIC 1 Midterrn 1 " Midterm 2 " ATPP COBOL course Pascal course FORTRAN course Pascal course ICC, COBOL, ACSC

r = -.10" r = -.15"* r = -.11" r = -.13"* r = ~.19"* r = -.11" r = -.19"* r= r= r= r= r= r= r= r= r= r= r= r= r= NS F(5, F(5, F(5, F(5, F(5, F(5, F(5, NS NS NS NS NS NS r~,) = .78" rf~) .24" rf~) 1.19" rf~) .47" n,) ,66" rP) 1,01" nI) -1.57" -.38"* -.29" .30" -.38** -.29" -.31" -.42** -.34"* -.48"* -.32"* -.27** -.24** -.23"

Petersen & Howe (1979)

113

Hostetler (1983) Evans & Simkin (1989)

120 > 117

Calaway (1982) Corman (1986) Whipkey & Stephens (1984) Lee (1985) Werth (1986) Nowaczyk (1984) Kagan & Pietron (1986)

86 51 88 71 58 160, 60, 66 90

PAT "

r = -.19* r = -.26"*

Note. Apt/Ach = aptitude/achievement; EPI = Eysenck Personality Inventory; SMEB = Self-Monitoring of Expressive Behavior; HI = Hostility Inventory; TTS = Thurstone Temperament Schedule; 16PF = S i x t e e n P e r s o n a l i t y Factor Questionnaire; MBTI = M y e r s - B r i g g s Type Indicator; TCI = The Compulsiveness Inventory; DSS = Dispositional Stress Scales; BAL = Basic Assembly Language; ICC -- Introductory computer course; ATPP = IBM Aptitude Test for Programming Personnel; ACSC = Advanced computer science course; PAT = Programming Aptitude Test. aThe dependent variable is an exam, homework, course grade, or aptitude test score, bn = > 117. **p < .01. *p < .05. NS = not significant.

Personality factors in human-computer interaction

107

or an extravert, Eysenck (1964) proposes that individuals fall somewhere along a continuum ranging from introversion to extraversion. One early study investigating the relation between introversion-extraversion and grades received in an introductory FORTRAN programming course found no significant correlation (Newsted, 1975). However, categorization of a subject as an introvert or an extravert was weakly measured by a single yes/no question, namely, "How would you classify yourself on an introvert/extrovert scale?" Several subsequent studies have reported a positive relation between introversion or traits commonly attributed to the introvert and successful performance in introductory programming courses. Based on the expectation that achievement in computer programming would be strongly associated with introversion, Kagan and Douthat (1985) employed a well-designed study investigating the relation between achievement in a programming course and scores on five personality instruments measuring some aspect of introversion. Three hundred and twenty-six students enrolled in an introductory FORTRAN course completed: (a) Eysenck's Personality Inventory (Eysenck, 1964), (b) the Crowne-Marlowe Social Desirability Scale (Crowne & Marlowe, 1960), (c) the Self-Monitoring of Expressive Behavior (Snyder, 1974), (d) the Hostility Inventory (Buss & Durkee, 1957), and (e) Type A Behavior (Bormer, 1969). Based on previous studies linking achievement in math, science, and engineering to a distinct cluster of personality traits characterizing introversion, these authors hypothesized that low neuroticism, need for social approval, sensitivity toward social cues, hostility, and Type A personality would be positively related to introversion. Eysenck's Personality Inventory measures extroversion-introversion and neuroticism. Neuroticism is characterized as a tendency to be moody, changeable, and anxious. The Crowne-Marlowe Social Desirability Scale measures a respondent's need for social approval. The Self-Monitoring of Expressive Behavior measures sensitivity to the cues in social situations, self-awareness of one's own behavior, and concern for its appropriateness. Two scales of the Hostility Inventory were used in this study: Irritability and Verbal Hostility. Finally, the Type A Behavior instrument measures the overt behavior pattern associated with coronary heart disease. Type A individuals are defined as being more competitive, hard-driving, ambitious, and time-conscious than their slower-moving, easy-going Type B counterparts. Although 91 students eventually withdrew from the course, these students did not differ significantly in terms of personality scores from those students who completed the course. Results were determined by bivariate and canonical correlational analyses and multiple regression analyses. Initial differences in intelligence and computer experience were controlled by partialing out the variance attributed to Exam 1, which was used as a measure of initial aptitude. Low scores on Extraversion and Self-Monitoring of Expressive Behavior were positively related to achievement on both the first and last exams. In other words, an introverted attitude and an insensitivity to the nuances of social situations were compatible with success in both the first and last quarter of the course. None of the personality variables correlated significantly with Exam 2 scores. However, high scores on Exam 3 were negatively associated with Neuroticism, Irritability, and Type A behavior, suggesting that a relatively relaxed temperament was conducive to attaining the programming skills necessary in the third quarter of the course. Although multiple regression analyses failed to yield significant equations for Exams 2 and 3, all of the personality dimensions contributed significantly to the

108

P ocius

equation for Exam 4. Low scores on the Self-Monitoring of Expressive Behavior scale and on Verbal Hostility and high scores on Need for Social Approval contributed the greatest variance toward programming achievement. These authors concluded that as the course progressed, temperament became more - - rather than less - - relative to programming achievement. Canonical correlation analysis produced two significant functions. Function one indicated that achievement on exam 4 was associated with a tendency to be ambitious, hard-driving (Type A) and socially insensitive. Interestingly, Type A behavior was associated with low achievement earlier in the course. A statistical comparison of exam means indicates that the course became progressively more difficult, suggesting that Type A individuals excelled in situations that required greater challenge. Function two indicated that introversion and emotional stability were relevant to success early in the course. Although these authors propose that all the personality traits used in this study represent a single global dimension of introversion-extraversion, intercorrelations among these measures indicate that this assumption is only partially supported. Introversion was negatively related to Type A personality and Self-Monitoring of Expressive Behavior. However, contrary to assumptions, introversion was positively related to Irritability and Neuroticism, and exhibited no relation to need for social approval. This study suggests that introversion as measured by the Eysenck Personality Inventory and certain traits characterizing introversion such as an insensitivity to social cues are relevant to success in both the early and latter parts of the programming course. In addition, a relaxed attitude early in the course followed by an ambitious, hard-driving (Type A behavior) approach toward the end of the course is conducive to programming success. Corman (1986) reports that Type A behavior is not significantly related to performance in a COBOL class. Two studies illustrate the relation between introverted characteristics and successful p e r f o r m a n c e in a p r o g r a m m i n g c o u r s e by using the T h u r s t o n e Temperament Schedule to identify personality traits associated with programming aptitude. The Thurstone Temperament Schedule employs seven measures of personality: active, vigorous, impulsive, dominant, stable, sociable, and reflective. Alspaugh (1972) administered this instrument to 50 students enrolled in an introductory computer science course. Students spent the first half of the course studying Basic Assembly Language for the IBM 360 computer. The second half of the course was spent studying FORTRAN IV. Programming students with the highest comprehensive grades for both Basic Assembly Language and FORTRAN IV possessed a personality associated with a low level of impulsiveness and sociability. In addition, students with the highest grades in the Basic Assembly Language portion of the course also scored relatively high on reflectiveness. In a second study Petersen and Howe (1979) collected Thurstone Temperament Schedule data from 119 fall semester and 113 spring semester students enrolled in an Introduction to Computers course. Although Thurstone Temperament Schedule scores between the fall and spring semester students did not significantly differ, several of these variables for the spring semester students were significantly correlated with their performance for that semester: vigorous, impulsive, dominant, sociable, and active. These results are congruent with those findings of Alspaugh individuals who score low on these characteristics tend to do well in a computer course. The Thurstone Temperament Schedule personality traits identified in both of these studies (i.e., low impulsiveness and sociability and high reflectiveness) characterize the introverted personality. Moreover, the results of these studies sup-

Personality factors in human-computer interaction

109

port Kagan and Douthat's findings that an introverted personality style is positively related to computer aptitude and achievement. In one exception to these findings, Hostetler (1983) found no relation between personality factors and performance in an introductory computer science business course that covered the basic concepts of structured programming using Fortran. The Sixteen Personality Factor Questionnaire, which measures 16 dichotomous personality traits, was administered to 120 students enrolled in the course. Although extraversion is identified as a second-order trait (Cattell, Eber, & Tatsuoka, 1970), none of the variables was significantly related to successful performance in the course, thereby failing to support the previous findings. In summary, there is some evidence to suggest that introversion is related to programming aptitude and achievement. Introverts and individuals who are reserved, insensitive to social nuances, and have a low level of sociability and impulsivity, traits that characterize the introvert, perform better in computer programming courses (Alspaugh, 1972; Kagan & Douthat, 1985; Petersen & Howe, 1979). In the following section, the literature that has explored the relation between Jungian personality dimensions and programming aptitude and acheivment is presented. Jung (1921/1971) identifies three bipolar personality dimensions: introversion-extraversion (I-E), sensing-intuitive (S-N), and thinking-feeling (T-F). The Myers-Briggs Type Indicator (MBTI), an instrument based on Jungian typology, measures these personality dimensions. Studies using the MBTI will be treated in a separate section.

Jungian Personality Dimensions


In 1975 Myers published the MBTI, an instrument predicated on Jung's theory of typology. In addition to the three scales measuring Jung's original personality dimensions, Myers included a fourth scale that measures a dimension that was implicit in Jung's theory --judging-perceiving (J-P). According to Myers and McCaulley (1987), individuals with a sensing preference perceive information via the observable, or by way of the senses. In contrast, individuals with an intuitive preference perceive information via the world of possibilities, meanings, and relationships, in other words, by way of insight. Individuals with a thinking preference make decisions that are impersonal and guided by logic and the principles of cause and effect. Conversely, individuals with a feeling preference make decisions based on the relative values and merits of the issues. Judging individuals have a preference for organization and rules. In contrast, perceiving individuals have a preference for flexibility and new possibilities. A more complete description of these four personality dimensions is provided in Figure 1. Two kinds of descriptive measures are obtained from the MBTI. First, each individual receives a score that is indicative of a preference for one of the poles on each index (e.g., thinking or feeling). Second, the four preferences from these indices are combined to form one of 16 possible MBTI personality types (e.g., INTP). To maintain clarity in this review, an individual's score on a particular index is designated as an individual's preference (e.g., thinking), while the combination of an individual's four preferences, for example, introversion-intuitivethinking-perceiving, is designated as the individual's type (i.e., INTP). Several investigations have examined the relation between Jungian personality dimensions and computer aptitude and achievement. To date, there is little evidence to support a relationship between these variables. The one exception is a recent study by Evans and Simkin (1989) in which the abbreviated version of the MBTI was administered to students in an entry-level business computer course. Six

110

P oc ius

The Four Personality Dimensions of the Myers-Briggs Type Indicator


EXTRAVERSION VS INTROVERSION

Extraverts like variety and action; are relaxed and


confident; are understandable and accessible; tend to reach conclusions by thinking out loud; are often good at greeting people; tend to be faster, dislike complicated procedures; are often impatient with long slow jobs; often act quickly, sometimes without thinking; welcome telephone interruptions and intrusions; like to have people around; and usually communicate freely.

Introverts like quiet for coneentratiou; are reserved and


questioning; are subtle and impenetrable; do not state conclusions until weU-thought out; have trouble remembering names and faces; tend to be careful with details; dislike sweeping statements; tend not to mind working on one project for a long time uninterruptedly; like to think a lot before they act, sometimes without acting; dislike telephone intrusions and interruptions; work contently alone; and have some problems communicating.

SENSING VS INTUITIVE

Sensing individuals face life observingly, at the


expense of imagination; are by nature pleasure-lovers, craving enjoyment; are present-oriented and dependent on their physical surroundings; like an established way of doing things; work steadily, with realistic idea of how long it will take; usually reach a conclusion step by step; seldom make errors of fact; are patient with precise work and routine details; and are impatient when the details get complicated. THINKING VS FEELING

intuitive individuals face life expectandy, at the


expense of observation; are by nature initiators, inventors, and promotors, craving inspiration; are future-oriented and independent of their physical surroundings; dislike doing the same thing repeatedly; work in bursts of energy powered by enthusiasm, with slack periods in between; usually reach a conclusion quickly; frequently make errors of fact; dislike taking time for precision and routine details; and are patient with complicated situations.

Thinking individuals value logic over sentiment; prefer


truth over tact; are naturally brief and business-like; are analytically oriented, responding to people's thoughts; do not show emotion readily and are often uncomfortable dealing with people's feelings; may hurt people's feelings without knowing it; can get along without harmony; tend to decide impersonally, sometimes paying insufficient attention to people's wishes; tend to be firm-minded; and need to be treated fairly. JUDGING VS PERCEIVING

Feeling individuals value sentiment over logic; prefer


tact over truth; are naturally friendly; are people oriented, responding to people's values; tend to be very aware of other people and their feelings; enjoy pleasing people, even in unimportam things; need harmony; often let decisions be influenced by their own or other people's likes and wishes; tend to be sympathetic; and need occasional praise.

Judging individuals are more decisive than curious; are


rational, self-regimented, purposeful, and exacting; aim to be right; live according to plans, work best when they can plan their work and follow their plans; like to get things settled and finished; may decide things too quickly; and may dislike to interrupt the project they are on for a more urgent one.

Perceiving individuals are more curious than decisive;


are empirical, flexible, adaptable, and tolerant; aim to miss nothing; live according to the situation of the moment, adapt well to changing situations; do not mind leaving things open for alterations; may have trouble making decisons; and may start too many projects and have difficulty in finishing them.

M y e r s & M y e r s (1980)

Figure 1. The four personality dimensions of the Myers-Briggs type Indicator.

measures of computer proficiency were used: a homework score consisting of BASIC programming assignments, two BASIC programming scores taken from the programming questions on the first midterm and final, respectively, two midterm scores, and one final score. According to the authors, the final did not test programming skills as did the midterms. Using multiple regression, several significant correlations were reported: Homework scores and sensing preferences were positively correlated; BASIC 1 scores and introversion were correlated; Midterm 1 scores and intuitive, thinking, and judging preferences were correlated; and Midterm 2 and introverted and feeling scores were correlated.

Personality factors in human-computer interaction

111

Consistent with studies in the previous section, there is some evidence that introversion is positively related to programming success. In addition, these results suggest that intuitive individuals excel in test-taking situations whereas sensing individuals function more effectively in less time-constrained situations. Myers and Myers (1980) note that sensing individuals do not perform as well as intuitive individuals on intelligence and scholastic aptitude tests. This is because sensing types have a greater concern for accuracy and therefore take the time to read a question several times before answering. As a result, they end up answering fewer questions than intuitives. This may explain why intuitive individuals performed better on the first midterm whereas sensing individuals performed better on homework assignments. It is unclear why feeling individuals performed better than thinking individuals on the second midterm. Individuals who prefer thinking tend to be more skillful in handling matters that deal with inanimate objects, machinery, principles, or theories - - all of which can be handled logically. Feeling individuals are better skilled in situations involving people, and how they can be persuaded or helped. The remaining studies report no significant findings between these dimensions and computer aptitude and achievement. For instance, in one study the relation between MBTI measures and scores on a programming aptitude test was investigated. Calaway (1982) administered the Aptitude Test for Programmer Personnel and the MBTI to 86 graduate students enrolled in four research and statistics courses. No significant differences were found between scores on the four preference indices of the MBTI and computer aptitude scores. Thus, Jungian personality dimensions were not related to programming aptitude when a programming aptitude test was used as a criterion. Other studies investigating the relation between these dimensions and performance in an introductory computer science course also report negative findings. Consistent with Calaway's findings that Jungian personality dimensions and programming aptitude test performance are not related, there is no evidence to indicate a relation between these dimensions and introductory programming course performance. For instance, Corman (1986) administered the MBTI to 51 students enrolled in an introductory COBOL class. A regression model with current course average as the dependent variable did not reflect any changes when personality variables were considered in the analysis. Whipkey and Stephens (1984) obtained the MBTI preference scores of 88 students enrolled in an introductory Pascal programming course. These authors reported that judging-perceiving preference scores were not significantly correlated with course performance at the .05 level, although correlations between judgingperceiving preference scores and programming performance of introverted and sensing individuals were significant. However, these authors did not correct for the probability of an increase in Type I error when these interactions were investigated. Given that the probability of a Type I error increases as the number of comparisons increases when the null hypothesis is true, the cz level must be corrected to account for this increase (Keppel, 1982). The corrected cz level for this comparison is tx2 = 1 - (1 - oq) c = o~ 2 = 1 - (1 - .05) 2 = .0975, where c is the number of comparisons, tx1 is the comparison-based error rate, and o~ 2 is the experiment-based error rate when multiple comparisons are made. Therefore, contrary to the findings reported, these comparisons are not significant at the .05 level. In another study Lee (1985) examined conventional methods vs library-based CAI for teaching an introductory FORTRAN course to undergraduate students. The MBTI was used to determine whether these personality indices would account for a

112

P ocius

significant amount of the performance variance in the course. Although the class that utilized the library-based CAI scored significantly higher on programming assignments than the class using conventional teaching methods, none of the MBTI indices accounted for a significant amount of variance in the programming assignment scores. In a final study, Werth (1986) administered the MBTI to 58 students who had completed an introductory PASCAL course. A majority of these students were computer science or electrical engineering majors. There were no significant correlations between grade and personality measures. However, the MBTI personality profile of these beginning computer science (CS) students differs dramatically from the estimated norms for the general population. The population of computer science students consist of more individuals with introverted (CS students, 63% : general population, 25%), intuitive (CS students, 51% : general population, 25%) and thinking (CS students, 77% : general population, 50%) preferences than the general population. Thus, although little relation between Jungian personality dimensions and computer aptitude and achievement is noted, there is a correlation between these dimensions and the selection of computer science as an academic major.
Internal-External Locus of Control

Although the most researched personality dimension in the human-computer interaction literature is introversion-extraversion, a few other constructs have been investigated. An additional construct that has been examined in relation to programming aptitude and achievement is personal locus of control. This construct is defined as a generalized expectancy for internal versus external control of reinforcements (Rotter, 1966). Internalizers see a contingency between events and their own actions and are characterized by a high degree of personal control. Externalizers do not perceive a contingency between their own actions and events, but rather, believe that events are caused by factors beyond their control, such as luck or task difficulty. Nowaczyk (1984) administered a test form that included nine statements taken from the Rotter (I-E) scale of personal control as modified by Collins (1974) to male and female students enrolled in either an introductory programming course (N = 160), a Cobol course (N = 60), or an upper-level computer course (i.e., Design and Implementation of Programming languages, The Theory of Computation) (N = 66). Multiple regression analysis indicated that perceived locus of control failed to predict performance in any of the courses. However, Nowaczyk notes that the questions did not provide a sufficiently large range of scores for locus of control. In fact, the majority of students were identified as internalizers. The large proportion of internalizers found in these courses suggest several possibilities. One is that internalizers may be less threatened by computers and are more likely to engage in computer-related activities than externalizers. Wesley, Krockover, & Hicks (1985) report that intemalizers exhibited greater knowledge of computer hardware, software, programming, algorithms, computer applications, and the impact of computers on society prior to taking a computer literacy course. It is also plausible that externalizers did not complete the course. Nowaczyk reports that completion rate for the Introductory Programming course was 78% while 36% of the students completed the Cobol course. It would be interesting to determine whether internalizers and externalizers differ in enrollment and attrition rate in computer programming courses.

Personality factors in human-computer interaction

113

Compulsivenessand Stress-inducing Personality Traits


Compulsiveness and stress-inducing personality traits have also been examined in relation to programming aptitude and achievement. Kagan & Pietron (1986) investigated the relation between these personality traits and success on a computer programming test. During the second week of the semester, 90 students enrolled in a computer literacy course called "Computers in Business" (i.e., a course designed to teach noncomputer science majors how to use simple software packages related to their own field of specialization) were administered a modified version of the programming aptitude test (Konalina, Wileman, & Stephens, 1983) and two personality measures: (a) The Compulsiveness Inventory (Kagan & Squires, 1985), containing three subscales: Indecision and Double-checking, Order and Regularity, Detail and Perfectionism; and (b) Dispositional Stress Scales, containing five subscales developed by Kagan to assess personality traits that tend to induce stress across a wide variety of situations: Overload, Lack of Self-confidence, Time Urgency, Need to Keep Busy, Anxiety. Kagan and Pietron reported significant correlations between compulsive and stress-inducing personality traits and scores on the programming aptitude test. Individuals with high scores on the Overload (stress) and Order and Regularity (compulsiveness) scales obtained lower scores on the programming aptitude test. These results indicate that subjects who tended to feel a general sense of being overwhelmed by work and who were more concerned with accomplishing tasks according to rigid procedures performed more poorly on the programming aptitude test. Kagan and Pietron suggest that a certain amount of relaxation and flexibility in attempting new methods is compatible with programming aptitude. This interpretation is consistent with the positive relation noted between a relaxed attitude and successful performance during the first half of a programming course in Kagan and Douthat's study.

Summary
Three conclusions are drawn from these findings concerning the relation between personality factors and programming aptitude and achievement. First, there is some evidence that introverts and individuals with traits characterizing the introvert tend to perform better in computer programming courses (Alspaugh, 1972; Evans & Simkin, 1989; Kagan & Douhat, 1985; Petersen & Howe, 1979). In addition, a low level of stress and anxiety and the ability to be flexible appear to be positively related to programming aptitude and achievement (Kagan & Douhat, 1985; Kagan & Pietron, 1986). Second, with the exception of one study (Evans & Simkin, 1989), there is no evidence that Jungian personality dimensions (i.e., introversion-extraversion, sensing-intuitive, thinking-feeling, judging-perceiving) correlate with programming aptitude or achievement (Calaway, 1982; Corman, 1986; Lee, 1985; Werth, 1986; Whipkey & Stephens, 1984). Third, although there is little support for a relation between Jungian personality dimensions and programming aptitude or achievement, there is evidence to suggest a relation between these personality dimensions and the selection of computer science as an academic major (Werth, 1986). In particular, computer science majors reported introverted, thinking, and intuitive preferences. It is also speculated that internalizers may be more drawn to programming courses and to computer-related activities in general.

114

P ocius

THE PROGRAMMER PERSONALITY

Although the term "programmer personality" has become clich6 (Lyons, 1985), there is no single job description to which the role of the programmer is adequately confined. Programming skills are found in many job title descriptions - - computer programmers, data processors, programmer analysts, systems analysts, and computer scientists, to name a few. Therefore, the second realm of study involves the investigation of the personality of individuals who at some capacity in their profession are programmers. The variability in job title terminology presented by the authors of these studies is noted, with the caveat that each job title can represent a multitude of different programming and nonprogramming skills. This literature is divided into two sections: (a) Jungian personality dimensions, and (b) other personality dimensions. A summary of these studies is presented in Table 2.

Jungian Personality Dimensions


Although the need for additional research on the personality characteristics of programmers is recognized (Shneiderman, 1980), only a small amount of research has
Table 2. Summary of Studies Investigating the Programmer Personality
Study Lyons (1985) N 1,229 Personality MBTI preference
ii ti

Results 1(67%) N (54%) T a (68%)

type

Tb (85%) Is-rJ (23%)


INTJ (16%) INTP (12%)

Bush & Schkade (1985)

58

MBTI preference type MBTI preference


4=

T (74%) ISTJ (25%) INTJ (16%)

Barnes (1984)

59

I (61%)
N (54%)

T (58%)
MMPI SVIB Sitton & Chmelir (1984) 27 KTS preference no elevated profiles

little preference for


extraverted activities

N (> 50%)
ENTP (% not reported) _S/T/J combinations (% not reported) I (63%) N (51%) T a (77%)

type
Denham (1979) 17 MBTI prelerence MBTI

Werth (1986)

58c

Note. MBTI = Myers-Briggs Type Indicator; MMPI = Minnesota Multiphasic Personality Inventory; SVlB = Strong Vocational Interest Blank; KTS = Keirsey
Temperament Sorter; I = introversion; N = intuitive; T = thinking; ISTJ = introversion/sensing/thinking/judging; INTJ = introversion/intuitive/thinking/ judging;INTP=introversion/intuitive/thinking/perceiving;ENTP= extraversion/intuitive/thinking/perceiving; _STJ = _/sensing/thinking/judging. Frequency estimates for the general population are presented in the text. aFemales, bMales, cComputer science students.

Personality factors in human-computer interaction

115

been devoted to the investigation of the programmer personality. Yet, although there is little scientific study in this area, stereotypes of computer programmers exist (Matheson & Strickland, 1986; Sitton & Chmelir, 1984). Most of the studies that investigate the personalities of computer programmers have focused on the Jungian personality dimensions. In a majority of these studies, significance tests are not reported. However, both MBTI preference and type frequencies are presented and compared with frequencies that are estimated for the general population (see Myers & McCaulley, 1985). The available data presented in each study is reported as completely as possible. Results from studies investigating the programmer personality suggest consistent preference differences between computer programmers and the general population. Estimates of preference frequencies in the United States population have shown that about 75% prefer extraversion, 75% prefer sensing, 60% of males prefer thinking while 65% of females prefer feeling, and 55 to 65% prefer judging. Compared to the general population, computer programmers report more introverted, intuitive, and thinking preferences. For instance, in an ongoing survey Lyons (1985) investigated the personalities and work preferences of computer professionals employed in over 100 different companies. At the time of this analysis, 1,229 computer professionals had participated in the survey. Sixty-seven percent of the computer professionals surveyed were introverts, and 54% percent of the sample consisted of individuals with intuitive preferences. There was a strong bias toward a thinking preference for both females and males, about 68% and 85% respectively. Consistent with this latter finding, a study in which the MBTI was administered to 58 computer professionals (40 programmer analysts and 18 systems analysts) at a north Texas high-technology aerospace firm found that 74% of their population reported a thinking preference (Bush & Schkade, 1985). In another study, Barnes (1974) administered the MBTI to 34 computer programmers and 25 programmer trainees. Computer programmers were volunteers who had been employed as programmers for at least three months, whereas trainees were volunteer programmers who had been employed less than three months and volunteer students enrolled in an Advanced Assembler course. Consistent with the previous MBTI studies, both groups reported an introverted and intuitive bias relative to the general population. Inspection of the data indicates that 61% of the programmers and programmer trainees were introverts, 54% exhibited an intuitive preference, 58% exhibited a thinking preference, and 54% exhibited a judging preference. Because the MBTI data for males (N = 36) and females (N = 23) were combined in this study, it cannot be determined whether this sample exhibited a thinking bias relative to males and females in the general population. Using a short form of the MBTI, the Keirsey Temperament Sorter (KTS), Sitton and Chmelir (1984) obtained similar findings. The KTS was administered to 27 volunteer data processors from four Texas computer installations. Computer programmers were more than twice as likely as the general population to score in the intuitive category. Furthermore, although the majority of computer programmers in this sample were extraverts, these authors report that they were significantly less extraverted than the general population. Type frequencies have also been investigated. Lyons (1985) found that out of the 16 possible personality types, three personality types - - ISTJ, INTJ, and INTP accounted for just over 50% of the computer programmer population (23%, 16%, and 12%, respectively). Six percent of the general population is reported in the ISTJ category, while 1% of the general population is categorized as INTJ types. INTP types also characterize 1% of the general population. Bush and Schkade (1985) and

116

P ocius

Sitton and Chmelir (1984) also report type distributions, although the large sample size necessary for the adequate analysis of type distributions precludes any generalization of these two studies to a larger population. Bush and Schkade found that the largest number of MBTI types among computer professionals was represented by the ISTJ personality, in which almost 25% of this population was categorized. The second most frequently reported type was INTJ, with almost 16% of the programmers found in this category. Given the small sample size employed in this study (N = 58), it is interesting that these results mirror Lyons' findings. Contrary to these results, Sitton and Chmelir found that ENTP types were the most commonly represented type in their sample, although the percentage of ENTP types was not reported. ENTP types represent 5% of the general population. In addition to studies investigating the personality of programmers at the professional level, the MBTI was given to individuals in skilled/semi-skilled computer occupations. Denham (1979) administered the MBTI to 17 female data entry operators. A significantly greater frequency of _S/T/J combinations among high-speed operators than low-speed operators was reported. Furthermore, low-speed operators had a greater number of bipolar opposite (_N/F/P) combinations than highspeed operators. Although Yate's correction for continuity is recommended for small cell size (< 10) to avoid overestimates of the true Z2 value (Bruning & Kintz, 1987), it is unclear whether this correction was applied to the data. At best, these findings suggest some consistency with the results of Bush and Schkade and Lyons - - the greater frequency of S[I'/J combinations reported among high-speed operators is congruent with the most common personality type (i.e., ISTJ) found among computer programmers.

Other Personality Dimensions


One study measured other aspects of the programmer's personality. In addition to the MBTI, Barnes administered the Minnesota Multiphasic Personality Inventory (MMPI) (Hathaway & McKinley, 1967) and an early form of the Strong Vocational Interest Blank (SVIB) (Campbell, 1969) to computer programmers and programmer trainees. The MMPI measures the primary personality characteristics that affect personal and social adjustment and provides a personality profile based on 10 clinical scales: Hs (hypochondriasis), D (depression), Hy (hysteria), Pd (psychopathic deviate), Mf (masculinity-femininity), Pa (paranoia), Pt (psychasthenia), Sc (schizophrenia), Ma (hypomania), and Si (social introversion). This instrument was originally designed to differentiate between normal and pathological tendencies, but in subsequent usage, these scales have been treated more as linear measures of personality traits (e.g., a high elevated score on the paranoia scale may suggest a high degree of resentfulness and suspiciousness) (Anastasi, 1976). The SVIB is designed to identify the interests of individuals in a manner that permits the individual to compare his or her interests with persons employed in a specific occupation. Although the SVIB is an interest inventory, interest inventories are argued by some to be expressions of personality (Holland, 1985). By inspecting histograms for each of the MMPI scales, Barnes reported no pathological tendencies in these samples. A flat histogram of the Si scale revealed a large range of scores with the scores fairly evenly distributed along the continuum. Given the strong tendency toward introversion among computer programmers, it would be interesting to determine whether this sample had an elevated score on the social introversion scale relative to the general population. Interpretation of the SVIB results suggested that programmers manifest little interest in extraverted or

Personality factors in human-computer interaction

117

authoritative activities (i.e., Public Speaking, Law/Politics, Business Management, Sales, Military Activities, and Technical Supervision) but preferred the activities Mathematics, Nature, and Music. Early studies examining SVIB profiles of computer programmers report remarkedly parallel findings - - computer programmers express a strong interest in music and mathematics (Perry & Cannon, 1967; 1968). These activities require a higti degree of concentration and reflectiveness, which are typical characteristics of the introvert. Further, these early studies report that a major striking characteristic of this population is their disinterest in people.

Summary
The literature investigating the programmer personality is not extensive, and is characterized by several limitations. Some of these studies are published in trade journals, and therefore do not present a detailed description of the analysis. In addition, some of the studies suffer from methodological problems such as small sample size and inadequate statistical analyses. Given these limitations, however, consistent personality differences are noted between computer programmers and the general population. Compared to the general population, programmers report more introverted, intuitive, and thinking preferences (Barnes, 1974; Bush & Schkade, 1985; Lyons, 1985; Sitton & Chmelir, 1984). The results of Werth's study, in which computer science majors were found to exhibit introverted, intuitive, and thinking preferences, are consistent with the findings for computer programmers. The programmer's introverted, intuitive personality is also reflected in their occupational interests. Computer programmers prefer independent, nonroutine work that does not require close interpersonal contact (Barnes, 1974; Perry & Cannon, 1967; 1968). Programmers also appear to be commonly represented by certain types. The most frequent MBTI personality types found among computer programmers are the ISTJ, INTJ, and INTP types (Bush & Schkade, 1985; Lyons, 1985). The ISTJ type is the most prevalent personality type found among programming professionals. Even among semi-skilled computer operators, the more skilled data-entry operators are frequently represented by S/T/J combinations (Denham, 1979). Overall, these studies suggest that personality is a salient factor in the selection of computer programming as both an academic and career choice.

COMPUTER-ASSISTED INSTRUCTION (CAI)


The final topic to be reviewed is the domain of computer-assisted instruction (CAB. In CAI, the computer is employed as a tool for the acquisition of knowledge that generally is not computer-related. In courses that use CAI, course content and tests are stored on the computer with which the student interacts in real time by means of on-line terminals. In general, CAI is typified by a noninterpersonal context. The individual works in a self-pacing environment in which there is little social, emotional, or cognitive exchange between individuals. Information to be learned is divided into sessions, in which the learner receives immediate reinforcement for his or her responses. The forerunner to CAI was a method of instruction called P r o g r a m m e d Instruction (PI). Unlike CAI, it did not involve a direct interaction between the individual and a computer terminal, but did involve several of the same advantages

118

P ocius

and constraints. Because of the similarities between the PI and CAI task, this literature is also reviewed, but in less detail. This literature is grouped into four sections: (a) introversion-extraversion, (b) Jungian personality dimensions, (c) anxiety, and (d) achievement motivation. A summary of these studies is presented in Table 3.
Introversion-Extraversion

Early studies investigating the effects of personality on PI performance typically explored constructs that characterize the introvert and extravert. Two of these constructs are the degree of sociability and social need of the individual. Exlxaverts are identified as individuals with a high degree of sociability and social need, while the converse is true for introverts. In a review of this literature, Dallos (1975) reports that significant correlations between PI performance and these characteristics exist. However, many of these correlations are mediated by other factors, such as the amount of structure and guidance in the program, and group versus individual use of the program. The PI task conventionally involves a noninterpersonal context and a maximum amount of structure and guidance. Dallos notes that individuals with introverted characteristics perform better on conventional linear or "reception programs" with "maximal structure and guidance," while individuals with extraverted characteristics perform better on nonconventional "discovery programs" with "minimal structure and guidance" (p. 18). For example, extraverts perform more poorly on a PI task in which a set of rules is imposed prior to practice vs rules given following practice, whereas introverts tend to perform better if they are given their rules before practice, although not significantly so (Leith & Trown, 1970). Performance is also related to the interaction between the individual's personality and the degree of interpersonal context of the PI task. For instance, individuals characterized by low social need perform well (Doty & Doty, 1964), while children described as "withdrawn" are more successful than less withdrawn children in a PI learning situation (Traweek, 1964). Interestingly, Greer (cited in Dallos, 1975) reports that extraverted eight-year-olds perform better working with a friend than introverts in the same situation. In one exception, Sawiris (1966) reports no significant relation between extraversion and PI performance for grammar school students. CAI performance is also influenced by the interaction between the individual's personality and the degree of interpersonal context of the CAI task. Sutter and Reid (1969) compared the achievement and attitude of students who took a CAI problem-solving course with a partner (interpersonal context) with those who took the course alone (noninterpersonal context). The level of sociability of 82 male students was assessed via the sociability scale of the California Psychological Inventory prior to undertaking the CAI course. The sociability scale identifies individuals characterized by an outgoing, sociable, and participative temperament. Degree of sociability did not affect attitude toward CAI. However, individuals with a high degree of sociability performed significantly better in the interpersonal context, while it was suggested that individuals with a low degree of sociability performed better working alone at the CAI terminal. This finding was repeated in a subsequent study. Reid, Palmer, Whitlock, and Jones (1973) investigated the CAI performance of 81 pairs of undergraduate algebra students learning scientific notation and exponentiation. Again, pairs in which both students reported a high degree of sociability performed significantly better than low sociability pairs. Overall, these findings suggest that individuals with introverted qualities perform better than individuals with extraverted qualities in the conventionally defined noninter-

Personality factors in human-computer interaction

119

Table 3. Summary of Studies Investigating Relationship Between Personality and ComputerAssisted Instruction (CAI) Performance
Measure Study Dallos (1975) Leith & Trown (1970) N NA 124 Personality extraversion CAI/PI PI achievement Results NR with task conventionality NR t = 2.46* NS r = -.43** NS PR t = 2.95** NS PR t = 2.99** NS r = NS, PCA factor loading = -.374 NS tb =-.12, tc = .31" i b -- -.20, tc = -.36* tb = .26, tc = -.32* tc = .39** NS NS NS NS NS r = -.60* NS NR F = 22.83* PR F = 5.14" NS NS t e = -.46*, t~ -- -.48* NS PR rg** 38% h undetermined NR F = 5.08* NR F = 5.18" NS NR F=11.20" NR F = 9.25* PR F = 8.00* F = 4.16 *i

HBPI extraversion PI achievement a general anxiety PI achievement G -ZTS social need EPPS achM PI achievement "

Doty & Dory (1964)

100

Traweek (1964)

186

CTP withdrawn PI achievement SAT general anxiety" test anxiety HSPQ extraversion PI achievement general anxiety" CPI sociability TAQ test anxiety

Sawiris (1966)

25

Sutter & Reid (1969)

82

Attitude toward CAI CAI achievement Attitude toward CAI CAI achievement CAI CAI CAI CAI achievement completion time achievement completion time

Reid, Palmer, Whitlock, & Jones (1973)

162

C PI sociability achM TAQ test anxiety

CAI achievement CAI completion time PI completion time PI number of errors Attitude toward CAI d CAI number of errors Attitude toward CAI CAI achievement Attitude toward CAT CAT achievement CAI completion time CAI attrition rate CAI achievement PI completion time PI number of errors PI achievement PI completion time PI number of errors PI achievement

Apter & Murgatroyd (1975)

24

EPI extraversion MBTI S/N E/I MBTI MBTI T/F MBTI SIN E-P MBTI TAQ test anxiety IPIT achM

White & Smith (1974)

105

Howard (1986) Johnson (1984)

79 18

Hoffman & Waters (1982)

155

Kern & Matta (1987)


Kight & Sassenrath (1966)

110 139

(Table 3 continued)

120

P ocius

Table 3. Continued
Measure Study Morris, Blank, McKie, & Rankine (1970) N 252 Personality TASC test anxiety CAI/PI PI number of errors PI completion time PI achievement PI achievement " PI number of correct responses Tobias (1973) O'Neil, Spielberger, & Hansen (1969) O'Neil (1972) 121 29 TAS test anxiety STAI A-Trait A-State STAI A-Trait A-State STAI A-Trait A-State STAI A-Trait A-State (Experiment II) 128 A-Traif A-State CAI achievement CAI number of errors " CAI number of errors " CAI number of errors " " CAI " CAI CAI CAI CAI ,, CAI CAI achievement completion time achievement completion time achievement completion time achievement Results PR F = 4.09" NS F = 5.15"*J NS k PR F = 4.01.1 PR F = 5.43 *k F = 3.14 *m NS

Tobias & Abramson (1971)

144

AAT anxiety . . . . achM

NS NS F = 5.08 *n NS PR Fs = 4.4, 11.2* NS PR F = 7.22** F = 3.51"P F = 2.48"q I NR F = 4.67 *k NS NS PR F = 3.46* F = 4.48 *rk NS k NS F = 3.18 *sl NS k

64

Leherissey, O'Neil, & Hansen (1971)

60

Leherissey, O'Neil, Heinrich, & Hansen (1973) (Experiment I)

148

Note. NA = not applicable; HBPI = H. B. Personality Inventory; G-ZTS = Guilford-Zimmerman T e m p e r a m e n t Survey; EPPS = Edward's Personal Preference Schedule; achM = a c h i e v e m e n t motivation; CTP = California Test of Personality; SAT = Sarason's Anxiety Test; HSPQ = High School Personality Questionnaire; CPI = California Psychological Inventory; TAQ = Sarason's Test Anxiety Questionnaire; EPI = Eysenck Personality Inventory; MBTI = Myers-Briggs Type Indicator; SIN = sensing/intuitive; F_JI = extraversion/introversion; T/F = thinking/feeling; E-P = extraverted-perceiving; IPIT = Iowa Picture Interpretation Test; TASC = Sarason's Test Anxiety Scale for Children; AAT = Achievement Anxiety Test; TAS = Sarason's Test Anxiety Scale; STAI = State-Trait Anxiety Inventory; PI = programmed instruction; CAI = computer-assisted instruction; CAT = computer-assisted testing; NR = negative relationship; NS = not significant; PR = positive relationship; PCA = principle components analysis. Significance scores are tabled when available. aRules before condition, bAIone. Cpaired. dWith increasing CAI responsibility. ?Test 1. fTest 2. gNot reported, hTotal attrition rate. iAchievement motivation x test anxiety interaction, lTest anxiety x pretest condition interaction, kTechnical material. IFamiliar material, mAchievement motivation x response mode x stress x sex interaction, hA-State x tasks interaction. Easy tasks. PA-State x memory condition interaction, qA-Trait x response mode interaction, rA-Trait x response mode x program length interaction, sA-State x response mode interaction. **p < .01. *p ,= .05. NS = not significant.

personal context of the PI task. However, when the opportunity for increased social interaction is introduced into the PI task, extraverted types perform better than introverts in the same situation.

Personality factors in human-computer interaction

121

Finally, performance speed differences between introverts and extraverts are reported. In a study published subsequent to Dallos' review, extraverts were found to work significantly more quickly through a PI task than introverts. Apter and Murgatroyd (1975) administered the Eysenck Personality Inventory to 24 undergraduate psychology students who were assessed on their overall performance using a "teaching machine." There were no significant differences between introverts and extraverts on total number of errors made during the program. However, extraverts took significantly less time than introverts to finish the program. According to the authors, all the subjects described the program as boring. Because extraverts tend to be less tolerant than introverts of repetitive, monotonous tasks (Pardes, 1965), extraverts may have completed the program more quickly to escape the tediousness of the task. This finding also reflects the basic decision-making style of these two types of individuals. Extraverts by nature tend to act quickly and sometimes without thinking, whereas introverts are reflective and do not state conclusions until they are well-thought out. Given the introvert's reflective nature and the extravert's tendency toward impulsiveness, the finding that there is no difference in number of errors between introverts and extraverts is unexpected. These authors report that errors were rarely made in the PI task; therefore, task difficulty level may not have been sufficient to differentiate between the number of errors typically made by introverts and extraverts. With increased task difficulty level, it is plausible that introverts will make fewer errors than extraverts.

Jungian Personality Dimensions


Several available studies address the relation between Jungian personality dimensions and CAI performance. These studies predominantly involve the investigation of two variables - - attitude toward using CAI and CAI performance. The results of these investigations have been mixed. In one early study, White and Smith (1974) investigated the effects of personality on attitude toward using CAI and CAI performance in a course designed to provide students with the attitudes and skills necessary for planning effective learning. The extraversion-introversion and sensingintuitive indices of the MBTI were used to assess the effects of degree of student responsibility for selecting learning activities in a CAI program. One hundred and five elementary-school teacher education majors assigned to one of four treatment groups of increasing levels of CAI responsibility completed the program. There were no performance differences between introverts and extraverts with the exception that introverts assigned to the second treatment group had more incorrect answers than extraverts in the same group. An interaction was reported between these personality dimensions and attitude toward using CAI. As the responsibility for the selection of learning activities was shifted to the students, intuitive individuals became more dissatisfied with CAI as a means of instruction, while sensing individuals became more satisfied. White and Smith suggest that sensing individuals require flexible CAI whereas intuitive individuals may benefit from more structured computer-assisted instruction. This positive attitude toward CAI was stronger for the introverted-sensing individual than the extraverted-sensing individual, suggesting that among sensing individuals, introverts regard a flexible CAI program more favorably than extraverts. Howard (1986) assessed the impact of personality characteristics on individuals' performance and attitude toward using CA1. Seventy-nine students were divided into experimental and control groups in which the experimental group received computer-assisted instruction of the metric system. There were no significant cor-

122

P oci us

relations between any of the four MBTI preference indices and achievement or retention of the CAI course material. In addition, there were no significant differences between any of these indices and attitude toward using CA1. The relation between personality factors and testing using CAI was also studied. Johnson (1984) investigated the effects of Jungian personality dimensions on both attitude toward computer-assisted-testing (CAT) and on CAT performance. Eighteen nursing students were administered the MBTI prior to undergoing CAT of medical terminology. Consistent with Howard's results, no significant relationships were found between CAT scores and MBTI measures. However, individuals with thinking preferences had more favorable attitudes toward CAT than individuals with feeling preferences. This finding is theoretically consistent with the thinking individual's preference for making decisions based on impersonal analysis and logic. Although the previous studies found no differences in retention of the material learned using CAI, Hoffman and Waters (1982) found several performance differences among MBTI personality types when completion and attrition rate in a CAI course were examined. The MBTI was administered to 155 students at the Naval Technical Training Center at Pensacola, Florida to assess whether there were significant differences between personality measures and students' performance in learning and transcribing Morse code using CA1. Sensing individuals completed the computer-assisted portion of the program significantly faster than intuitive individuals. In addition, a much higher than expected rate of EN-P types did not finish the course. Fifty-three percent of extraverted-perceiving individuals failed to complete the course, comprising 38% of all individuals who dropped the CAI program. The most common reasons reported for dropping the course were "lack of concentration powers and motivation problems connected with difficulty with Morse code" (p. 21). These authors suggest that CAI favors persons with a sensing preference - - individuals who can pay attention to details, have an affinity for memorizing facts, and can stay with a single task until completion. A final study by Kern and Matta (1987) examined the relation between Jungian personality dimensions and CAI performance. One hundred and ten students from the College of Business at the University of Notre Dame were administered the MBTI followed by computer-assisted instruction in Lotus 123, a microcomputer spread-sheet software package. These students were categorized into one of four degree programs offered by the College of Business. Although the results reported by the authors of this study are presented in the summary table, inspection of the MBTI data provided does not reveal how these findings were determined. It is unclear whether these conclusions were drawn from the sample as a whole, or from a subset of the sample defined by one of the four programs offered by the College of Business. In addition, no tests of significance were reported. Given the lack of clarity of the findings presented, any conclusions drawn from this study are suspect. In summary, there is currently little support for a relation between Jungian personality dimensions and achievement and retention of CAI course material. However, performance differences not related to CAI achievement are noted. Individuals with a sensing preference will tend to complete a CAI course, and will complete the course at a faster rate than individuals with an intuitive preference. On the other hand, EN-P types and extraverted-perceiving individuals in particular will have difficulty completing CAI courses (Hoffman & Waters, 1982). Further, some attitude differences exist. There is evidence that thinking individuals respond

Personality factors in human-computer interaction

123

more favorably to CAT than feeling individuals (Johnson, 1984), and that sensing individuals are more satisfied with a flexible CAI program in which they are given the opportunity to make choices (White & Smith, 1974).

Anxiety
Another personality construct that has received a great deal of attention in the PI and CAI literature is anxiety. One of the advantages of PI and CAI is that they offer a fairly well-structured, nonstressful situation in which the learner works at his or her own pace, competes only with himself, and is rewarded for cautiousness. Therefore, it is has been anticipated that individuals who are prone to anxiety will benefit from PI or CAI (Kight & Sassenrath, 1966; Ripple, O'Reilly, Wightman, & Dacey, 1965; Tobias, 1973; Tobias & Abramson, 1971). Several kinds of anxiety are defined in the literature. For instance, the Test Anxiety Questionnaires (Mandler & Sarason, 1952; Sarason, 1959) measure two kinds of anxiety: anxiety in test-like situations and general anxiety. Early studies suggest that test anxiety is positively correlated with successful PI or CAI performance. Traweek (1964) investigated the effect of test anxiety on PI performance using Sarason's Test Anxiety Scale for Children. Successful fourth-grade students (Total N = 186) in the PI learning task reported more tendencies to be test-anxious than unsuccessful fourth-grade students. A similar finding was partially replicated for older students. Kight and Sassenrath (1966) administered the Test Anxiety Questionnaire to 139 undergraduate students in educational psychology prior to their completing a PI phase of a course on test construction and evaluation. Students who reported high test anxiety required less time to complete the programmed instruction and made fewer errors, but failed to exhibit higher retention scores than students scoring low on test anxiety. Subsequent studies suggest that this relationship is more complex. Several studies report that the relation between test anxiety and PI and CAI performance is mediated by certain factors in the learning situation. For instance, the performance level of test anxious individuals improves if the PI or CAI task is not accompanied by a pretest, the individual works alone, and stress is introduced into a task that involves the use of familiar material. Morris, Blank, McKie and Rankine (1970) gave Sarason's Test Anxiety Scale for Children to 252 sixth-grade students prior to their completion of a PI program on English money. Contrary to Kight and Sassenrath's findings, children who reported high test anxiety made more errors on a small-step linear version of the program, and did not differ from low test anxiety students in overall time to completion. However, high test anxiety students performed as well as or better than low test anxiety students in a no pretest condition, whereas low test anxiety students did better than high test anxiety students in a pretest condition. During a pretest, high test anxiety students may fear they are being "evaluated," resulting in a threat to their self-esteem. Wine's (1971) review of the test anxiety literature indicates that on difficult tasks in which evaluative stress is present, high test anxiety individuals perform more poorly than low test anxiety individuals. Wine proposes that, compared to low test anxiety individuals, high test anxiety individuals become more self-preoccupied during testing situations and therefore shift their attention away from task relevant problems during performance. The interaction between test anxiety and the interpersonal context of the task also affects CAI performance. Sutter and Reid (1969) compared the achievement and attitudes of low and high test anxiety college students who took a CAI problem-solving course with a parmer and those who took the course alone. Test anxi-

124

Pocius

ety was measured using the Test Anxiety Questionnaire. Although individuals with high test anxiety had greater negative attitudes toward CAI than low test anxiety individuals in both the paired and alone groups, students reporting high test anxiety achieved better working alone, whereas those low in test anxiety achieved better working with a partner. These authors speculate that, for high test anxiety students, working alone lessens the threat presented by the presence of a partner. Consistent with Wine's interpretation, individuals with high test anxiety are more free to make task-relevant responses if they do not face competition or evaluation from a peer. The level of test anxiety of one's parmer also affects the amount of time needed to complete a CAI course. Reid, Palmer, Whitlock, and Jones (1973) found that for 81 pairs of college algebra students completing a CAI course designed to teach scientific notation and exponentiation, pairs in which both individuals reported high test anxiety took longer to complete the course than low test anxiety and mixed test anxiety pairs, while low test anxiety pairs completed the CAI course in the least amount of time (F = 1.65, p < .20). The overall correlation between test anxiety and time approached significance (.19, p < .10). No significant correlations were reported between test anxiety and CAI achievement. Tobias and Abramson (1971) report that anxiety interacts with stress to raise PI performance if the task involves the use of familiar material. The Achievement Anxiety Test (AAT) (Alpert & Haber, 1960) was administered to 144 educational psychology students prior to completing a programmed instruction course. Subjects were classified as either low or high anxious based on the AAT - - scale of this test, which taps the debilitating effects of anxiety on performance in achievement situations, such as the testing situation. Based on previous findings that differences between high and low anxiety groups occur only under conditions of stress, a stress and no-stress condition were introduced into the programmed instruction. Stress was induced by instructions relating achievement on the program to intelligence and learning ability, thus incorporating a threat to the individual's self-esteem. Contrary to expectations, there were no interactions between debilitating anxiety and stress on PI performance for technical subject matter. However, debilitating anxiety interacted with stress to raise achievement in the PI program using easy, familiar content. Tobias (1973) investigated the effects of distraction, mode of responding, and test anxiety on CAI achievement. One hundred and twenty-one general psychology students completed the Test Anxiety Scale prior to completing a CAI task under neutral or distraction conditions and using either constructed response mode with feedback or reading mode. Although it was hypothesized that the performance of high anxiety students would be poorer compared to low anxiety students during the distraction condition, there were no significant differences between these two groups for any of these conditions. The distraction condition, which required students to memorize consonant-vowel-consonant syllables while working on the program, may not have been effective in distinguishing between the performance of low- and high-test anxiety students because the distraction condition was not evaluative in nature. Studies using instruments measuring the effects of general anxiety on PI and CAI performance have reported few significant findings. Traweek (1964) found no significant differences between successful and unsuccessful students with respect to general anxiety. Level of general anxiety as measured by the H. B. Personality Inventory (Hallworth, 1962) was ascertained for 128 junior high students prior to their working through a programmed text on modem mathematics (Leith & Trown,

Personality factors in human-computer interaction

125

1970). No significant relation was noted between general anxiety and PI performance. Sawiris (1966) reports that anxiety as measured by the High School Personality Questionnaire by Cattell and Beloff is not correlated with achievement using programmed instruction to learn geometry, although principal components analysis revealed an "achievement" factor with a high negative loading on the anxiety variable. Similar to Sutter and Reid's (1969) findings, he reports that the absence of anxiety is associated with a favorable attitude to programmed instruction. The findings that test anxiety but not general anxiety is related to performance suggests that only anxiety characterized by a threat to self-esteem is related to PI and CAI performance. A series of studies investigating the effects of anxiety on PI and CAI performance has focused on Trait-State Anxiety Theory (Spielberger, 1971). The instrument designed to measure trait-state anxiety is the State-Trait Anxiety Inventory (Spielberger, Gorsuch, & Lushene, 1970). State anxiety (A-State) refers to a transitory state or condition that is characterized by feelings of tension, apprehension, and heightened autonomic nervous system activity. Trait anxiety (A-Trait) refers to a relatively stable personality trait reflecting individual differences in anxiety proneness. High A-Trait individuals are more predisposed than low A-Trait individuals to perceive certain situations as dangerous, particularly situations characterized by a threat to self-esteem. Trait-State Anxiety Theory predicts that high ATrait individuals will exhibit higher levels of state anxiety during situations perceived as threatening. This prediction has been supported in the PI and CAI literature (Leherissey, O'Neil, Heinrech, & Hansen, 1973; O'Neil, 1972; O'Neil, Spielberger, & Hanson, 1969). The effects of trait-state anxiety on PI and CAI performance have been investigated within the framework of Drive Theory. Drive Theory (Spence, 1958; Taylor, 1956) predicts that the performance of high-anxious individuals will be inferior to that of low-anxious individuals on complex or difficult learning tasks in which competing error tendencies are stronger than correct responses. In contrast, on simple leaming tasks, in which correct responses are dominant relative to incorrect response tendencies, it is expected that the performance of high-anxious subjects will be superior to that of low-anxious subjects. Similar to the relationships found between test anxiety and PI and CAI performance, trait-state anxiety studies indicate that the relation between anxiety and CAI performance is a complex relationship that is mediated by certain factors in the learning situation. Task difficulty, the presence of memory support, and the type of response mode required of the user is reported to mediate the effects of trait-state anxiety on PI and CAI performance. O'Neil, Spielberger, and Hansen (1969) investigated the effects of trait-state anxiety on learning mathematical materials presented in an easy and a difficult format via CAI. The anxiety level of 29 male and female introductory psychology students was measured using the StateTrait Anxiety Inventory. High A-State students made more errors on the difficult portion of the learning task and fewer errors on the easier portion of the task than low A-State students, thus supporting the predictions made by Drive Theory. While there was no relation between A-Trait and CAI performance, high A-Trait students responded throughout the learning task with higher levels of A-State than low ATrait students. These authors explain this discrepancy by noting that a small number of subjects with A-State scores that were inconsistent with their A-Trait scores contributed so much variability that the analysis of errors on the learning task as a function of A-Trait failed to yield statistically significant F ratios.

126

Pocius

O'Neil (1972) reports contradictory findings. Sixty-four females completed the State-Trait Anxiety Inventory prior to completing a CAI task designed to teach mathematical concepts. Contrary to the predictions of Drive Theory, high A-State students made more errors than low A-State students on the easier sections of a CAI task, but not on the most difficult section of the task. A-Trait was not related to number of CAI errors. One explanation for the failure to support Drive theory is that the interaction between anxiety and task difficulty may be different for males and females. In an unpublished study by O'Neil, Hansen, & Spielberger (1969) in which male students were used, the relation between A-State and CAI performance was consistent with Drive Theory - - high A-State students made significantly more errors on the difficult materials than low A-State students (cited in O'Neil, 1972). Additional studies examining sex differences with respect to the interaction between anxiety and CAI task difficulty may provide information pertaining to the importance of sex differences in CAI performance. Interestingly, these researchers reported that unless an explicit threat to self-esteem was introduced into the learning program, there was no differential change in A-State for subjects who differed in A-Trait. This suggests that that component of anxiety that is characterized by a threat to self-esteem may be the salient factor that affects CAI performance. One study reports that the relation between anxiety and CAI performance is mediated by the presence of memory support. Based on the hypothesis that anxiety interferes with memory, Leherissey, O'Neil, and Hansen (1971) tested the effects of memory support on CAI task performance for individuals with differing levels of state anxiety. Sixty male introductory psychology students were assigned to either a memory support or a no memory support CAI program designed to teach mathematics concepts. Consistent with the previous studies, there was no relation between A-Trait and CAI performance. Although high A-State students made more errors overall, memory support was found to reduce the errors made by high AState students, whereas medium and low A-State students performed equally well with or without memory support. In a subsequent study Leherissey, O'Neil, Heinrich, and Hansen (1973) found that the type of response mode required of the user mediates the relation between anxiety and CAI performance. The State-Trait Anxiety Inventory was administered to 148 female undergraduate students prior to their completion of a CAI program concerning familiar and technical material on heart disease. Four response modes were investigated: reading (read information/no overt response), covert (read information/mentally choose correct answer), modified multiple-choice (overt response required by choosing correct choice), and constructed response (overt response required by drawing correct choice). Unlike the previous studies, A-Trait was significantly related to CAI performance. For familiar material, high A-Trait students performed better in the modified multiple-choice and the constructed response groups and performed poorly on the reading and covert tasks. The performance of low A-Trait students did not differ among the response conditions. For technical material, low A-Trait students performed better than medium or high A-Trait students, again supporting the predictions of Drive Theory. An unexpected finding was that A-State was not related to CAI performance although the main effect of A-State did approach significance (F = 2.98, p < .10). Level of A-Trait was not found to be related to total learning time, although high A-State students took longer to complete the program than either medium or low A-State students. A second study was undertaken in order to replicate the findings of this first experiment and to investigate whether shortening the amount of time required to

Personality factors in human-computer interaction

127

spend on a CAI task would improve performance for 128 female students. Only the reading and constructed response versions were used in this study. Complex relationships were found between A-Trait and response mode, type of material learned, and length of program. For the familiar material, low A-Trait students in the constructed response-long group had better performance than either medium or high A-Trait groups, and performed better in the constructed response-short group than in the other three groups. High A-Trait students performed more poorly in the constructed response-long group than in the other three groups. There was no significant relation noted between A-Trait anxiety and the technical material performance. A-State was also correlated with CAI performance. Increasing levels of AState were negatively related to performance using familiar materials for all groups. While the reading-short group resulted in the best performance for low AState students, it was debilitating for high A-State students. For the reading-long group using familiar materials and for the long version groups using the technical material, low, medium, and high A-State students performed relatively the same. There was no relation between A-State and technical material posttest. Consistent with the results from the first experiment, level of A-Trait was not related to program completion time. The results of these experiments indicate that the effects of anxiety on CAI performance are mediated by the type of response mode required of the user. With the exception of one study (Leherissey, O'Neil, Heinrich, & Hansen, 1973), there is no relation noted between A-Trait and CAI or PI performance. Leherissey, O'Neil, and Hansen (1971) suggest that while measures of A-Trait provide useful information regarding the probability that high levels of A-State will be aroused, taking measurements of A-State in the learning situation affords a more accurate assessment of the relation between anxiety and performance. Measures of state anxiety appear to be sensitive in detecting performance differences, perhaps because they are more relevant to the task at hand. In summary, the literature examining the effects of anxiety on CAI performance suggests that anxiety characterized by a threat to self-esteem is related to PI or CAI performance; in addition, this anxiety facilitates or debilitates PI and CAI performance contingent on specific aspects of the learning task (Kight & Sassenrath, 1966; Leherissey, O'Neil, & Hansen, 1971; Leherissey, O'Neil, Heinrich, & Hansen, 1973; Morris, Blank, McKie, & Rankine, 1970; O'Neil, 1972; O'Neil, Spielberger, & Hansen, 1969; Reid, Palmer, Whitlock, & Jones, 1973; Sutter & Reid, 1969; Tobias, 1973; Tobias & Abramson, 1971; Traweek, 1964).
Achievement Motivation

Achievement motivation - - the drive to be able to do something - - has not been studied extensively in the PI and CAI literature, although it has been proposed that individuals high in achievement motivation have an opportunity to work faster and to show higher overall proficiency in a PI or CAI task than individuals low in achievement motivation (Kight & Sassenrath, 1966). Findings from the few studies examining this relationship have been mixed. Kight and Sassenrath investigated the influence of achievement motivation on PI performance in which achievement motivation was measured using the Iowa Picture Interpretation Test (Hedlund, 1953), a quasi-projective measure of achievement imagery. Undergraduate students who were high in achievement motivation worked through the program at a faster rate, committed fewer errors, and scored higher on the retention test than subjects low in achievement motivation. An interaction between achievement motivation

128

P ocius

and test anxiety was also noted. They conclude that students high in achievement motivation are able to gain greater benefit from the immediate reinforcement characteristic of programmed instruction. Tobias and Abramson (1971) also found a positive relation between achievement motivation and PI performance. The AAT+ scale of the Achievement Anxiety Test (AAT) (Alpert & Haber, 1960) measures the kind of anxiety that facilitates performance in achievement situations, and is reported to be more closely related to achievement motivation than traditional conceptions of anxiety. These researchers found that achievement motivation as measured by the AAT+ scale was positively related to PI posttest scores for technical material, although achievement motivation was not correlated with number of correct responses to the program. A complex interaction between achievement motivation, response mode, sex, and stress was also reported for the program using technical material. Two studies found no support for a relationship between achievement motivation and PI or CAI performance. In one study, achievement motivation was measured using the Achievement Need Scale of Edward's Personal Preference Schedule (Doty & Doty, 1964). There was no relation between need for achievement and PI performance for 100 introductory psychology students assigned to a PI program designed to teach physiological psychology. Similarly, Reid, Palmer, Whitlock, and Jones (1973) investigated the relation between achievement motivation and CAI p e r f o r m a n c e in which a c h i e v e m e n t m o t i v a t i o n was m e a s u r e d using the Achievement via Independence scale of the California Psychological Inventory. No significant overall correlations were noted between achievement motivation and posttest scores or completion time. These results are consistent with negative findings from two unpublished studies (Reid, Whitlock, & Palmer, 1972; Sutter, 1967, cited in Reid et al., 1973). However, there was a significant interaction between achievement motivation and test anxiety for CAI achievement, which was especially notable for female pairs (r = .75, p < .05). Because achievement motivation is defined using different instruments in each of these studies, these findings may be reflecting different facets of achievement motivation. For example, Reid et al. point out that achievement motivation as measured by the Achievement via Independence scale of the California Psychological Inventory emphasizes the motivation to achieve in situations where independence of behavior is important. In CAI, the learning situation allows for little independent behavior, and therefore, achievement motivation via independence may have little relation to CAI performance. Studies incorporating several measures of achievement motivation can be helpful in identifying more precisely what aspects of achievement motivation are related to CAI performance. This literature also suggests that, like anxiety, the effects of achievement motivation on CAI performance is mediated by other personality traits or task characteristics (Kight & Sassenrath, 1966; Reid et al., 1973; Tobias & Abramson, 1971). Perhaps the more important question is not whether individuals high in achievement motivation perform better than individuals low in achievement motivation in a CAI setting, but whether individuals with differing levels of achievement motivation may benefit from different kinds of leaming programs, such as a CAI program vs. a traditional learning program.

Summary
The investigations that assess the relation between introversion-extraversion and PI and CAI performance suggest that this personality trait interacts with performance contingent upon the conventionality of the task at hand (Dallos, 1975; Doty &

Personality factors in human-computer interaction

129

Doty, 1964; Leith & Trown, 1970; Reid, Palmer, Whitlock, & Jones, 1973; Sawiris, 1966; Sutter & Reid, 1969; Traweek, 1964). The maximal structure and guidance and the noninterpersonal context of the conventional CAI task favors the individual with an introverted personality. However, as the CAI task moves toward a more discovery-type program with an increased interpersonal context, the extravert is favored. Studies investigating the relation between Jungian personality dimensions and attitude toward CAI have revealed mixed results. Howard (1986) found no significant correlations between attitude toward using CAI and these personality dimensions while Johnson (1984) reported that thinking individuals held more favorable attitudes than feeling individuals toward using CAT. Intuitive individuals were more satisfied with a CAI task that allowed a minimal amount of responsibility, while satisfaction for sensing individuals increased as the task allowed greater decision-making on the part of the individual (White & Smith, 1974). In addition, satisfaction was greater for introverted-sensing individuals than extraverted-sensing individuals. There is no evidence to indicate a relation between Jungian personality dimensions and performance reflecting the amount of CAI course material learned (Howard, 1986; Johnson, 1984). However, CAI performance differences that do not reflect course content have been noted. Sensing individuals complete the CAI portion of a program more quickly than intuitive individuals, while extravertedperceiving individuals tend to drop out of a CAI program (Hoffman & Waters, 1982). In addition, extraverts finish a task more quickly than introverts with little compromise in performance (Apter & Murgatroyd, 1975), although it is speculated that with greater task difficulty, extraverts will make more errors than introverts. Thus, although Jungian personality dimensions do not appear to be related to academic achievement in CAI-based courses, these personality dimensions may determine speed of completion or attrition rate in a CAI program. It is plausible that only anxiety characterized by a threat to self-esteem (e.g., being evaluated or compared to other individuals) significantly affects PI or CAI performance. Further, task performance appears to be either facilitated or debilitated by anxiety contingent on the qualitative nature of the task. The performance of individuals with test anxiety improves if the PI or CAI task is not accompanied by a pretest, the individual works alone, and stress is introduced into a task that involves the use of familiar material. In contrast, individuals with low test anxiety benefit from a pretest and perform better working with a parmer (Morris, Blank, McKie, & Rankine, 1970; Sutter & Reid, 1969; Tobias & Abramson, 1971). In the trait-state anxiety literature, the performance of high anxious individuals is facilitated by easier vs. difficult CAI tasks, the presence of memory support, and certain types of response modes (e.g., multiple-choice format) (Leherissey, O'Neil, & Hansen, 1971; Leherissey, O'Neil, Heinrich, & Hansen, 1973; O'Neil, Spielberger, & Hansen, 1969). In contrast, low anxious individuals perform better on more difficult CAI tasks, do not benefit from memory support, and are not affected by different types of response modes when learning familiar material, although their performance is affected by type of response mode when learning technical material. Finally, there is evidence that individuals high in achievement motivation benefit more than individuals low in achievement motivation from PI and CAI programs, although complex interactions between achievement motivation and other personality traits and factors in the learning situation are noted (Kight & Sassenrath, 1966; Reid, Palmer, Whitlock, & Jones, 1973; Tobias & Abramson, 1971).

130

P ocius

GENERAL CONCLUSIONS AND DISCUSSION

Given the previous review, four general conclusions are drawn. First, the introversion-extraversion dimension appears to be an important dimension that is related to many aspects of human-computer interaction. For instance, there is evidence that introverts and individuals with traits characterizing introversion perform better than extraverts in introductory programming courses. In addition, both computer science majors and computer professionals are almost overwhelmingly represented by introverts. This personality dimension is also related to CAI performance. Introverts perform better on conventional CAI tasks while extraverts perform better on tasks that are less conventionally defined. Further, although findings from one study suggest that extraverts may complete a CAI task more quickly than introverts, another study indicates that extraverts with a perceiving preference will have a higher attrition rate in CAI-based courses. All in all, introverts appear to be more suited to the human-computer interaction environment that is currently defined in the working world. Second, there is little support that Jungian personality dimensions (i.e., introversion-extraversion, sensing-intuitive, thinking-feeling, judging-perceiving) correlate with programming aptitude and achievement. Given the findings of studies using other introversion-extraversion measures, it is surprising that the introversionextraversion index of the MBTI does not correlate with programming aptitude and achievement. Correlations between the MBTI and instruments measuring introversion-extraversion or related traits are moderate to high (e.g., Eysenck Personality Questionnaires [introversion r = .74, p < .001], Sixteen Personality Factor Questionnaire [extraversion r = -.74, p < .001 ], California Psychological Inventory [sociability r = -.66, p < .001]) (reported in Myers & McCaulley, 1985). This lack of correlation is also reflected in the CAI literature. There is no evidence to support a relation between Jungian personality dimensions and academic achievement in CAI-based courses. However, certain personality dimensions are related to attrition rate and speed of completion of CAI-based courses. One possibility for the absence of a relationship between introversion-extraversion as measured by the MBTI and programming aptitude and achievement and CAI performance is that limited characteristics of this construct may account for performance. For instance, low sociability and impulsivity and high reflectiveness, traits found among introverts but lacking in extraverts, influence programming and CAI success. Because the MBTI assesses a much broader definition of introversion-extraversion, it may be targeting traits not relevant to programming or CAI performance. It is also plausible that investigating the interaction between certain Jungian dimensions is necessary for detecting differences in programming or CAI performance. Carlson and Levy (1973) report that introverts with thinking preferences are more effective in remembering neutral, objective, impersonal material whereas extraverts with feeling preferences are more effective in remembering novel, social, emotionally-toned stimulus material. Given the inherent nature of the programming task, which is characterized by the assimilation and manipulation of objective, impersonal information, it is speculated that an introvert with a thinking preference will perform better than an extravert with a feeling preference at a programming or CAI task. Third, although little relation between Jungian personality dimensions and programming aptitude and achievement has been found, there is a relation between

Personality factors in human-computer interaction

131

these personality dimensions and the selection of computer programming as an academic or career choice. Introverted, intuitive, and thinking individuals are highly represented in the computer science population. This is true not only for computer professionals but, at least in one available study, for computer science majors as well. One reason that individuals with particular personality traits are drawn toward specific careers is that individuals attempt to match self-perceptions with occupational stereotypes (Holland, 1985). In other words, individuals prefer occupations in which their self-perceptions match the stereotype of the typical employee. Conversely, individuals avoid occupations in which their self-perceptions and the stereotype of the typical employee are incongruent. For example, programmers express little interest in activities that demand gregariousness and extraversion or that require close personal interaction or responsibility in helping people (Barnes, 1974; Perry & Cannon, 1967; 1968). In addition, the typical "computer scientist" is stereotyped as an introvert (Matheson & Strickland, 1986). Therefore, it is not surprising that a high frequency of introverts and thinking individuals gravitate toward computer science majors and careers. Certain personality types are also highly represented in this population. Programming professionals are most commonly represented by ISTJ, INTJ, and INTP types. According to Myers (1980), the sensing-intuitive preference has the most influence on occupational choice, because it determines in large part what will interest individuals. Sensing types like "the practical application, the what and the how" and are drawn to occupations that let them deal with a constant stream of facts. Intuitive types prefer "the principle, the theory, the why" and are drawn to situations where they can look for possibilities. It would be interesting to determine whether ISTJ types are attracted to programming vocations that are qualitatively different from those programming vocations preferred by INTJ and INTP types. Descriptions of the ISTJ, INTJ, and INTP types suggest that these types may prefer different kinds of programming careers. For instance, the ISTJ type seems well-suited to the tasks found in a data-processing career. Data-processing tasks generally demand a high degree of concentration, attention to detail, and a tolerance for routine. Myers and McCaulley (1985) characterize the ISTJ type as painstaking, serious, quiet, practical, orderly, matter-of-fact, logical, realistic, dependable, and thriving on routine. ISTJ types absorb a large number of facts and are careful about details and accuracy. They are well-organized and work steadily toward accomplishment, earning success by concentration and thoroughness. The suggestive evidence that more skilled data entry operators are frequently represented by S/T/J combinations also imply that these types are particularly well-suited for data-processing careers. In contrast, INTJ and INTP types appear suited for computer science careers that involve research and theory. Both INTJ and INTP types are likely to be interested in scientific and technical matters and are often found in research-oriented careers and in higher academics. Both types are skeptical, critical, imaginative, inventive, possibilities-oriented, and dislike routine. While the INTP type tends to be better at pure theory, the INTJ type is more apt to accomplish tasks. It is speculated that the computer scientist is most commonly represented by the INTJ and INTP type. Given the current evidence, it is concluded that the Jungian personality dimensions are not an effective predictor of performance reflecting programming aptitude or achievement or of academic achievement in CAI-based courses. However, these personality dimensions may be suitable for determining whether an individual will be drawn toward a computer-related career.

132

Pocius

Finally, several other personality dimensions affecting human-computer interaction have been identified. There is some evidence that individuals with compulsive or stress-inducing personality traits do not perform well in a programming course. Instead, a certain amount of relaxation and flexibility is positively related to programming achievement. Although there is little research investigating the relation between internal-external locus of control and programming achievement, it is speculated that intemalizers may be more drawn to programming courses and to computer-related activities in general. The current literature indicates that anxiety characterized by a threat to self-esteem is related to PI and CAI performance. In addition, the relation between both anxiety and achievement motivation and task performance is affected by specific aspects of the learning task. In conclusion, understanding the effects of personality variables on computer use can be used to improve the quality of human-computer interaction. For example, acknowledging the importance of the interaction between personality traits and human-computer interaction can alert educators to individuals or groups who may be at risk for experiencing problems in human-computer interaction. In programming courses, individuals with certain compulsive or stress-inducing traits may be at risk for poorer performance than individuals without these traits. Anxious individuals may have more negative attitudes toward PI and CAI, while personality types such as the EN-P type are more likely to drop out of a CAI program than other types. Along a similar vein, there may be personality types who are more prone to experience computer anxiety. Artistic and Social vocational personality types, for example, report a greater amount of computer anxiety compared to Realistic, Investigative, Enterprising, and Conventional personality types (Bellando &Winer, 1985). Second, the awareness of which personality traits introduce individual differences in human-computer interaction can illuminate ways in which a particular human-computer interaction task can be altered to accommodate different users. For instance, not everyone benefits equally from a PI or CAI program. Although introverts tend to perform well in a PI or CAI program, extraverts perform poorly in a traditional CAI program. Altering the program to allow extraverts to work in pairs or groups or to interact via a system network may serve to increase CAI performance and reduce attrition rate. In addition, introducing more novelty into the CAI task may help to reduce the potential tediousness of the task to which extraverts are easily prone. Further, considering the level of anxiety of an individual may influence whether a CAI program should include a pretest, interpersonal context, memory support, or a required mode of responding. Third, identifying personality variables that are conducive to programming success can be useful in forecasting computer aptitude and achievement. Combining personality variables with demographic, academic background, previous computer experience, and cognitive variables to create a set of predictor variables of computer proficiency can be used to discriminate among computer science enrollment applicants. Finally, awareness of an individual's personality traits can help individuals in making better decisions about their study majors or career choice. For instance, vocational counselors may encourage an individual to pursue a computer science career should they share the interests and personality traits of individuals employed in computer science fields. Conversely, individuals considering a computer science career who do not exhibit traits conducive to programming success or express similar traits or interests with computer science professionals can be alerted to the diffi-

Personality factors in human-computer interaction

133

culties they may encounter, and can be directed toward alternative careers that appeal to the individual, or introduced to strategies for dealing with these potential difficulties.
Acknowledgment - - I would like to thank the referees for their comments on this paper. The detailed and invaluable suggestions that were made have greatly improved the quality of the paper.

REFERENCES
Alspaugh, C. A. (1972). Identification of some components of computer programming aptitude. Journal for Research in Mathematics Education, 3, 89-98. Anastasi, A. (1976). Psychological testing. New York: Macmillan Publishing Co., Inc. Apter, M. J., & Murgatroyd, S. (1975). Concentration, personality and self-pacing in programmed learning. Programmed Learning and Educational Technology, 12,208-215. Barnes, P. H. (1974). A study of personality characteristics of selected computer programmers and computer programmer trainees (Doctoral dissertation, Auburn University, 1974). Dissertation Abstracts International, 35, 1440A. Bellando, J., & Winer, J. L. (1985, April). Computer anxiety: Relationship to math anxiety and Holland types. Paper presented at the Annual Convention of the Southwestern Psychological Association, Austin, TX. Bruning, J. L., & Kintz, B. L. (1987). Computational handbook of statistics. Glenview, Illinois: Scott, Foresman and Company. Bush, C. M., & Schkade, L. L. (1985, March). In search of the perfect programmer. Datamation, 31, 128-132. Butcher, D. F., & Muth, W. A. (1985). Predicting performance in an introductory computer science course. Communications of the ACM, 3, 263-268. Calaway, F. W. (1982). An exploration of computer aptitude (Doctoral dissertation, Memphis State University, 1982). Dissertation Abstracts International, 43, 2280A. Campbell, P. E, & McCabe, G. P. (1984). Predicting the success of freshman in a computer science major. Communications of the ACM, 27, 1108-1113. Card, S. K., Moran, T. P., & Newell, A. (1986). The psychology of human-computer interaction. HiUsdale, NJ: Erlbaum. Carlson, R., & Levy, N. (1973). Studies of Jungian typology: I. Memory, social perception, and social action. Journal of Personality, 41,559-576. Cattell, R. B., Eber, H. W., & Tatsuoka, M. M. (1970). Handbook for the Sixteen Personality Factor Questionnaire (16PF). Champaign, IL: Institute for Personality and Ability Testing. Corman, L. S. (1986). Cognitive style, personality type, and learning ability as factors in predicting the success of the beginning programming student. ACM SIGCSE Bulletin, 18, 80--89. Dallos, R. (1975). Programmed learning and personality: A review and a preliminary study. Programmed Learning and Educational Technology, 12, 12-20. Denham, J. (1979). Use of the Myers-Briggs Type Indicator as a predictive discriminator between high- and low-speed operators of data entry equipment. Australian Psychologist, 14, 207. Doty, B. A., & Doty, L. A. (1964). Programmed instructional effectiveness in relation to certain student characteristics. Journal of Educational Psychology, 55, 334-338. Evans, G. E., & Simkin, M. G. (1989). What best predicts computer proficiency? Communications of the ACM, 32, 1322-1327. Eysenck, H. J. (1964). Principles and methods of personality description, classification, and diagnosis. British Journal of Psychology, 55, 284--294. Fowler, G. C., & Glorfeld, L. W. (1981). Predicting aptitude in introductory computing: A classification model. AEDS Journal, 96-109. Goldstein, K. M., & Blackman, S. (1978). Cognitive style. New York: John W~fley& Sons, Inc. Hoffman, J. L., & Waters, K. (1982). Some effects of student personality on success with computerassisted instruction. Educational Technology, 20-21. Holland, J. L. (1985). Making vocational choices: A theory of vocational personalities & work environments. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. Hostetler, T. R. (1983). Predicting student success in an introductory programming course. ACM SIGCSE Bulletin, 15, 4ff-43.

134

P ocius

Howard, M. J. (1986). The impact of personality characteristics and attitude on learning by computerassisted instruction (Doctoral dissertation, Florida State University, 1986). Dissertation Abstracts International, 47, 801A. Johnson, J. S. (1984). A comparison of student performance on computer-assisted and traditional testing in a baccalaureate nursing program (Doctoral dissertation, University of Kansas, 1984). Dissertation Abstracts International, 46, 1512B. Jung, C. G. (1971). Psychological types (H. G. Baynes, Trans. revised by R. E C. Hull). Volume 6 of The collected works of C. G. Jung. Princeton, NJ: Princeton University Press. (Original work published in 1921) Kagan, D. M., & Douthat, J. M. (1985). Personality and learning FORTRAN. International Journal of Man-Machine Studies, 22, 395-402. Kagan, D. M., & Pietron, L. R. (1986). Aptitude for computer literacy. International Journal of ManMachine Studies, 25,685-696. Keppel, G. (1982). Design and analysis: A researcher's handbook. Englewood Cliffs, NJ: PrenticeHall, Inc. Kern, G. M., & Matta, K. E (1987). Leaming style as an influence on the effectiveness of self-paced computer-assisted instruction: Preliminary results. Proceedings of the 9th Annual Conference on Computers Industrial Engineering, 13, 203-207. Kight, H. R., & Sassenrath, J. M. (1966). Relation of achievement motivation and test anxiety to performance in programmed instruction. Journal of Educational Psychology, 57, 14-17. Konvalina, J., Stephens, L., & Wileman, S. (1983). Identifying factors influencing computer science aptitude and achievement. AEDS Journal, 106-112. Lee, T. W. (1985). The effects of library-based computer assisted instruction on student programming performance and an implementation of the grading support system in FORTRAN language courses at the university (Doctoral dissertation, Florida Institute of Technology, 1985). Dissertation Abstracts International, 46, 2648A. Leherissey, B. L., O'Neil, H. E, Jr., & Hansen, D. N. (1971). Effects of memory support on state anxiety and performance in computer-assisted leaming. Journal of Educational Psychology, 62, 413-420. Leherissey, B. L., O'Neil, H. F., Jr., Heinrich, D. L., & Hansen, D. N. (1973). Effect of anxiety, response mode, subject matter familiarity, and program length on achievement in computerassisted learning. Journal of Educational Psychology, 64, 310-324. Leith, G. O. M., & Trown, E. A. (1970). The influence of personality and task conditions on learning and transfer. Programmed Learning and Educational Technology, 7, 181-188. Lyons, M. L. (1985, August). The DP psyche. Datamation, 31, 103-110. Martin, M. P., & Fuerst, W. L. (1987). Using computer knowledge in the design of interactive systems. International Journal of Man-Machine Studies, 26,333-342. Matheson, K., & Strickland, L. (1986). The stereotype of the computer scientist. Canadian Journal of Behavioral Science, 18, 15-24. Mazlack, L. J. (1980). Identifying potential to acquire programming skill. Communications of the ACM, 23, 14-17. Morris, V. A., Blank, S. S., McKie, D., & Rankine, E C. (1970). Motivation, step size and selected learner variables in relation to performance with programmed instruction. Programmed Learning and Educational Technology, 7, 257-267. Morrison, P. R., & Noble, G. (1987). Individual differences and ergonomic factors in performance on a videotex-type task. Behaviour and Information Technology, 6, 69-88. Myers, I. B., & McCaulley, M. H. (1985). Manual: A guide to the development and use of the MyersBriggs Type Indicator. Palo Alto, CA: Consulting Psychologists Press. Myers, I. B., & Myers, P. B. (1980). Gifts differing. Palo Alto, CA: Consulting Psychologists Press. Newsted, P. R. (1975). Grade and ability predictions in an introductory programming course. ACM SIGCSE Bulletin, 7, 87-91. Nowaczyk, R. H. (1984). The relationship of problem-solving ability and course performance among novice programmers. International Journal of Man-Machine Studies, 21, 149-160. O'Neil, H. F., Jr. (1972). Effects of stress on state anxiety and performance in computer-assisted learning. Journal of Educational Psychology, 63,473-481. O'Neil, H. E, Jr., Spielberger, C. D., & Hansen, D. N. (1969). Effects of state anxiety and task difficulty on computer-assisted learning. Journal of Educational Psychology, 60, 343-350. Pardes, M. (1965). Eysenck's introversion-extraversion, boredom, and time estimation (Doctoral dissertation, Columbia University, 1965). Dissertation Abstracts International, 26, 3489. Perry, D. K., & Cannon, W. M. (1967). Vocational interests of computer programmers. Journal of Applied Psychology, 51, 28-34.

Personality factors in human-computer interaction

135

perry, D. K., & Cannon, W. M. (1968). Vocational interests of female computer programmers. Journal of Applied Psychology, 52, 31-35. Petersen, C. G., & Howe, T. G. (1979). Predicting academic success in introduction to computers. AEDS Journal, 182-191. Reid, J. B., Palmer, R. L., Whitlock, J., & Jones, J. (1973). Computer-assisted instruction performance of student pairs as related to individual differences. Journal of Educational Psychology, 65, 65-73. Rich, E. (1983). Users are individuals: Individualizing user models. International Journal of ManMachine Studies, 18, 199-214. Ripple, R. O., O'Reilly, R., Wightman, L., & Dacey, J. (1965). Programmed instruction and learner characteristics: Preliminary data. Psychological Reports, 17,633-634. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80 (1, Whole No. 609). Sawiris, M. Y. (1966). A factorial study of some variables relevant to a programmed learning situation. Programmed Learning and Educational Technology, 3, 30--34. Schneiderman, B. (1980). Software psychology: Human factors in computer and information systems. New York: Winthrop. Sharma, S. (1986-87). Learners' cognitive styles and psychological types as intervening variables influencing performance in computer science courses. Journal of Educational Technology Systems, 15, 391-399. Sitton, S., & Chmelir, G. (1984, October). The intuitive computer programmer. Datamation, 30, 137-141. Spence, K. W. (1958). A theory of emotional based drive (D) and its relationship to performance in simple learning situations. American Psychologist, 13, 131-141. Spielberger, C. D. (1971). Anxiety as an emotional state. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research. New York: Academic Press. Sutter, E. G., & Reid, J. B. (1969). Learner variables and interpersonal conditions in computerassisted instruction. Journal of Educational Psychology, 60, 153-157. Taylor, J. A. (1956). Drive theory and manifest anxiety. Psychological Bulletin, 53, 303-320. Tobias, S. (1973). Distraction, response mode, anxiety, and achievement in computer-assisted instruction. Journal of Educational Psychology, 65, 233-237. Tobias, S., & Abramson, T. (1971). Interaction among anxiety, stress, response mode, and familiarity of subject matter on achievement from programmed instruction. Journal of Educational Psychology, 62,357-364. Traweek, M. W. (1964). The relationship between certain personality variables and achievement through programmed instruction. California Journal of Educational Research, 15, 215-220. Van Muylwijk, B., Van Der Veer, G., & Waern, Y. (1983). On the implications of user variability in open systems. An overview of the little we know and the lot we have to find out. Behaviour and Information Technology, 3, 313-326. Van Der Veer, G. C., Tauber, M. J., Waem, Y., & Van Muylwijk, B. (1985). On the interaction between system and user characteristics. Behaviour and Information Technology, 4, 289-308. Weinberg, G. M. (1971). The psychology of computer programming. New York: Van Nostrand Reinhold. Werth, L. H. (1986). Predicting student performance in a beginning computer science class. ACM SIGCSE Bulletin, 18, 138-143. Wesley, B. E., Krockover, G. H., & Hicks, C. R. (1985). Locus of control and acquisition of computer literacy. Journal of Computer-Based Instruction, 12, 12-16. Whipkey, K. L., & Stephens, J. T. (1984). Identifying predictors of programming skill. ACM SIGCSE Bulletin, 16, 36-42. White, A. L., & Smith, D. D. (1974). A study of the effects of varying student responsibility for instructional decisions in a CAI course. Journal of Computer-Based Instruction, 1, 13-21. Wine, J. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104.

You might also like