University of Southern California Dissertations and Theses

Empathy And Social Perception

(USC Thesis Other)

Empathy And Social Perception

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content EMPATHY AND SOCIAL PERCEPTION by -f- ^ Forrest K. Strayer A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Psychology) June i960 UNIVERSITY O F S O U TH ER N CALIFORNIA G R A D U A T E SC H O O L U N IV E R SIT Y PARK LOS A N G E L E S 7 , C A L IF O R N IA Th„s dissertation, written by ?orrest._K,_..Strajer................ under the direction of his-.-Dissertation Com mittee, and approved by all its members, has been presented to and accepted by the Graduate School, in partial fulfillment of requirements for the degree of D O C T O R OF P H IL O S O P H Y Dean D ate.. DIS! /IITTEE fdt... Chairman TABLE OP CONTENTS Page LIST OF TABLES v LIST OF FIGURES vi Chapter I. THE PROBLEM AND BACKGROUND 1 Statement of the Problem Background Early studies of empathy The assumed similarity approach to interpersonal perception The Empathy Test Later studies of social perception The critical articles Recent trends In social perception research Re-statement of the Problem Experimental Hypotheses Experimental Subjects Experimental Measures Social perception measures Assumed similarity measures The aptitude tests The educational development (achieve ment) tests The Vocational Interest Test The job satisfaction scale The Social Situations Test The sociometric "popularity” score The Task Check List "amount of super vision" scores The combined mean-peer and supervisor job proficiency ratings The combined self and supervisor task leadership ratings Summary of the experimental measures II. METHOD k3 ii Chapter Page Data Collection Data Analysis The reliability of the measures and other preliminary statistics The basic intercorrelation matrix The multiple regression analysis Further correlational studies III. RESULTS..................................... 7S The Reliability of the Experimental Measures The Equivalence of Measures Used to Define Social Perception Operationally Hypothesi s I Results of the test of Hypothesis I Discussion of the results of the test of Hypothesis I The Equivalence of Measures Used to Define Assumed Similarity Operationally Hypothesis II Results of the test of Hypothesis II Discussion of the results of the test of Hypothesis II The Relations between Social Perception and Assumed Similarity Measures and Measures of Job Satisfaction and Job Performance Hypothesis III Results of the test of Hypothesis III Discussion of the results of the test of Hypothesis III The Predictive Validity of Social Percep tion and Assumed Similarity Measures for the Job Performance Criterion as Compared with the Predictive Validity of Conventional Predictors for that Criterion Hypothesis IV Results of the test of Hypothesis IV Discussion of the results of the test of Hypothesis IV Some Nonhypothesized Relations between the Social Perception and Assumed Similarity Measures, Sociometric Popularity, Job Satisfaction, and Job Performance HI Chapter Page Relations among social perception and assumed similarity scores derived from a common base Relations among social perception and assumed similarity scores not derived from a common base Some nonhypothesized relations between sociometric popularity and job performance IV. SUMMARY AND CONCLUSIONS ................... 138 REFERENCES li+8 APPENDIX A Basic Intercorrelation Matrix .... l$k APPENDIX B Cronbach’s Analysis of Variance Model for Social Perception Scores ...» 156 APPENDIX C Data Collection Instruments........... 160 iv LIST OP TABLES Table Pag© I. Reliability of the Experimental Measures • . 77 II. Intercorrelations among Measures of Social Perception......................... 87 III. The Means and Standard Deviations of the Accurate and Inaccurate Groups on the JPP-raw D and JPP-Acc Scores......... 90 IV. Intercorrelations among AS Measures ..... 95 V. Relations between Social Perception and Assumed Similarity Measures and Measures of Job Satisfaction and Job Performance . . . 107 VI. Beta Weights and Per Cents of JPC Variance Accounted for by Selected SP, AS, and Aptitude Predictors ............ lllj. VII. Means and Standard Deviations of Scores of Subjects In This Sample and Unselected Basic Airmen on Nine Air Force Tests . . . Il6 VIII. Intercorrelations among Aptitude, Achieve ment, and Interest Tests, and the JPC • . • 118 IX. Intercorrelations between Selected Social Perception and Assumed Similarity Scores not Derived from a Common Profile..... 131 X. Beta Weights and Per Cents of JPC Variance Accounted for by Selected SP, AS, and Aptitude Scores, Plus the Sociometric Popularity Scores ................. 13^ v LIST OP FIGURES Figure Page 1, Weights for Sociometric Accuracy Scores . • • 50 2. Weights for Sociometric AS Scores 5& vi CHAPTER I THE PROBLEM AND BACKGROUND The current Interest In the concepts of empathy and social perception began about ten years ago* This interest coincided with the promulgation of Sullivan’s "interpersonal theory of psychiatry," with the "new look" in perception, with the increasing attempts by the Rogers’ group to combine research with psychotherapy, with the advocation by many clinicians of the phenomenological point of view, and with the general rise of clinical psychology as a profession. The early studies of empathy asked such questions as, What is the relation between empathy (awareness of other’s thoughts and feelings) and insight (awareness of one’s own thoughts and feelings)? and, Does the good therapist see himself as similar to or different from his patients? After the concepts of empathy and social perception had been operationally defined and some correlations with other psychological concepts had been obtained, the studies became more critical. Questions were raised about the generality of accuracy (or verld- icality) in social perception, about how to separate 2 accuracy from projection (both, person A and person B privately express the opinion that politics cannot be separated from religion; person A is asked to predict person B fs attitude on the subject, and he predicts that B*s attitude Is the same a3 his: is his prediction based on an accurate perception or is it primarily a projection of his own attitude?), and about artifacts in some of the correlations already observed* As a result of the criticisms, most psychologists investigating social perception changed the direction of their Investigations. They became Interested In how social perceptions were made, and In how they differed from non-social perceptions, instead of In accuracy-of- predlction scores and their correlates. But the change in direction was too abrupt, and It was not based on experimental findings. Further, accuracy-of-predictlon scores remained of necessity the primary Independent variable In moat of the studies of the "process” of social perception, because those scores operationally defined the process being investigated. Statement of the Problem Two kinds of scores have been used to operationally define the related concepts of empathy and social 3 perception. Most often, investigators have used accuracy-of-predictlon of other persons1 responses as the measure of Individual differences on the variable social perception. The usual operation was to have a group of subjects respond to the items of some attitude or personality questionnaire. Then each subject was asked to predict the responses made by one or more of the other subjects. Accuracy was measured by the agreement between predicted responses and actual responses. The second kind of operational definition of social perception was the assumed-similarity score. This score was obtained by the same kind of operation used to obtain the accuracy-of-predictlon score. Subjects first described themselves on a questionnaire, and then they predicted how one or more other subjects would describe themselves on the same questionnaire. Assumed-similarity was measured by the agreement between the subjects* self descriptions and the self-descriptions they predicted for "Social perception" Is more widely used at present as the name for the concept that would account for individual differences In people*s ability to Infer from observable cues what other persons may be thinking or feeling or what other persons may be about to do. "Empathy" was used as the name for this concept In most of the early studies done, and it has a connotation that "social perception" does not have. "Empathy" usually connotes some kind of sharing of the phenomenological field of the other person, which would Imply a more direct and Immediate knowledge of the other person*s thoughts, feelings, or Intentions. the other person(s). If a subject predicted that another person would describe himself in much the same way that the subject had described himself, then the investigators who used these scores inferred that the subject "assumed similarity" between himself and the person whose responses he predicted. The accuracy-of-prediction and assumed-similarity scores were both u3ed as operational definitions of social perception. The Investigators who used them assumed that they represented a variable that was called "empathy," "social perception," or "Interpersonal perception." But except for two Investigators who examined the relation between two kinds of acouracy-of-prediction scores (Bell and Hall: 195^)* no attempt was made to find out If aocuracy-of-prediction under one condition was correlated with accuracy-of-predictlon under other conditions. Ho Investigator correlated assumed-similarity scores with accuracy-of-prediction scores, and only Fiedler (1958) indicated that any attempt had been made to determine the relations among assumed-similarity scores obtained under different conditions* Because both accuracy-of-prediction and assumed- similarity scores had continued to be used as independent variables In studies of social perception (cf. Tagiuri and Petrullo, 1958), and because these scores might have 5 practical value as predictors of other behavior variables (cf. Fiedler, 195^5 Van Zelst, 1952; Gage, 1952), the writer felt that both accuracy-of-prediction and assumed- similarity scores, as well as scores on other test and behavior variables, should be obtained from a single group of experimental subjects* The relations among the social perception scores could then be determined with some precision, and some of their relations with other test scores and with criteria of performance could also be investigated* Four things were attempted in the dissertation study reported here, within the limits imposed by the experimental subjects and measures used! 1* To determine the equivalence of various kinds of operations used to measure accuracy of social perception, 2. To determine the equivalence of various kinds of operations used to measure the tendency to assume similarity between oneself and others* 3* To determine some of the correlates of accuracy of social perception scores. i j . . To determine some of the correlates of assumed- similarity scores. Background The review of the literature is divided into six part3. Following a brief introduction, the early studies of empathy are described* Then Fiedler’s continuing work with assumed-similarity measures is presented. Next, various studies of social perception performed in the early 1950's described. Then Cronbach^s critical articles are discussed in some detail. Finally, brief mention is given to recent trends in social perception research. Early studies of empathy Empathy had long been felt to be an important psychological variable, but few facts about it were known. The tendency of spectators at athletic events to duck with the fighter, to lift themselves with the pole vaulter, to strain towards the tape with the sprinter led some authors to infer that empathy is a kind of "kinesthetic inference" (G. W. Allport, 1937s P« 532)9 but only superficial evidence was presented to help support this conclusion. The modern approach to the measurement of empathy or social perception probably began with the work of Steinmetz (19^5)• In an article dealing with various techniques to be used in "directive psychotherapy," he gave case studies of methods that appeared to b© based on the same hypotheses that guided more recent investigations of "assumed-similarity" and "empathetlc-accuracy-of- predictlon" scores. He said: The measurement of specific psychological percep tion Is the approach to the ability one should expect of the clinical psychologist and psychiatrist. I refer to the ability to respond in a test or rating 7 scale of psvchopathological traits as other persons might, (1945s P» 334) Very little could be concluded about the value of Steinmetz*s methods, since the few case studies he presented could only be considered illustrative; but he evidently was the pioneer in using differences between p self and predicted-Other scores on tests like the Guilford-Martin Personnel Inventory to measure what he called "psychological understanding,1 1 The first of the modem investigators of empathy as & psychological trait or ability was Rosalind Dymond, She became interested in it while doing a study (1949) of "insight,1 1 Her subjects gave personal histories and took the TAT, She assumed that Interpersonal relations described in the TAT stories were projections of experi enced self-Other relations, and so she predicted from each subject's TAT stories what kind of relationships the subjects had with the important people in their lives. To explain why some of her subjects denied the real-world relationships predicted for them, she hypothesized that they might be deficient in 1 1 ’role-taking' or empathic ability," She assumed that empathy would result in a subject's giving "richer" characterizations in his TAT ^Cronbach's (1955) convention of referring to the person whose responses are being predicted as the "Other" will be observed here. stories, and thus by rating as good, fair, or poor the richness of the descriptions accorded to the pictured characters In each story, she came up with an index of empathy: the per cent of the characters which fell Into the "good” category* She found that the subjects whose TAT characterizations were the least rich were responsible for 81^ of the denied relationships, and so she concluded, "It would seem that the ability to feel and describe the thoughts and feelings of others, (empathy) is accompanied by a better understanding of the relationships one has with others, (insight)" (19i j - 8: p, 232). In her next paper, Dymond described a scale for measuring empathic ability. The scale consisted of six traitss self-confidence; superior-inferior, selfish- unselfish; friendly-unfriendly; leader-follower; and sense of humor. Each subject rated himself and the members of his discussion group on the traits, and then each subject had to predict how each Other would (1) rate himself and (2) rate the predicting subject. Empathic ability was taken to be the degree of correctness In predicting the Other'3 rating of the self (predictor) and the Other's rating of himself, with the self ratings and the ratings of Others by the self providing the standards of correctness. In her study with the members of eight student groups Dymond found that empathy scores on her 9 experimental scale were related to the degree of empathy shown in telling TAT stories (as defined above). She reported only the test-retest reliability of the self- ratings, which was low: .60; she said nothing of the reliability of the predictive accuracy scores* Dymond used essentially the same rating technique in her next study (1950) except that each subject made predictive ratings for one other subject only in the group of which he was a member* She reported that women were somewhat better predictors than men (p<: *03), and that women were more easily predicted than men* She again reported a positive correlation between her TAT empathy index and accuracy of prediction on her empathy scale* She also studied some of the personality characteristics of the six most accurate predictors and the seven least accurate predictors, using the Wechsler-Bellevue (Form I), the Rorschach, the TAT, and the California F (ethno- centrism) scale. Her preliminary conclusions were too lengthy to reproduce here, but her good empathizers evidently had a monopoly on all the "good" traits (warm, optimistic, flexible, and so on) whereas the poor empathizers were almost unremittingly "bad" (rigid, introverted, emotionally uncontrolled, self-eentered, suspicious, and so on). 10 The assumed similarity approach to interpersonal perception Fiedler began his work with interpersonal percep tion about the time Dymond became interested in the relation between empathy and insight, but instead of working with accuracy-of-perception scores he worked with what he called "assumed-similarity" (AS) scores. He asked his subjects to describe themselves on conventional questionnaires or rating scales, and then to describe some other subject or group of subjects using the same test form. The difference between the self-description and the Other-description(s) was the measure of ASi the smaller the difference, the greater the tendency to assume similarity. AS was usually measured with Cronbach and Gleser’s ”Dn statistic (1953)* Fiedler’s first finding of importance (19^0) was that psychotherapists rated "good" by their supervisor(s) assumed more similarity between themselves and their patients than did therapists rated "poor." That is, therapists who predicted that their patients would describe themselves on a questionnaire much a3 the therapists had described themselves on the same question naire were more likely to be rated "good" than therapists who predicted their patients would describe themselves differently from the therapists* self-descriptions. 11 Following two more "clinical” studies concerned with patients’ and therapists’ attitudes as reflected by AS scores (1951* 3.952), Fiedler started a series of studies to determine the general usefulness and meaning of AS scores. He first demonstrated that subjects tended to assume more similarity between themselves and persons they liked than between themselves and persons they did not like (1952). Davitz corroborated this finding (1955)* Fiedler then (195^) hypothesized that groups with high AS scores (a ”high” AS score meant a small difference between a self-description and a predicted self-description of an Other) would be more cohesive and therefore more effective than groups with low AS scores. He used basketball teams to test this hypothesis. He asked each team member to describe himself, the team member he could cooperate with best, and the team member he could cooper ate with least well on a 100-item forced-choice questionnaire. The difference between the self description and the predicted self-description of the team member that could be cooperated with best was labeled "ASp,” where the "p" stood for "positive.” The difference between the self-description and the predicted self description of the team member that could be cooperated with least well was labeled "ASn," where the "n" stood for "negative." The difference between the predicted 12 self-descriptions for the team members that could be cooperated with best and least well was labeled "ASo," where "ow stood for "opposites.n Fiedler found that the ASo score of the most popular member of each team was related to the team’s effectiveness (measured by number of games won). The most popular members of the better teams saw lea3 similarity between the men they could and could not cooperate with than did the most popular members of the poorer teams; i.e., the ASo scores of the winning teams mo3t popular members were larger (the distance between the profiles predicted for most and least cooperative team members vas greater) than were the ASo scores of the losing teams most popular members. In recent years Fiedler continued to investigate the relation between the ASo scores of preferred co-workers and group effectiveness, and he learned that the relation was by no means a simple one (1958). These studies did indicate, however, that depending on whether a group member was a formal leader or informal leader and depending on his sociometric status, his ASo score was predictive of the work-effectiveness of his group. In his most recently published study, Fiedler (1959) reported on the relation of individual AS scores and group AS scores (the average similarity subjects assumed to other members of their group, and the average AS of the group members toward the individual subjects) to certain criteria of psychological adjustment and to changes in adjustment. He found that the average tendency of an individual to see himself as similar to other members in a group and the average tendency of the group members to see the individual as similar to them was related not only to adjustment but to change in adjustment while the individual was a member of the group. The Empathy Test In 19^1 Kerr and Speroff published their Empathy Test. In its first form, which has been used in all published research to date, The Ehipathy Test required examinees to rank l i j . music forms according to their popularity with factory workers, to rank 1£ magazines in terms of paid circulation, and to rank 10 annoying experiences as persons to 39 years of age would rank them. "Empathy” was the extent of agreement between guesses and poll findings or published figures. Van Zelst (1952) reported that empathy on one part of the to3ts correlated on the average about .65 with empathy on other parts of the test. These correlations indicated that the ability to gue3s group tendencies or tastes had some generality. Several "validation" studies of The Ekapathy Test were published* In their manual for the teat, Kerr and Speroff (1951) indicated that scores on it correlated •3I 4 . with sociometric popularity among college men, .32 and .62 with scores on the How Supervise? test, and *44 with sales records of automobile salesmen* They also report© . test-retest reliability coefficients ranging from .67 to •81* Van Zelst (1952) reported that Empathy Test scores of business agents for AFL building trades unions correlated *67 with ratings of the agents by their superiors on leadership ability, *38 with the percentage of votes received by the agents in their most recent election, *55 with How Supervise? scores, *64 with the per cent Increase in new members the agents were respons ible for, *44 with the agents' ability to settle grievances (records were rated by three judges), and *44 with the agents' records In enforcing rules and regula tions. Using Sknpathy Test scores as the dependent variable, Van Zelst found a multiple correlation of .76 between it and the above six criteria. He also reported that the corrected odd-even internal consistency coefficient computed for his sample was .87* In another study, Van Zelst (1953) found a significant correlation between Empathy Test and job-satisfaction scores* Using leadership rankings among five-member groups 15 as the criterion, Bell and Hall (195^) obtained a validity coefficient of .25 (significant at the .01 level) for Empathy Test scores. Generally speaking, therefore, some relationship was demonstrated between ability to guess group standards and various criteria of leadership. This conclusion was supported by the finding of Chowdry and Newcomb (1952) in another setting (The Empathy Te3t was not used) that ability to estimate group opinions on group-relevant material was related to group leadership. In further validation studies of The Empathy Test, Bell and Stolper (1955) found no relation between scores on it and scores on the Sensitivity to Other Persons test, but Speroff (1953) found a significant relation between scores on it and accident-proneness when the sociometric popularity of the workers studied was held constant. Jarrard (195&) in reviewing SperoffTs work, said that Speroff did not find significant differences between Empathy Test scores for accident-prone and non-accident- prone groups, but Speroff himself said: The results reveal that the group with the high Interpersonal Desirability Values [Speroff did not Indicate exactly how these were obtained, but evi dently they came from some sort of sociometric ratings 3 and the low accident records Is superior to the group with the low Interpersonal Desirability Values and high accident records in their ability to empathize or respond to others as called for in the test. (1953: P* 300) Unfortunately Speroff did not indicate the correlations l6 between popularity and accident-proneness, or between popularity and empathy. Considering the confusion arising from this study, and considering a statement Jarrard quoted from Bell and Stolper, ”... this attempt at validating [italics added] The Empathy Test was not successful,” (19^6: p. 160), there was some reason to doubt the thoroughness with which the null hypothesis was tested in some of the studies reported. Later studies of social perception In 1952 Gage performed a most interesting 3tudy of interpersonal perception. His experimental "judges" watched six "strangers” singly perform several "expressive movements." After each stranger was finished he took part of the Kuder Preference Record. At the same time the judges filled out that part of the Kuder as they predicted the stranger would. Subsequently, the judges responded to the Kuder Items as they predicted each of three of their classmates would. Each of the judges also took the Kuder himself and predicted how the typical classmate ("stereotype") of his sex would respond to the Kuder items. Gage was interested primarily In the generality of the accuracy-of-prediction scores. He wanted to know if subjects who were accurate in predicting the responses of 17 one stranger would also be accurate In predicting the responses of another. He also wanted to know if the subjects who were accurate in predicting the responses of strangers were equally accurate in predicting the responses of their classmates. His results demonstrated three kinds of generality: 1. Over items in the Kuder. Accuracy in predict ing a response to one item was substantially related to accuracy in predicting a response to another item. The average corrected correlation coefficient was .70. This coeffi cient gave an indication of the reliability of the total accuracy scores across all items used in the Kuder. 2. Over strangers. On the average, the judge who was accurate in predicting one stranger's response was also accurate in predicting other strangers' responses. The average corrected correlation coefficient between accuracy scores for two groups of three strangers was .q.9. 3. Between strangers and classmates. The judge who was accurate in predicting strangers' responses on the average was also accurate in predicting his classmates' responses on the average. The correlation between accuracy in predicting strangers' responses and accuracy in predicting classmates' responses was .71* Gage concluded from these results that there was consider able generality in ability to predict Others' responses accurately. Gage next examined the nature of the accuracy scores he had obtained. The differential accuracy scores represented the sum of the itera-by-itera differences between the judges* predictions for strangers and the strangers1 actual responses. Gage also had the Judges take the Kuder themselves and also respond to It as the typical classmate (stereotype) of their sex would. From all these administrations of the Kuder, Gage could compute six separate difference scores. One of these, the difference between the judges * predicted responses for strangers and the strangers' actual responses (the differ ential accuracy score) has already been discussed* Two more of the difference scores will be considered here! the difference between the stereotype predictions and the strangers' responses (stereotype accuracy), and the difference between the self responses of the judges and the responses they predicted for the strangers (assumed similarity)• Gage found that the stereotype accuracy scores were more closely related to the individual stranger's responses on the average than were the differential accuracy scores. If each judge's stereotype predictions had been used to predict each stranger's responses rather than the differ ential predictions the judge made for each stranger separately, the judge would have been more accurate on the average in his predictions of how each stranger would respond. This result was obtained even though it was the stereotype of the high school "classmate' 1 rather than the college "stranger" that was used. Gage did not inquire 19 into this unexpected finding, but the stereotype predic tions may have conformed more to modal response tendencies— if such tendencies exist on the Kuder— than the differential predictions. The judges may have erred in the direction of trying to discriminate too much among the strangers in making their differential predictions. Gage also compared the differential accuracy and stereotype accuracy scores plus the assumed-similarity scores in their average relations with six sociometric indexes.3 The average correlation of the differential accuracy scores with all six indexes was .37; the average correlation of the stereotype accuracy scores with the indexes was . l j . 1; and the average correlation for the assumed-similarity scores was .33* Gage did not, however, report the correlations between the two accuracy scores and the assumed-similarity scores. This was regrettable, because much of the later confusion (cf. Cronbach, 1955) as to the meaning of accuracy-of-prediction and assumed- similarity scores might have been avoided if Gage had investigated the interrelations and possible interdepend- icies among the various scores he obtained. He did ^The judges were asked to pick the three boys and three girls in their class that were (1) their closest friends, (2) their preferred study companions, (3) their preferred social companions, ( i j . ) the quickest in catching on to other students1 feelings, ($) the most tactful, and (6) the quickest at sizing up a newcomer. 20 demonstrate, however, that reliable individual differences existed in ability to predict test responses, and that this ability was correlated with various sociometric ratings and was not correlated with intelligence test or interest test scores. In the following year, Gage (1953) presented an analysis of the interpersonal perception situation and reported the findings of another study. He suggested that four components were involved in social perception: the perceiver, the Other, the input (the stimulus cues the perceiver had to work with), and the outtake (tbs judgment the perceiver was expected to make). In the experiment he reported, he was primarily interested in how his subjects used the input. The study was concerned with the accuracy of subjects in perceiving the responses of several Others. All subjects had equal experience with the Others, and so input was controlled. Gage made use of the fact that on some Items of the questionnaire he used the majority of the subjects were either accurate or Inaccurate In their predictions. That Is, on particular items most subjects would correctly predict the responses of the Others whereas on other Items most subjects would be Incorrect in their predictions. In evaluating the accuracy of Individual subjects in comparison with these group trends, Gage made two Inferences: 21 1. Accuracy In perceiving manifest stimulus values. If, after Input, a subject agreed with the group’s average prediction and was accurate, Gage inferred that the subject’s accuracy was based on the input. That is, the subject and the rest of the group evidently made their accurate prediction on the basis of shared input information. 2. Accuracy in taking the role of the other. If, after input, a subject did not agree with the group prediction and was accurate, then Gage inferred that the subject had resisted the stereotypical implications of the input so his accuracy must have been based on something other than the input. Gage found internal consistency for these two kinds of accuracy, and he found that they were negatively corre lated. Hastorf and Bender (1952) did the first critical studies in the area of social perception. They questioned the effect of projection on accuracy of social perception scores. They reasoned that if both the subject and the Other responded similarly to test items, then if the subject wa3 accurate in his predictions he might merely have been projecting his own attitudes rather than inferring or intuiting the attitudes of the Other. In the situation where a subject takes a test, predicts how some Other subject will answer the test items, and the Others’ answers are available for compari son, Bender and Hastorf (1953) suggested that two scores be derived: the conventional accuracy score (the difference between the predicted responses for the Other and the Others1 actual responses); and a "projection” score (the difference between the subjects’ own responses and the responses predicted for the Others)* There was nothing really new here, for the latter is the assumed- similarity score. But Hastorf and Bender evidently felt that the tendency to assume similarity was, if not equal to unabashed projection, at least loaded on the projection factor. So they considered it to be an error component that should be suhtracted out of the accuracy score. They did this by comparing the projection 3core and accuracy (or "raw empathy") score on each item of the test instru ment. Where the projection score was smaller (where the subject’s response and the response he predicted for the Other were the same or similar), a negative "refined empathy" score resulted. Where the projection score was larger than the accuracy score, a positive refined empathy score resulted. The total refined empathy score was the algebraic sum of the positive and negative refined empathy item scores. Bender and Hastorf found that the total refined empathy scores correlated negatively with both the actual similarity between subjects and Others and with the projection (or AS) scores. It was later pointed out by Gage and Cronbach (1955) that these correlations were 23 artifactual. In the same study (1953)# Bender and Hastorf observed a relation that deserves mention* They had f>0 subjects predict how four of their associates would respond to i\.2 temperament items* They correlated the projection (AS) scores for all six subject-associate pairs. The correlation coefficients ranged from *ij.5 to *73* The tendency to assume similarity was, In the authors* words, "dramatically consistent*" But dramat ically consistent or not, they chose to ignore It and continued to view projection as an error-laden component. In their final experimental study, Hastorf and Bender, along with Weintraub (1955)# found little or no correlation between total refined empathy scores and other variables, but they noticed that subjects with high total refined empathy scores tended to respond primarily with "like" or "dislike" choices on the Strong Vocational Interest Inventory; i.e., subjects with high total refined empathy scores seldom chose the "indifferent" category of response. The authors then discovered that these subjects also had selected the extreme categories of response on the I ( . 2 item questionnaire used to obtain the empathy scores. The correlation (rho) between the tend ency to select either the "almost always” or "almost never" categories on the questionnaire and the "like" or "dislike" categories on the Strong was *^7* The authors also found that subjects who used the extreme categories on the questionnaire tended to choose as associates whose responses to the questionnaire they were going to predict subjects who used the middle categories ("seldom” and "frequently") most often* The rho in the correlation of these subject-associate response tendencies was .32. These findings led Hastorf, Bender, and Welntraub to conclude that The refined empathy scores gives credit for the subjectfa moving away from himself when he predicts the other. However, a subject who makes mid-scale self-responses thereby limits the amount by which he can move away from himself in his predictions; there fore, he loses out on assumed similarity. (1955: p. 3I 1 . 3) This was not well put, but, as will be apparent when Gage and Cronbachfs (1955) discussion of this point is reviewed, the point is that the refined empathy score does not distinguish between accuracy of social perception and the tendency to assume dissimilar1tv. The relation between assumed similarity (projec tion) and accuracy of social perception was presented in a new kind of prediction ta3k developed by Tagiuri in a paper on what he called "relational analysis" (1952). He added a "guessing" task to the conventional sociometric rating task; i.e., in addition to asking his subjects to indicate who they liked and who they didnrt like in their 25 defined group, he asked them to predict which subjects had indicated they liked them and which had indicated they didn't like them* His main finding in this first study was that if a subject indicated that he liked another subject, it was unlikely that he would predict that the other would dislike him, and if a subject indicated he disliked another subject, he was unlikely to predict that the other would like him. The following year, in conjunction with Blake and Bruner, Tagluri (1953) discussed his method of relational analysis further. H© indicated that with the help of Luce and Macy (1955) 116 worked out a statistical model for estimating the chance occurrence of combinations of "like," "indifferent,” and "dislike" ratings in the triangular situation where one subject sociometrically rates another and predicts how the Other will rate him when the rating of the Other for the subject is given (but not known to the subject). Tagiuri was especially inter ested in whether agreement within any of the three possible pairs of ratings occurred more often than would be expected by chance: He labeled the case where the self rating and the Other rating agreed "mutuality"; i.e., the feeling each had for the other was "mutual," whether it was liking, disliking, or indifference. The case where the self rating and the rating predicted for the Other agreed was called "congruence"; i.e., the subject’s prediction of the OtherTs rating was congruent with his own feelings for the Other. The case where the predicted rating for the Other and the Other’s rating agreed was labeled "accuracy." For the data reported on in this study, Tagiuri, Blake, and Bruner found that their subjects were no more accurate than "chance robots" in perceiving rejection and indifference. In other words, when the Other’s rating was a "dislike" or an omission ("indifferent"), the number of agreements between the predicted rating for the Other and the Other’s rating was not different from what would be expected by chance. They also found that when the subject either liked or disliked the Other, he usually predicted that the other either liked or disliked him, respectively; but congruency was not the rule when the subject omitted the Other from his ratings (was indifferent to him). Agreement between self and other ratings— mutuality— was not different from chance expectancy In this study, but the same authors later reported (Newcomb, 1958) that they did find more mutuality than would be expected by chance. The existence of mutuality in excess of chance expectancy was Important because of the logical Inter dependence of the three variables: Where congruency and mutuality existed, accuracy had to exist; where congruency and nonmutuality existed, accuracy could not exist* Since the early studies in social perception were primarily concerned with accuracy, the realization by Tagiuri and his co-workers of the merely logical necessity of accuracy or inaccuracy under these conditions was noteworthy, especially in view of the empirical finding that congru ency far exceeded chance expectations. (Bender and Hastorf had already realized the same thing with regard to their scores: "Congruency” was another name for "projection" or "assumed similarity." To say that accuracy scores were confounded with congruency when mutuality was present was the same thing as saying that empathy scores were confounded with projection when the subject and the other responded similarly to the test instrument.) Norman and Ainsworth (19£>3) performed an interest ing and imaginative variant of the conventional social perception study. They examined the relations among scores they derived from administrations of the Guilford- Martin Inventory of Factors GAMIN. First they asked subjects (7^ college men) to fill out the inventory as a self-description, then they asked them to fill it out as "most others of their own age and sex" would. A "projec tion" score was derived by giving subjects a point for each item on which they denied possession of an 28 undesirable trait but predicted that most of their group would possess it when the majority (£1%) of the members of the group did not, in fact, indicate on their self description that they possessed the trait. This operational definition of projection was closer to the psychoanalytic definition of it as a defense mechanism than was the definition given by Hastorf and Bender (similarity between self-description and predicted self description of some other person). Norman and Ainsworth derived an "empathy” score by giving subjects a point for every item on which their prediction of "most Others" response agreed with the response given by most Others on the item. This was similar to the conventional empathy score, except that the subjects were asked to predict a stereotype and the prediction was compared to the actual stereotype. The prediction task was more similar, therefore, to that required of subjects taking The Empathy Test or that required of their subjects by Chowdry and Newcomb, than it was to that which Dymond, Fiedler, Gage, Hastorf and Bender, and others required of their subjects. (Gage (1902) did ask his subjects to predict the responses of a stereotype, but he did not compare these predictions to any group norm.) Norman and Ainsworth also derived a kind of group 29 assumed-similarity score, which they called "reality”: A point was given on each item for which the individual subject’s prediction jf "most Others" response agreed with the majority of the Other’s predictions of how most Others would respond. The authors called this kind of agreement "reality" because they felt that one component of reality would be that the subject saw the world as others saw it. The authors were aware that the projection and empathy scores were derived from the same base (the subjects’ self-descriptions),^- so they derived projection, empathy, and reality scores for each of three traits measured by GAMIN, A, I, and N. Intertrait comparisons of the three scores were then made. The median correla tion between the six projection scores and the six empathy scores wa3 -.29* The median correlation between projec tion and reality was -.16 (not significant). The median ^Cronbach (19^5) implied that Norman and Ainsworth were not aware that their projection and empathy scores were derived from the same base, whereas they explicitly state that they were. Cronbach concluded, therefore, that the correlations Norman and Ainsworth reported had no "psychological meaning." He was wrong on two counts: (1) Norman and Ainsworth correlated projection and empathy scores computed from different items, and thus no arti- factual overlay could exist between the scores; (2) Cronbach’a model did not apply to the AS and accuracy-of- predictlon scores used by Noman and Ainsworth, for they asked their subjects to make only one prediction for each Item on GAMIN, and so the accuracy scores did not, as Cronbach claimed, contain AS components. 30 correlation between empathy and reality was *30. The Internal consistency of the three scores was demonstrated by their median correlations across the three traits: projection, *50; empathy, .50; reality, .lj.8* (These were not corrected coefficients.) Using "adjustment" scores from the GAMIN key on the three traits as criteria, the authors found the following median Intertrait correlations: projection vs. adjustment, .13 (this was insignificant, but an arti- factual positive correlation might have been expected because denying the possession of undesirable traits went with adjustment on the inventory); empathy vs. adjustment, .38; reality vs. adjustment, .21 (not significant). Prom these results it appeared that social percep tion scores--and "projection," as defined by Norman and Ainsworth, could be considered social perception— were reliable, bore reasonable and statistically significant relations with one another, and had some relation to social adjustment as measured by three of the factor scores from GAMIN. The critical articles Before 1955 there was some criticism of methods and goals within the various camps of social-perception investigators (cf. Hastorf and Bender, 1952? Tagluri, 31 Blake, and Bruner, and Gage, 1953)» but these were the relative quiet before the storm. In his article on "Processes affecting scores on 'understanding of others' and 'assumed similarity,Cronbach (1955) made an apparently devastating analysis of the components of social perception scores, an analysis that seemed to have coincided with the withdrawal from the field of most of the early investigators. Cronbach was concerned with what he called the "myopic operationism" that characterized modern (19^5- 1955) research in social perception, and so he analyzed into separable components what he took to be the conven tional operations used in measuring "accuracy” and assumed similarity. His purpose was to show "... that investigators run much risk of giving psychological interpretation to mathematical artifacts when they use measure which combine the components" (1955* P* 177). He was also interested in demonstrating the psychological usefulness of the scores when properly analyzed. Cronbach said that the "usual" accuracv-of-social- perception score was the sum of four components which he called "elevation" (E), "differential elevation" (DE), "stereotype accuracy" (SA), and "differential accuracy," (DA). By "usual" he meant that each subject (or "judge") predicted the responses of more than one Other person on 32 more than one trait or item* The E component was separable if the judge predicted either the response of more than one Other on a single item or the responses of one Other on more than one item, but the other components were neither separable nor existent unless Cronbach’s standard of "usualness” was met. His method of analysis was not applicable, for example, to the sociometric predictions that Tagiuri and his co-workers dealt with, where predic tions were made for one item only, "like-dislike." The E component Cronbach spoke of was like the familiar "leniency" error known from rating studies. If the Judge’s central tendency of response averaged over items or over Others was different from the central tend ency of predicted responses, E was present* The DE component Indicated how close the judge’s average prediction for the Other was to the Other’s actual central tendency of response; this component seemed to reflect, therefore, the judge’s ability to predict the relative leniency errors of Others. The SA component showed how close the judge's predictions on each item approximated the item mean and scatter of the Others’ responses. In Mead’s terminology, it reflected accuracy In predicting the "generalized other." The DA component Indicated accuracy in predicting the responses of each Other Item by item; i.e., holding both ability to predict the Other’s 33 central tendency, and the central tendency of Items constant, DA indicated how accurately the judge varied his predictions for individual Others on each item, Cronbach concluded: Surely these aspects of social perception do not all reflect the same trait. A Judge who happened to use the same region of the response scale as other persons (Elevation is small) need not have superior insight. Judging which items have the highest mean (SA) seems to require acquaintance with the norms of the group; but a person might possess such knowledge and yet be unable to differentiate accurately between individuals. (1955* PP* 179-180) As an example of what he meant, Cronbach criticized the work of Chowdry and Newcomb (19^2). He said that their accuracy score combined E and SA, but Chowdry and Newcomb*s subjects did not predict the attitudes of each of the members of their groups; rather, they made a single estimate of what they thought the group norm was for each attitude or opinion rated. In other words, they asked their subjects to predict the stereotype; the stereotype was not statistically inferred in accordance with Cronbach's analytic technique. Granted that an elevation component could be separated from their scores, it also had to be granted that Cronbach evidently did not care at the time to distinguish between these two ways of obtain ing stereotype predictions: asking subjects to predict how the majority of a specified group would respond (Chowdry and Newcomb) vs. asking subjects to predict the responses of each member of a specified group and then averaging the responses to get the stereotype predictions* Later, along with Gage (195!?)» he did make the distinction and said that the two kinds of stereotypes might not coincide, but he and Gage went on to.say, "Such a discrep ancy between what might be called the Judge’s ’conscious’ and ’unconscious' concept of the group could be of considerable interest" (195£* P« I4 . 18) • Surely there was no justification for implying that a measure obtained indirectly (statistically inferred) was thereby a measure of something indirect or hidden, i.e., "unconscious." Measures of stereotypes obtained directly and indirectly probably should have been compared, but the primary purpose would have been to evaluate the advisability of giving the two the same name if they did not quantita tively coincide. Cronbach analyzed assumed-similarity scores in the same way he analyzed accuracy-of-social-perception scores, but, after presenting some empirical data, he concluded that more evidence was needed to determine how to divide AS. He thought AS wa3 perhaps a general mental set, and therefore some sort of "global index" might be a satis factory score. In his recommendations, Cronbach saidi Studies of perception may be concerned either with constant processes or with variable processes. . . . 35 We have seen that methods currently used are affected by both constant (within judge) and reactive (between judge and Other) processes, and therefore cannot serve well to investigate either* . . . An argument can be presented for concentrating attention on constant processes, taking up interactions between the judge and the other only after the constant processes characteristic of the judge are dependably measured. (1955: P. 190) Cronbach'3 first critical analysis of social perception scores was soon followed by another that he co-authored with Gage (1955)* The authors represented the social perception situation by a four-cell contingency table. The personal choices (self-descriptions) of the judge were identified by the columns and the judge's predictions of Others by the rows. The self-descriptions were called ’ ’ real dissimilarity" (RD) or "real similarity (RS), indicating those occasions when the Judge's self- ratings were different from or the same as the Other's self-ratings, respectively. The predictions were called "assumed-similarity" (AS) or "assumed-disslmilarlty" (AD), indicating, respectively, those occasions when the judge's predictions were similar to and dissimilar from his self-ratings: Self-Descriptions AS Predictions AD RD RS The relation between every Judge and every Other could be characterized by a tally in a particular cell* For example* if the Judge was similar to the Other (both made the same response) and also assumed that the Other was similar to him (predicted that the Other would respond as he had responded), the tally went in cell (b), Tallies in (b) and (c) summed to an accuracy score. Tallies in (a) and (b) summed to an AS score. The authors said: When a Judge is predicting an Other, we may regard the real similarity or real dissimilarity of this pair on any item as fixed independently of any social perception by the Judge. Now we may ask, within the real similarity items: If the Judge predicts correctly, is he accurate? Or does he assume similar ity? Obviously, these questions are operationally Identical, The count of such items represents ^warranted assumed similarity,1 and there is no way to distinguish whether this represents the mental set to assume similarity or the ability to Judge accurately. • . , Among the real dissimilarity items where he predicts correctly, we might ask: Does the Judge recognize the dissimilarity or does he assume dissimilarity? (1955s P* 4- 16) To remove that part of the accuracy score that was due to real similarity would eliminate the questions "Is the Judge accurate? Or does he assume similarity?" but would leave the investigator with only the tallies in cell (c) as a measure of accuracy. He would not know whether those tallies represented the ability to judge accurately or the set to assume dissimilarity, as Hastorf, Bender, and Welntraub were to discover (1955)* 37 In discussing the "process” of social perception, Gage and Cronbach said that Various global iispositions of the Judge appear to account for much of the variance in accuracy scores* Two dispositions of this kind can. be identified* First, Judges seem to differ significantly in their over-all tendencies to react favorably or unfavorably toward Others, both before and after the others have been observed. A second kind of disposition has been termed the Judge's 'implicit personality theory*; this consists of 'built-in' correlations that the Judge consciously or unconsciously imposes on traits, characteristics, or behaviors of Others. • . • Judges have been found to differ in the closeness and direc tion of the associations they implicitly assume between traits.5 Hence, in the bulk of research to date, social perception as measured is a process dominated far more by what the Judge brings to it than by what he takes in during it. His favorability toward the Other, before or after he observes the Other, and his implicit personality theory, formed by his experiences prior to his interaction with the Other, seem to determine his perceptions. . . • . . . the process of perception is so laden with affect, and so highly overlearned in the course of socialization, that the dominant role of global dispositions might well be expected. (195£* P. ip20) Gage and Cronbach thus concluded that investigators would be likely to find social perceptions determined more by the perceiver than by the social situation. General traits or response sets probably were at work in social perception; it probably was better to study the perceiver first and the interpersonal relation sometime later. t L ^The first disposition might have been a combina tion of the familiar "leniency” and "halo” errors observed in many studies of ratings, and the second disposition was reminiscent of the "logical” error (Guilford, 195^)* 38 Recent trends Id aoclaX perception research The two critical articles that Cronbach authored and co-authored apparently had a powerful but strange effect upon investigators interested in social perception. He had two theses: 1. Social perception scores have many components. Investigators must isolate the components of the scores and use the components as independ ent variables. 2. Before looking at the interpersonal relation to determine the perception, look at the perceiver; i.e., study the main effects before bothering with the interactions* Since these critical articles appeared to have discouraged a continuation of earlier lines of research and presented suggested lines for future research, it might have been expected that investigators already In the field and those to follow would have concentrated on two things: 1. The meaning of the accuracy-of-prediction or assumed-similarity scores used in social perception research. 2. The contribution of the perceiver, as one experienced In social perception, to the perception of a social stimulus. Quite the opposite reaction occurred. The main effect of the critical studies apparently was to discourage the use of social perception prediction or difference scores as Independent variables. Only Fiedler of the early Investigators continued to use differ ence scores obtained from social perceptions as 39 independent variables, and he tacitly ignored Cronbach fs criticism by leaving his asstoned-similarity scores tin analyzed. The general reaction against accuracy-of-prediction as a meaningful variable in social perception prompted many psychologists to investigate the process of social L perception. Tit© term "process" meant that studies of social perception would no longer be concerned primarily with the prediction of Others1 responses to assorted stimuli. Rather, some experimental controls were intro duced before those predictions were called for. For example, Lundy (1955) first had his subjects talk to the subject whose responses they were going to predict. He gave half the subjects the instruction to learn as much as they could about the person they were to talk with and he gave the other half of the subjects the instruction to tell the other person as much about themselves as possible. He was thus interested in the effect of the instructions he gave on accuracy-of-prediction and AS scores. The studies reviewed here were concerned with the trait or ability of social perception in individuals. 6 Several studies could be cited, e.g., Blake and Helson (195&), Jones and deCharms (1958), Horwitz (1958), and Tagiuri (1958). Descriptions of these studies were not included in this literature review because they were not specifically relevant to the hypotheses to be tested. ko rather than with Individual differences in social perceiton which were a function of the persons or situa tions perceived* *t is the thesis of this dissertation that the research done between 19^+9 195£ not come up with an agreed-upon mode of defining social perception operationally, despite many varied and imaginative attempts to do so. Probably largely as a result of CronbachTs critique of "myopic operationism," attempts at definition stopped after 195> 5> before previous attempts at definition (or measurement) had been combined in a single study to determine either the equivalence or the corre lates of the operations used* Re-Statement of the Problem The studies reviewed here indicated that social perception had been defined by a variety of "operations" (procedures to measure Individual differences). For example, individual differences In social perception were measured by making the following kinds of comparisons: 1. By comparing individual predictions of how one or more Others would respond with the actual responses of the Other(s). a. Dymond's "erap athy s c ale," b. The procedure used by Steinmetz, Gage, Hastorf and Bender, and many others: Individuals first took a conventional personality, Interest, or attitude scale, and then predicted how specific Others would answer the test items* c. Tagiuri’s 3ociometrio predictions. 2. By comparing individual predictions of how one or more Others would respond with the individ ual’s own responses. This was Fiedler's "assumed similarity" measure. 3. By comparing Individual predictions of how certain defined groups of people would respond with actual evidence of the groups' response. a. The Empathy Teat. b. Gage's, and Norman and Ainsworth's stereo type predictions. c. CronbachTs derived "stereotype accuracy" scores. 1*. By comparing individual predictions of how someone the individual has positive feelings about would respond with the individual's predictions of how someone he has negative feelings about would respond. This was Fiedler's "assumed similarity of opposites" measure. Very little evidence of the reliability of these measures was offered, and virtually no evidence of their functional equivalence or lack of equivalence was presented. Accept able retest and internal consistency reliabilities were quoted for The Empathy Te3t: and Gage, and Norman and Ainsworth showed that their measures were general (or consistent) across predicted-Others and across sub-tests, respectively. But here the evidence for reliability ended. The only study of the relations between two different measures of social perception was that reported by Bell and Stolperj they found no correlation between scores on The Empathy Test and scores on a Dymond-like scale (predictions of the responses of individual Others). k* Various criteria (dependent variables) were correlated with measures of social perception* The gen eral picture was not clear, but there appeared to be some relation between leadership and/or popularity and ability to predict group opinion. There was also a possible relation between ability to predict Individuals* responses and various sociometric Indexes. But beyond this, little could be said. The problem, then, was this: Assuming they were reliable, did the various measures of social perception that had been used measure the same thing, or "aspects of’ 1 the same thing? This, of course, was not the whole problem. Even if substantial correlations were found between two or more measures of social perception, this would be no demonstration that the two (or more) measured "social perception." Some other characteristic of the subjects or the measures might account for the correlation without resort to the "trait level" of explanation. CHAPTER XI METHOD The description of the method used in this study will begin with general statements of the experimental hypotheses* This will be followed by a description of the subjects used in the experiment* Then the Independent and dependent variables used to test the hypotheses will be described. Next the data collection procedure will be described. The chapter will conclude with a description of the various stages of the data analysis* Experimental Hypotheses I* The ability to predict the responses of Other individuals on attitude or temperament scales or on sociometric ratings is correlated significantly with the ability to predict the average or modal response of members of specified groups on similar scales or ratings. II. The tendency to assume similarity between self responses and predicted-Other responses on attitude or temperament scales or on sociometric ratings is correlated signifi cantly with the tendency to assume similarity between positively-cathected and negatively- cathected predicted-Others1 responses on similar scales or ratings* III. The ability to make accurate predictions about the responses of Individuals or groups, the tendency to assume that the responses of i+ 3 M l Others will be similar to one’s own responses in the same situation, and the tendency to assume that the responses of liked and disliked persons will be similar in the same situation are correlated significantly with effectiveness or satisfaction as a worker. IV. The ability to make accurate social percep tions or the tendency to assume similarity are indications of personal adjustment, motivation, satisfaction, or some other personality or temperament variable, and so experimental measures of that ability or tendency will account for a significant amount of the variance in a criterion of job effectiveness over and above that accounted for by conventional cognitive predictors. Experimental Subjects The subjects who provided the data for this experiment were seventy Air Force enlisted men who main tained the electronic bombing and navigation equipment (the ’ ’K-system") aboard B-lj.7 bombers of the Strategic Air Command. The men were assigned to one of four Armament and Electronics squadrons at either Davis-Monthan Air Force Base (AFB), Tucson, Arizona, or March AFB, Riverside, California. All the men in these squadrons who were rated at the level of proficiency and who were assigned to flight line maintenance participated as subjects. This experimental sample was relatively homogeneous with respect to age, education, work experience, and aptitude variables. The average age of the subjects was 23*02, and the standard deviation of their ages was 2.$6; one man was 19» on© vas 36, and one was 31— ‘the rest were between 20 and 30* Seven men had either one or two semesters of college, and $Q of the remaining 63 either had finished high school or had passed the General Edu cational Development test that gave them the equivalent to a high school diploma. Thirty-one men came into the service from high school with no intervening work experi ence, and most of the other 39 bad held only general labor or clerical jobs. Their experience working on vhe K-system varied from 1 to years, with a mean of 3*3^ years and a standard deviation of 1.10. They had been in the squadrons in which they were tested from 8 months to 5 years, with a mean of 2.92 years and a standard deviation of 1.22. On 9 standard Air Force tests the means of the subjects in this sample were higher than those of stand ardization sample, and in all cases the standard deviations of the scores of the subjects in this sample were smaller than those of the standardization sample (cf. Table VII). Because of its demonstrated homogeneity, it was probable that the variance within this sample on the experimental variables was restricted, but the restriction did not turn out to be serious. The specialized nature of the sample did indicate, however, that the results obtained in testing the hypotheses could not be general ized to more representative groups. Experimental Measures The experimental measures used in this study can, for convenience, be divided into independent and dependent variables. Of the independent variables * the most important were the social perception and assumed similarity measures. These will be described first. Social perception measures The Job Perception Profile. Each subject filled out a rating scale called "Rate the Job of the 32l£OE Assigned to Plight Line Maintenance." A copy may be found in Appendix C. This scale will be referred to in this report as the "Job Perception Profile," or JPP. Down the left side of the page of the JPP were 12 job characteristics, such as "Opportunity to supervise and direct others" and "Danger and personal risk." Each subject indicated how much of the time, in terms of a rating scale going from "never" to "always,” his job was characterized by each of these 12 characteristics. The resulting profile of ratings represented the subject's "perception" of his job. k7 The mean ratings given on each characteristic by the subjects within each squadron were computed and used as estimates of the "true" characteristics of the Job, Within squadron means were used because the objective job could truly have varied from squadron to squadron as a result of administrative differences, personnel differ ences, and so forth. Two scores were derived from each subject’s personal rating of his job: 1, The raw difference score. The differences between each subject’s job-characteristic rating and the mean job-characteristic rating for his squadron were squared and summed across the 12 job characteristics, and the square root of this sum was obtained, This was a measure of the difference between the subject’s over-all perception of the job and the average perception of the job by all the men in his squadron. It was labeled "JPP-raw D." 2. The standard difference score. The job- characteristic ratings were first transformed to C scores within subjects,? Then the squadron means of these C scores were deter mined, Then the differences between each subject’s job-characteristic C-score and the mean job-characteristic C-score for his squadron were squared summed across the 12 Job characteristics, and the square root of this sum was obtained. This was a measure of the ?0n the C scale the mean was set at and the standard deviation at 2,0, A C score of 5 > was assigned if the raw score was within one-half S, D, below the mean; a C score of 6 was assigned if the raw score was within one-half S. D, above the mean; etc. In other words, a half sigma deviation on the raw score scale equaled a unit change in score on the C scale. 48 difference between the subject's over-all perception of the job and the average percep tion of it within his squadron with individual differences in elevation and scatter held constant* This score was labeled "JPP- standard D.ft8 Croribach's (1955) argument that raters who varied their ratings no more than and preferably less than those ratings that were to be predicted were varied had more "chance" of making correct predictions, an argument for which Crow (1957) found some experimental support, seemed to make mandatory the equalizing of raters' dispersions in prediction tasks. In addition to his own description of his job on the JPP, each subject was asked to predict how his super visor perceived his (the subject's)job, The supervisors also were asked to describe the job of the 5-l©vel mechanic using the JPP. One score was derived: The differences between each subject's predicted job- characteristic ratings and his supervisor's actual ratings were squared and summed across the 12 characteristics and the square root of the sum of the squared differences obtained* This was a measure of the subject's accuracy in Q Because the JPP-standard D scores proved to be valid predictors of performance criteria, they were recomputed from a standard score base rather than a C-score base to see if the variance thus retained from the raw scores would be reflected in larger validity coeffi cients. It was not. lj-9 predicting his supervisor’s ratings* It was labeled "JPP-Ace." The sociometrio predictions. Each subject filled out two forms, one labeled "Personal Preferences" and the other "Preferences of Others" (copies are included in Appendix C). On the first form he wrote down the names of the flight-line, 5-level mechanics in his squadron he really liked to work with and the names of those he did not like to work with. On the second form he wrote down the names of his fellow-workers whom he predicted had indicated in their "personal preferences" that they either really liked to work with him or did not like to work with him. Thus each man’s personal preferences was a criterion for each other man’s preferences of Others (or predictions of personal preferences). The derivation of scores from the sociometric predictions involved some adjustment or compensating procedures. Subjects tended to make more positive ratings than negative ratings (more "likes" than "dislikes"); the same tendency held for predictions about those ratings, probably because of the "congruence" phenomenon reported by Tagiuri (1952). It was easier (could occur more often by chance), therefore, to predict a positive rating than a negative. Adjustments were made, therefore, to reduce the value of correct positive predictions. Also, since subjects were not required to make any particular number of predictions, the subject who made fewer predictions probably had less chance of being wrong, assuming he was more cautious and only made predictions he was relatively sure of. For this reason, it was assumed that a subject’s failure to indicate that Others rated him either posi tively or negatively was a prediction that they were Indifferent to him. Three categories of ratings were, therefore, assumed in the scoring procedure: plus (like), zero (neither like nor dislike), and minus (dislike). To compute accuracy-of-prediction scores, the following weights were assigned for agreements and disagreements within a 3 X 3 contingency table of Otherf3 preferences vs. subjects’ predictions of those preferences: Preferences of Others Predictions + of 0 Other’s Preferences Figure 1. Weights for Sociometric Accuracy Scores 51 It will be noted that the negative preferences ("dislike" ratings) were weighted heavier than the positive preferences. Since fewer negative ratings were given, there was less "chance" of tallies appearing in those cells; yet agreements or disagreements occurring in those cells had equal meaning as far as accuracy of prediction went. Negative weights were given to the plus and minus columns in the zero prediction row to penalize those subjects who, for whatever reason, made few predic tions; a zero prediction combined with a plus preference was weighted less negatively than a zero prediction combined with a minus preference because there was more "chance” of a subject's being given a plus rating by others than a minus rating; giving the two combinations equal negative weights might have penalized too much the subject who was well liked and yet made few predictions. Since the tendency was to give both more positive predic tions and more positive preferences, the agreement in the plus prediction row was given less weight than the agreement In the minus prediction row, and the disagree ments were given more negative weight. Each subject's accuracy-of-prediction score was the % algebraic sum of the agreements and disagreements between his predictions and the actual preferences of the Others in his work group, using the weights given above. For example, if a subject predicted that an Other liked him and the Other Indicated that he did like the subject, then the subject received a score of "2" for that particular relation. In the same way he received a score for every other dyadic relation in his group (the number of such relations varied from 13 for the 3°3**d squadron to 19 for the 320th squadron). The scores thus obtained were labeled "W-Socio-Ace." By ignoring each subjects omissions ("indiffer ence" ratings), their preferences and predictions were also plotted in four-cell contingency tables. The columns were labeled "real dissimilarity” (RD) and "real similar ity” (RS), and the rows were labeled "assumed similarity" (AS) and "assumed dissimilarity" (AD), following Gage and Cronbaohfs schema (1955)i AS AD The tallies In cells (b) and (c) were summed to obtain unweighted accuracy scores. These scores were labeled "TJ-Socio-Acc." RD RS a b c d The Social Situation Test. Each subject took the Social Situations Test, devised by Guilford and his staff, under three kinds of instruction. (The test was reduced from 30 to 25 items since it had to be administered so many times, and the wording of some of the items was changed to fit the educational level of the experimental subjects. A copy may be found in Appendix C.) The first time they took it as a test, giving their personal choices for each item. Next, they were asked to answer the items as they predicted the 5-level mechanic in their squadron they liked to work with most would answer them. Finally, they were asked to predict how the subject they liked to work with least well would answer them.9 Before the Social Situations Test answer sheets could be scored, it was necessary to weight each alterna tive response for each item so that the "distance," if there were any, between a predicted response and a real response could be determined. Ten judges^ took the test and ranked the four alternatives for each item in the order of their "correctness" or "fitness." The mean rank ^The answer sheets obtained -under this final instruction were not used to compute social perception scores, since it was felt that the introduction of an "opposites set" would work against whatever set to make accurate predictions the subjects may have assumed. ■^The judges were all employees of Human Factors Research, Inc. Eight were men and 2 were women. The age range was from 22 to Iff?. All judges were college graduates. Six had degrees in psychology. for each alternative was determined, and this was the weight or value of the alternative.3"1 The response each subject predicted for the Other he liked to work with most was then compared with each response of that Other. Where the responses were different, the differences between the weights of the alternatives involved were squared. These squared differences,t where they existed, were summed across the 25 items„ The square root of this sum was obtained. This was the measure of the subjectTs accuracy of prediction of Social Situations Test responses. It was labeled "SS-Acc." Six SP measures were thus derived from the experi mental instruments used in this dissertation: the JPP-raw D scores (differences between subjects’ raw score JPP profiles and their squadron mean JPP profiles); the JPP-standard D scores (differences between subjects1 C-score JPP profiles and their squadron mean C-score profiles); the JPP accuracy-of-predietion scores ("JPP- Acc1 ': differences between subjects’ predicted JPP profiles for their supervisors and the supervisors’ actual profiles); the weighted sociometric accuracy scores ("W-Socio-Acc": the algebraic sum of the agreements and disagreements between the predictions subjects made about ^The reliability of these mean rank values was •93, as determined with Ebel’s (1951) intraclass formula. 55 the so dome trie ratings they received and the actual ratings their co-workers made, where the agreements and disagreements were weighted in terms of a priori prob abilities); the unweighted sociometric accuracy scores ("U-Socio-Acc": same as the W-Socio-Acc scores except that a priori weights were not assigned); and the Social Situations Test accuracy-of-prediction scores ("SS-Acc": differences between Social Situations Test responses predicted for liked co-workers and the co-workers' actual responses)• Assumed similarity measures The Job Perception Profile. Since each subject predicted how his supervisor would describe the job of the 5-level, flight-line mechanic, a measure of subjects' tendencies to assume similarity between themselves and their supervisors was obtained by computing the "distance” (D) between each subject's JPP and the JPP he predicted for his supervisor. This measure was labeled "JPP-AS." Each subject also filled out the JPP tinder two further kinds of instruction. First, each was asked to describe his job as the mechanic he could cooper ate with best would describe it; then he was asked to describe it as the man he could cooperate with least well would describe it. By computing the distance between the two JPPfs filled out under these instructions, Fiedler's (19A) assumed similarity of opposites measure (ASo) was obtained. This was labeled nJPP-ASo." The sooiometrlc ratings. Two AS (or "congruency") scores were obtained from the sociometric ratings. To obtain a weighted score that took into consideration the tendency of subjects to make more positive than negative ratings and to make more omissions on the prediction than on the preference task, a nine-cell contingency table similar to the one on page $0 was used: Personal Preferences 0 - Predictions *► 1 — 1 -3 of Other 1s 0 -1 0 -1 Preferences - -k -2 1 Figure 2. Weights for Sociometric AS Scores Each subjects AS score was the algebraic sum of the agreements and disagreements between his personal prefer ences and his predictions of the preferences of Others, using the weights given above. If he liked a given subject and predicted that subject liked him, he received a score of "I” for that particular relation, and so forth. 57 The scores thus obtained were labeled "W-Socio-AS." Unweighted AS scores were obtained from the same four-cell contingency table as illustrated on page 52. They were the sum of the tallies in cells (a) and (b). These scores were labeled "U-Socio-AS." The Social Situations Test. Since each subject took the Social Situations Test under three conditions, two AS scores were obtained: 1. The distance between each subject’s own test profile and the profile he predicted for the man he liked to work with most. This was an "ASp" score, in Fiedler’s (195^) terms, and indicated the extent to which each subject perceived the responses of his most-preferred co-worker on the Social Situations Test as being similar to his own. It was labeled "SS-ASp." 2. The distance between the test profiles each subject predicted for his most-preferred and least-preferred co-workers. This was another ASo score and was labeled "SS-A8o." Six assumed similarity scores were thus obtained from the experimental measures. Three were "general" AS scores: the JPP-AS (similarity between self and predicted- supervisor JPP profiles) the W-Socio-AS (similarity between personal preferences for particular Others and predicted preferences of those Others for the self), and the U-Socio-AS scores (same as the W-Socio-AS scores except that a priori weights were not assigned); one indicated the similarity between subjects and liked 58 Others: the SS-ASp scores; and the other two were ASo scores: the JPP-ASo scores (similarities between pre dicted JPP profiles for co-workers who would and would not be easy to cooperate with), and the SS-ASo scores (similarities between predicted Soc* Sit. test responses for liked and disliked co-workers who were named). The aptitude tests Three tests from the Factor Reference Battery II developed by Morsh (1957) and the Visual Attention test from the Wilson Driver Selection Battery (19^8) were administered to the experimental sample. A copy of each is Included in Appendix B. These tests, in the order of their administration, were: 1. Pattern Comprehension. This test was loaded on the "visualization" factor, which Morsh said was confused with a "perceptual speed" factor In some studies and a "general reason ing" factor in others. 2. Word Knowledge. This was a synonyms test devised by W.E. North. It was Included in the battery to measure the verbal comprehension factor. 3. Mechanical Information. This test was included in the battery to measure the "mechanical knowledge" factor. i j . . Visual Attention. To the writer fs knowledge, this test had not been included In a battery for factor analysis, but it would appear to measure both perceptual speed and concentration 59 (or perseverance). It was Included in this battery because it had some face validity for the K-system mechanic^ common task of reading schematic diagrams of electronic circuits. All these aptitude teats were speed tests. The time limit for each was quite short. The score derived for each was the number of correct responses. The educational development (achievement) tests Following a classification scheme developed by the 17. S. Department of Labor (no date), a test of "language development" (the Reading Comprehension test) and of "mathematical development" (the Mathematics test) were administered. Copies are included In Appendix B, The Reading Comprehension Test. This test was constructed by William Harris and Wallace High of Human Factors Research, Inc. It consisted of five paragraphs of technical material. Four questions were asked about each paragraph. The material was selected from equipment maintenance and operation manuals and was designed to match the difficulty and technical style of the material the mechanics read in the Technical Orders dealing with maintaining the K-system. There was no time limit on the test. The score was the number of correct answers. The Mathematics Test. "Mathematics Test BI703A," as used for screening purposes by the Air Force, was 6o adapted for use In this study. Nine additional items were written by Edward Parker of Human Factors Research, Inc., to cover elements of geometry and trigonometry needed to understand the current flow In electronic circuits. Two of these items were subsequently deleted. This test was designed to assess relative scholastic achievement In mathematics. It wa3 administered as a power test with no time limit. The score was the number of correct responses. The Vocational Interest Test The Vocational Interest Test consisted of the I j . 1 triads making up the "electronic technician’s key" of Kenneth Clark’s U. S. Navy Vocational Interest Inventory. A copy is included in Appendix B. A single score repre senting comparability to the average electronic technician in Clark's sample was obtained. This score was the sum of the occasions when the subjects’ likes or dislikes agreed with the likes or dislikes of Clark's Navy subjects minus the number of occasions when the subjects’ likes or dislikes were opposite to those expressed by Clark’s subjects. The job satisfaction 3cale A job satisfaction scale (JSS), called for adminis trative purposes the "Bomb-Navigation System Question naire," was administered to each subject. A copy Is 6 l included in Appendix B. The i |6 items were culled from 12 previous questionnaires used in Air Force research. Subjects could either ''agree strongly," "agree," be "undecided," "disagree," "disagree strongly" in response to each item. The items were phrased in either a positive or negative way, and so each subject’s score was a function of how often he agreed with positive statements and disagreed with negative statements. That is, a high score indicated a generally positive attitude toward the job, supervisor, and fellow workers. Part scores on (1) "Sense of Personal Achievement," (2) "Attitude Toward Supervisor," and {3) "Attitude Toward Co-workers," as well as total scores, were computed. The Social Situations Test As indicated earlier, each subject took the Social Situations Test as a test. In addition to scoring each subject’s personal choices with the "weighted key" already described, they were also scored with Guilford’s "right answer" key. Each subject's score on the Social Situa tions Test as a conventional aptitude predictor (rather than as an SP or AS predictor), therefore, was the number of right answers. 12 Cf. Hausman and Strupp (1953)» Schmid, Morsh, and Detter (1957), Twery, Schmid, and Wrigley (1958)« 62 The sociometric "popularity” score An indication of each subject’s popularity as a worker was obtained by giving him a plus every time one of his fellow workers indicated they liked to work with him and a minus every time one of them indicated they did not like to work with him. The popularity score (labeled "Socio-Pop.") was the algebraic sum of the likes and dislikes* Scores on the measures listed above were the independent variables used in this study. The SP and AS measures alone were used to test hypotheses (I) and (II), and they were used to help test hypotheses (III) and (IV). The other independent variables were needed to help test hypothesis (IV). The tests of hypotheses (III) and (IV) also required a dependent variable: a criterion of job performance. This job performance criterion (JPG) was the unweighted composite of the following three measures: The Task Check List "amount of supervision” scores The job of the 5-level, K-system mechanic assigned to flight line maintenance was broken down into I30 different tasks. On the Task Check List (a copy is included in Appendix C), each subject Indicated for each task how much supervision he required to perform the task, if any, if he had ever performed the task. Each subject’s 63 supervisor also indicated how much supervision, if any, each subject needed on each task. Any discrepancy between subject and supervisor in these task by task evaluations was ironed out during a personal interview. A three category scoring system was used: 0 if the subject had never performed a task; 1 if he needed some supervision in performing a task; 2 if he performed a task without super vision. These task 3Core3 were summed across the 130 tasks to obtain a total amount of supervision score. The combined mean-peer and supervisor job proficiency ratings Each subject rated the over-all job proficiency of those men in his work group he knew well enough to evalu ate on a 25 point scale. A copy of the scale is included in the TCL in Appendix C. The supervisors of each work group also rated the men on the same scale. No subject was rated by fewer than 5 ? of his peers, and many subjects were rated by as many as 17 of their peers. The mean number of peer ratings made per subject was 12.2?. Each subject’s mean-peer proficiency rating and supervisor proficiency rating was summed into an unweighted total rating score. This was the second component of the JPC. 61 i. The combined 3elf and supervisor task leadership ratings For 27 tasks that required two or more men for their performance, each subject rated how often he "took the lead" in the performance of the task when he worked with another 5-level mechanic* "Taking the lead" meant directing the performance of the other(s). The tasks and the rating scale are included in Task Check List in Appendix C. Each subjects supervisor was also asked to indicate for each task how often the subject was likely to take the lead when performing the task with other 5-l©vel mechanics. Discrepancies between the supervisor and self ratings were not inquired into during the interview. The scale for rating task leadership ranged from "never" to "always," with four steps in between. Numeri cal scores ranging from 0 to 5 were assigned for each step. The total score for the self and for the supervisor ratings was the sum of the 2? task scores. These total scores were summed into an unweighted combination to serve as the third component of the JPC. Summary of the experimental measm*e3 A total of 27 experimental measures are considered in Chapter III on the results of this study. These measures will be summarized in this section and the abbreviations used in referring to them will be reviewed. 6* Social perception measures* Six kinds of accuracy- of-prediction scores were derived from subjects’ responses to the JPP (Job Perception Profile, or job rating scale), the Soc. Sit* test (Social Situations Test), and the socioraetrlc predictions: (1) The accuracy of subjects in predicting the average of the job descriptions made by the 32l50E ’s in their squadron on the JPP ("JPP-raw D" scores). (2) The accuracy of subjects in predicting average of the job descriptions made by the 32150E's in their squadron on the JPP when all job-description scale scores had been transformed to C scores ("JPP-standard D"). (3) The accuracy of subjects in predicting their supervisors' descriptions of the job of the 321%0E on the JPP (" JPP-Acc"). (J+) The accuracy of subjects in predicting which of their co-workers had indicated in sociometric ratings whether or not they liked to work with the subjects, when the agreements or disagreements between the predictions and the ratings were given a priori weights (KW-Socio-Acc")• (5) The accuracy of subjects in pre dicting their sociometric ratings when the agreements and disagreements between predictions and ratings were not weighted ("U-Socio-Acc11). (6) The accuracy of subjects in predicting the responses to the Soc. Sit. test of the 3215>OE’s they most liked to work with ("SS-Acc") • 66 Assumed similarity measures. Six kinds of assumed- similarity scones were derived from subjectsT responses to the JPP, the socioraetric predictions, and the Soc* Sit. test: (1) The tendency of subjects to assume that their supervisors would describe the 32l£0ETs job on the JPP similarly to the way the subjects had described it ("JPP- AS"). (2) The tendency of subjects to assume that 5-level mechanics they could cooperate with easily on the job would describe the job similarly to 5-level mechanics the subjects would have trouble cooperating with on the job (t ! JPP-ASo" ). (3) The tendency of subjects to assume that 32150E t3 they would or would not like to work with would also either like or dislike to work with them, when the agreements and disagreements among the sociometric ratings and predictions involved were given a priori weights (f , W-Socio-AS"). (1 4 -) The tendency of subjects to assume that 32150E's they would or would not like to work with would also either like or dislike to work with them, when the agreements and disagreements between the sociometric ratings and predictions were not weighted ("U^Socio-AS")• (5) The tendency of subjects to assume that the men they most liked to work with would respond to the Soc* Sit* test items similarly to the subjectsT own responses (SS-ASp”)• (6) The tendency of subjects to assume that the men they most and least liked to work with would respond similarly to the Soc. Sit. test items ("SS-ASo"). The aptitude tests. Four aptitude tests were administered to the experimental subjects: (1) Pattern Comprehension ("Pattern Comp."); (2) Word Knowledge ("Vocab."); (3) Mechanical Information ("Mech. Info."); and ( I } . ) Visual Attention ("Visual Attn."). The achievement tests. Two achievement tests were administered: (l) Reading Comprehension ("Read. Corap."); and (2) Mathematics ("Math."). Other tests. (l) A Job Satisfaction Scale ("JSS") was administered and three subtest scores were derived: (a) Sense of Personal Achievement ("Sense of Ach."); (b) Attitude toward Supervisor ("Attitude to Supervisor") and (c) Attitude toward Co-workers ("Attitude to^ Co workers"). (2) Guilford’s Social Situations Test ("Soc. Sit. test") was administered and scored with Guilford’s key. (3) A sociometric popularity-as-a-worker score ("Socio-Pop.") was derived from the sociometric ratings. Criterion measures of job performance. (1) The Task Check List ("TCL") amount of supervision scores. These scores are usually referred to as the TCL amount-of supervision scores, but this maybe abbreviated to "AOS." (2) The combined mean-peer and supervisor proficiency 68 ratings* These ratings may be referred to by the initials "PR.” (3) The composite criterion of job performance (Job Performance Criterion) is usually referred to as the "JPC." Data Collection The data for this study were collected in conjunc tion with data collected by Human Factors Research, Inc., under contract AF l j . 1 (657) 2 i j . 8 to the Wright Air Develop ment Command, and with the cooperation of the men of the i|-3rd and 303rd Armament and Electronics squadrons stationed at Davis-Monthan AFB and the 22nd and 320th Armament and Electronics squadrons stationed at March AFB. The writer collected all the data personally. The men who were to serve as subjects were assigned to be tested and interviewed as a part of their jobs, but the job of maintaining the K-system came first, and so no rigidly standardized data collection procedure could be established. Also, the test battery was quite long; it took most subjects three to four hours to perform all tasks* Most of the subjects (52 out of 70) finished the battery in two sittings, but as many as four sessions were required for the others. In all cases the tests (or tasks) were given in the same order: The Task Check List (TCL) came first, followed by the aptitude tests, the 69 achievement tests, the interest tests, the JSS, the JPP, the socioraetric predictions, and the Social Situations Test. As soon as the subjects’ supervisors had completed their version of the TCL, in which they evaluated their men, the disagreements between each subject and his super visor regarding the amount of supervision the subject needed for particular tasks were tallied. If there were more than 10 disagreements (out of a possible 130), each disagreement was inquired into separately during a final interview with each subject. Before the subjects were interviewed, the writer went over each one’s ! , Form 7" on which mo3t of the facts pertinent to each subject’s Air Force history were listed. Some of these facts, such as the length of time the subject had worked as a £-level mechanic, the time he had worked on the K-system, the time he had been in the squadron he was presently assigned to, and so forth, were verified (or corrected) during the interview. Most of the men were interviewed in the "ready room," but a few had to be interviewed while working aboard planes on the flight line. At the conclusion of the interview, which took from five minutes to a half hour, depending on the extent of disagreement between the subject and his supervisor on the 70 TCL ratings, each subject was asked to fill out the JPP again. He was asked first to describe his job as the 5-level mechanic he ’ 'could cooperate with BEST" would describe it, and then to describe it as the 5-level mechanic he "could cooperate with LEAST WELL" would describe it. The data were collected at Davis-Monthan APB during the period from March 2, 1959 to March l i ^ . , 1959* Data collection was started at March APB on the 26th of March and was concluded on the 30th of June. Most of the test ing and interviewing was concluded at March APB by the middle of April. The delay in the final completion was caused by two things: 1. Some subjects permanently assigned to the 22nd or 320th A & E squadrons were temporarily on duty in Guam or Alaska as part of a regular SAC deployment scheme. These subjects were tested and interviewed when they returned from the temporary assignment. 2. It was necessary to get test-retest reliability estimates for the sociometric ratings and predictions, and for the JPP ratings. An interval was allowed to reduce the effect of memory on the test-retest correlations. Data Analysis The description of the data analysis procedures used in this study will be divided into four sections: the reliability of the experimental measures and other preliminary statistics; the basic intercorrelation matrix; 71 the multiple regression analysis; and further correla tional studies. The reliability of the measures and other preliminary statistics Most of the measures used in this study were "experimental." It was necessary, therefore, wherever possible, to get some estimate of their reliability. For the social perception measures the test-retest method probably provided the best estimate of the measures1 reliabilities, but because of the number of measures obtained in this study, it was not possible to readminister all of them. Only the sociometrlc ratings and predictions and the JPP were readministered to the subjects. The reliability sample for the former was 2 l \ . and for the latter 19* All cases retested were from March AFB. Internal consistency (split-half correlations) reliabilities were computed for the JPP-AS, JPP-ASo, JPP-Acc, SS-Acc, SS-ASp, and SS-ASo scores, for the Reading Comprehension, the Mathematics, and Interest scores, and for the three part scores of the JSS. The Internal consistency reliability of the mean- peer proficiency ratings was computed using Ebelfs intraclass formula (195>1)* The Internal consistency of the combined mean-peer and supervisor ratings was 72 determined by correlating the two and applying the Spearman-Brown correction. The internal consistency of the TCL amount-of-supervision scores was estimated in two ways: by combining the subject-supervisor task ratings in a 3 X 3 contingency table and computing the contingency coefficient. The internal consistency of the combined self and supervisor task leadership ratings was estimated by correlating the total scores each subject received from each source and applying the Spearman-Brown correction. To combine the TCL amount-of-supervision scores, the combined mean-peer and supervisor proficiency ratings, and the combined self and supervisor task leadership ratings into the unweighted JPC required that each be reduced to standard scores, the standard scores combined, and a constant added to make all scores greater than zero. Numerous other scores required that constants be added to make all scores greater than zero: W-Socio-Acc, W-Socio-AS, W-Socio-M, the JSS scales, and the Interest scores. Scatterplots of the correlations between all independent variables and the JPC were tabulated to insure that the criteria of linearity of regression and homo- skedasticity required for use of the Pearson r were not violated. 73 The basic In tercorrelation matrix Because 32 variables were originally involved in this study, it was considered uneconomical to compute the intercorrelations among them (1 |9^ in number) using a desk calculator. A basic matrix of 20 variables was selected, therefore, and the scores on these variables for the 70 men in the sample were punched on IBM cards and the intercorrelations among them were electronically computed by the 6f>0 computer of the Service Bureau Corporation of Los Angeles. Following the machine correlation of the 20 variable matrix, multiple correlation coefficients between the independent variables and the JPC were computed using Gengerelli’s (19^8) approximation method. Corre lations between six independent variables (all either SP or AS variables) not included in the basic matrix and the other 20 variables were then computed by hand. The multiple regression analysis Using the Doolittle technique, the partial regres sion coefficients were obtained for two selected sets of predictors and the JPC were then computed. Further correlational studies It became apparent as the hypothesized relations were examined that certain nonhypothesized relations between SP and AS scores should be Investigated. In addition to further correlational analyses, this entailed the recomputation of certain SP and AS scores to determine if artifacts would have existed in the relations between the scores as originally computed. CHAPTER III RESULTS Pour hypotheses were presented in general form in Chapter II* The specific tests of these hypotheses will be described in this chapter. In addition, the relia bility studies of the experimental measures, relations between the accuracy-of-precLLction and assumed-similarity scores, and relations between sociometric popularity-as-a- worker scores and scores on other variables will be presented and discussed. The chapter is divided into six sections: one for the reliability studies, one for each of the hypothesized relations, and one for the nonhypo thesized relations between the social perception scores and between the sociometric ratings and other measures. The reliability coefficients computed for the experimental measures will be presented and discussed in the first section. The following order of presentation will be observed within the next four sections: The hypothesis tested will be specifically stated; the results of the test(s) of the hypothesis will then be given; and then the results will be discussed. The fifth section will begin with the reasons for investigating the 75 76 nonhypothesized relations. The results of the investiga tions will then be presented and discussed. The Reliability of the Experimental Measures The reliability coefficients computed for the experimental measures are listed in Table I. The number of cases for whom reliability data were available and the kinds of coefficients computed are also listed. The JPP-raw D and JPP-standard D scores were com puted for the retest sample by comparing the subjects1 second ratings of their job with their mean squadron ratings derived from the first ratings rather than with the mean squadron ratings from the retest administration. This procedure was necessary because not all subjects in the original sample of 70 could be retested, but the procedure undoubtedly reduced the comparability of the two sets of JPP distance scores. Of the 36 subjects retested, 19 were contacted in person and the other 17 responded by mail. The 17 were all from Davis-Monthan APB; 1$ of the 32150E,s at Davis-Monthan did not respond to the request by mail to fill out the JPP again. The 19 subjects retested in person at March AFB were the only ones available for testing when the writer returned to the air base in June, 1959» to do the reliability study. The retest coefficients of .26 and .13 obtained TABLE I RELIABILITY OP THE EXPERIMENTAL MEASURES Reliability So. of Experimental Measure Coefficient Cases Kind of Reliability JPP-raw D .26 36 Test-retest, Pearson r JPP-standard D 36 n 1 1 JPP-Ace •ks? 70 Split-half, internal consistency SS-Acc .56b 70 t t t t n Sociometric predictions :Z? 2k Test-retest, contingency coefficient JPP-AS 70 Split-half internal consistency JPP-ASo • 8& 70 t i t t i t SS-ASp .33? 70 t t n t t SS-ASo .83b 70 t t t t t t Combined mean-peer & super • 89b visor proficiency ratings 70 Internal consistency, Pearson r Combined self and super .63* t t t t t t visor leadership ratings 70 TCL amount of supervision •39b 70 Internal consistency approximation, .91b contingency coefficient JPC 70 Mosier’s formula for the reliability .50? of a composite (Guilford, 195k* 393) Read* Comp. 70 Odd-even Internal consistency Math. • 93? 70 Split-half internal consistency JSS-Sense of Ach. .90? 70 t t t t t t JSS-Attitude to Supervisor • 93? 70 n t t t t JSS-Attitude to Co-workers .91? 70 t t t t t t Interest .82b 70 « t t t t Soc. Sit. •k-2 70 Kuder-RIchardson internal consistency bSignifleant beyond the ,01 level -4 -4 for the JPP-raw D and JPP-standard D scores were not significantly different from zero. The lack of reliabil ity in the D scores computed between each subjects1 JPP and his squadron mean JPP could be attributed in part to the fact that the retest ratings were not compared to mean ratings derived from those retest ratings. The lack of reliability might also have been caused by real changes in the job rated, since a period of from one to three months elapsed between ratings, but such changes probably were negligible. A more likely explanation was that the subjects were tired of filling out the JPP by the time the reliability study was done, because they had already been asked to fill it out four times under differing instruc tions. It is frequently difficult to get sustained cooperation from service personnel. Asking them to perform virtually the same task five times would hardly improve their cooperative spirit. Fortunately, the lack of demonstrable reliability in the JPP-raw D and JPP- standard D scores did not confuse the interpretation of the results. Both scores were reliably related to other measures. It was not likely, therefore, that the null hypothesis was accepted In the testing of any of the relations between the JPP-raw D end JPP-standard D scores and other variables when a valid relationship was present. The split-half internal consistencies of the 79 JPP-Ace, SS-Acc, JPP-AS, SS-ASp, SS-ASo, and JPP-ASo were calculated by getting D scores for the first and second halves of the tests. Pearson r ’s were computed between these D scores and the Spearman-Brown correction applied. For example, the square root of the squared differences between each subject’s predicted JPP ratings for his supervisor on the first six job characteristics and his supervisor’s actual ratings for those character istics was correlated with the square root of the squared differences between the subject’s predicted ratings and the supervisor’s actual ratings for the last six job characteristics. Accuracy on the first half of the JPP was thus compared with accuracy on the second half. Separate reliabilities for the accuracy-of- prediction and assumed-similarity scores derived from the sociometric predictions were not computed. Rather, 3 X 3 contingency tables were tallied for each subject In the retest sample. The subjects’ predictions on the first administration were Identified by the columns and their predictions on the retest administration were identified by the rows. If a subject predicted that a certain Other had indicated that he liked to work with him on both administrations, a tally was placed in the "like-like" cell; If he predicted that a certain Other had indicated that he liked to work with him on the first administration but omitted the Other's name on the second administration, a tally was placed in the "like-indifferent" cell; and so forth. The tallies for all cells except the "indifferent-indifferent" cell were used in computing the contingency coefficient. Including the "indifferent- indifferent" agreements, which merely Indicated that subjects had not written down the names of certain Others on either administration of the "Preferences of Others" blank, would have inflated the reliability coefficient obtained. The .73 coefficient obtained by arbitrarily putting a zero In the "indifferent-Indlfferent" cell indicated that the 2i j . subjects involved made substantially the same predictions on both administrations of the "Preferences of Others" blank. Conventional Inter-rater agreement indexes of reliability (Pearson r f3 corrected by the Spearman-Brown formula) were computed for the mean-peer vs. supervisor proficiency ratings and for the self vs. supervisor task leadership ratings. The reliability of the TCL amount-of-supervision scores was estimated by combining the 3 ^ 3 contingency tables in which each subject's task-by-task ratings of the amount of supervision he required were compared with the ratings his supervisor had given him on each task. If a subject had indicated for a task that he needed no 81 supervision and his supervisor had indicated the subject needed some supervision on that task, a disagreement was recorded; if a subject had indicated he had never performed a task whereas his supervisor had indicated he had performed it, a disagreement was recorded; and 30 forth. Prom the agreements and disagreements tallied in the 70 contingency tables, a contingency coefficient of agreement was computed. The resulting coefficient of .39 was significantly different from zero. The reliability of the composite criterion of job performance (JPC) was estimated, using Mosier's formula for the reliability of a composite, from the .89 relia bility of the combined mean-peer and supervisor proficiency ratings, the .63 reliability of the combined self and supervisor task leadership ratings, the .39 estimation of the reliability of the TCL amount-of- supervision scores, and from the .81 correlation between the proficiency and leadership ratings, the .61 correla tion between the proficiency ratings and supervision scores, and the .73 correlation between the leadership ratings and supervision scores. The odd-even or split-half internal consistencies of the Read. Comp. (.50), Math. (.93)» JSS-Sense of Ach. (.90), JSS-Attitude to Supervisor (.93)» JSS-Attitude to Co-workers (.91)# and the Interest (.32) tests were computed with Pearson r’s corrected by the Spearman-Brown formula. A corrected Pearson r (split-half internal consistency) was computed first for the Soc. Sit. tests scored with Guilford’s key, but the coefficient was so low (*33) that the item variances were computed and the Kuder-Riehardson formula #20 was applied. The resulting coefficient was still low (.1*2). The unreliability of the scores was due to the relatively low total variance and the relatively high item variances. The mean on the 25 item test was 16.51 and the standard deviation was only 2.61*. Reliabilities were not computed for the Factor Reference Battery tests since they had all been used previously and reliability coefficients reported (Morsh, 1957)* The reliability of the Visual Attention test was not estimated. Only test-retest or alternate forms methods would have been appropriate because it was a perceptual speed test on which few errors were made. An odd-even internal consistency coefficient would have been close to 1.0. The reliability coefficients obtained for the experimental measures were all significantly different from zero (p< .01) except those obtained for the JPP-raw D and JPP-3tandard D scores. As indicated above, the lack of demonstrable reliability for those two scores did not limit the interpretation of the results of the relations between the JPP distance scores and scores on other variables since both the JPP-raw D and JPP-standard scores were reliably related to scores on other measures* The Equivalence of Measures Used to Define Social Perception Operationally Two general abilities have been measured to define social perception operationally: the ability to describe or predict group responses or norms, and the ability to predict or describe individuals1 responses. The experi mental measures used in this dissertation were designed to measure those two abilities. The JPP-raw D and JPP-standard D scores represented attempts to measure subjects’ accuracy in perceiving a general social situation, with relative accuracy being taken as the difference between the individual subjects’ perceptions of their job and the average perception of the job by the other subjects within a given squadron. As described earlier, the JPP-raw D scores were standardized within subjects to eliminate the possible error or bias introduced by assumed irrelevant differences in the means or variabilities of individual subjects’ ratings. It was expected that subjects would interpret the JPP scale labels of "never,” "seldom,” "occasionally," "frequently,” "usually," and "always" differently. For example, some subjects might think of "frequently" as meaning as often a3 the average subjects* interpretation of "usually," and so all their ratings would be 20 points lower on the average. Mean differences like this would have inflated the JPP-raw D scores, and the variance thus induced might have had a large error component. To remove this kind of variance and that induced by subjects* tend encies to use varying amounts of the scale in making their ratings, the JPP ratings were standardized within subjects. The JPP-standard D scores were a measure of the relative scale values of the flight-line 321£0E*s job on the 12 characteristics given. That is, the JPP-standard D scores measured the relative shape of the job in terms of the 12 characteristics, whereas the JPP-raw D scores measured the absolute magnitude of the job on the 12 scales given. Both the JPP-raw D and JPP-standard D scores were intended as approximations of the kind of accuracy of social perception measured by The Empathy Test (Kerr and Speroff, 1951) and by the prediction-of-group-responses tasks used by Gage (1952), Chowdry and Newcomb (1952), and Norman and Ainsworth (195^). The JPP-Acc, SS-Acc, W-Socio-Acc, and U-Socio-Acc scores represented attempts to measure the accuracy of subjects in predicting the responses of specific Others. 85 Their relative accuracy was the difference between the predicted responses and the Others’ actual responses, where the predictions are made either across items or across Others. Subjects made predictions across the 12 characteristics on the JPP and across the 25 items on the Soc* Sit. test, or they made predictions across the 13 to 19 subjects they worked with on the T,Preferences of Others” sociometric prediction blank. Subjects were told that their predictions of their supervisors' ratings on the JPP and of their co-workers responses to the Soc. Sit. test would be scored for accuracy. The JPP-Acc, SS-Acc, W-Socio-Acc, and U-Socio-Acc scores were intended as approximations of the kind of accuracy-of-social-perception scores derived from Dymond’s scale (I9lp9) and of the prediction-of-other-individuals responses tasks used by Steinmetz (191+5K Gag© (1952), Hastorf and Bender (1952), and Tagiuri (1952). Neither the JPP-raw D and the JPP-3tandard D scores nor the W-Socio-Acc and U-Socio-Acc scores could be considered separate measures of accuracy of perception of the 32150E ’s job or accuracy of perception of co-workers’ likes and dislikes for the perceiver, respectively. The JPP-raw D and JPP-standard D scores were derived from the same ratings, and the W-Socio-Acc and U-Socio-Acc scores were derived from the same ratings. The relations between 86 these scores were investigated, but the results will not be considered in the test of Hypothesis I* Hypothesis I The ability of subjects (l) to perceive their job as it was perceived by other subjects in their work group, (2) to predict how their supervisors perceived their (the subjects’) job, (3) to predict the responses of liked fellow-workers on the Soc. Sit. test, and (l±) to predict which of their fellow-workers had indicated that they liked to work with them, disliked to work with them, or were indifferent to them are all related. Therefore, the JPP-raw D and JPP-standard D scores, the JPP-Acc scores, the W-Socio-Ace and U-Socio-Acc scores, and the SS-Acc scores will, in one or more of the 13 acceptable combina tions among them, intercorrelate positively above the .05 level of statistical significance. Results of the test of Hypothesis I The intercorrelations of the six measures of accuracy of social perception are presented in the matrix in Table II. Only two of the 13 correlations that tested Hypothesis I were significantly different from zero, those between the JPP-raw D and JPP-Acc scores (.I(.7) and between the JPP-Acc and U-Socio-Acc scores (.2^). As Indicated in Table II, the .lj.9 correlation between the JPP-raw D and JPP-standard D scores and the .57 correlation between the W-Socio-Acc and U-Socio-Acc scores had to be consid ered artifactual because the scores correlated were obtained in the same operation (i.e., from the same basic data)• TABLE II INTERCORRELATIONS AMONG MEASURES OP SOCIAL PERCEPTION 1. JPP-raw D 2. JPP-standard D 49b 3. JPP-Acc 4 7 b .15 k- SS-Acc i . o \ _ o -.10 .06 5. U-Socio-Acc .10 .02 . 2 1^ -.09 6. W-Socio-Acc .05 .06 .18 .16 .S7b 1 2 3 k 5 6 Significant beyond the .05 level. ^Significant beyond the .01 level. Hypothesis I was given provisional support by the results of the experimental correlations, but, in general, the correlations indicated that the operational defini tions of social perception used in this dissertation did not provide substantially equivalent measures. Social 88 perception defined by the W-Socio-Acc scores was different from social perception defined by the JPP-raw D, JPP- standard D, JPP-Acc, and SS-Acc scores* Social perception defined by the JPP-standard D scores was different from social perception defined by the JPP-Acc or SS-Acc scores. And so forth. Discussion of the results of the test of Hypothesis I The correlation between the JPP-raw D and JPP-Acc scores was substantial, but before it could be concluded that the two prediction- or description-tasks provided relatively comparable measures of an ability called "social perception," it was necessary to consider first that the two scores were derived from ratings made on the same scale under different instructions. It was possible that part or all of the *4? correlation between the JPP- raw D and JPP-Acc scores could be accounted for by systematic ways in which Individual subjects made their ratings on the JPP job-characteristic scales regardless of Instructions. Consider, for example, that the correla tion between the JPP-standard D scores and the JPP-Acc scores was only .15* Certain systematic rating tendencies had been removed from the JPP-raw D scores in deriving the JPP-standard scores: the tendency to confine ratings on the average to the upper or lower regions of the scale, for example, or the tendency to spread ratings out all over the scale as compared to the tendency to bunch them somewhere near the middle. To compare the rating tendencies of subjects whose JPP profiles were most similar to their squadron mean JPP profiles and whose predicted-supervisor JPP profiles were most similar to their supervisors* actual JPP profiles with the rating tendencies of subjects whose JPP profiles and predicted-supervisor JPP profiles were least similar to their respective criterion profiles, the JPP-raw D and JPP-Acc score distributions were split into "accurate" and "inaccurate" halves. The mean JPP and mean predicted- supervisor JPP scale (job characteristic) scores and the standard deviations of those scale scores were calculated for the accurate group and for the inaccurate group. (JPP and predicted-supervisor JPP scale scores could range from 0 to 100.) The results are presented in Table III. The differences between the accurate and inaccurate group means and standard deviations were not statistically significant, but for both the JPP-raw D and JPP-Acc scores the accurate group means were higher and the standard deviations narrower. Those subjects who used the JPP most accurately, therefore, tended to give more "always," "usually," and "frequently" ratings on the average and to use a narrower range of the scale, whether they were describing their job or predicting their supervisors’ description of their job* In performance rating terms, the more accurate subjects were characterized by two familiar rating errors; the leniency error, and the central tendency error. According to Cronbach’s (1955) deductions, the latter "error" alone could account for the greater accuracy shown by the subjects who tended to use a smaller range of the JPP scale in making their ratings. TABLE III THE MEANS AND STANDARD DEVIATIONS OF THE ACCURATE AND INACCURATE GROUPS ON THE JPP-RAW D AND JPP-ACC SCORES Measure Group Mean | Standard Deviation JPP-raw D Accurate 56.87 21.16 Inaccurate 53.81 26.if0 JPP-Acc Accurate 6J 4 .. 80 2543 Inaccurate 62,13 29.32 The • relation between the U-Socio-Acc and JPP- Acc scores was a raw score relation, like the relation between the JPP-raw D and JPP-Acc scores. When the socio metric predictions were weighted to reduce the effects of subjects’ tendencies to make more positive than negative sociometric choices and to omit more co-workers from their sociometric ratings than they included, the relation between the W-Socio-Acc scores and the JPP-Acc scores dropped to .18, which was not significantly different from zero. This drop in the relation between the socio metric predictions and accuracy in predicting supervisors* responses to the JPP when certain response tendencies in making sociometric ratings and predictions was controlled was some indication that the . 2 i ( . relation between the TJ-Socio-Acc and JPP-Acc scores might have been due in part to a concordance of response sets. There was evidently a tendency for subjects who made higher predictions with a narrower range on the average when making JPP ratings also to make more positive ("like”) sociometric ratings or to make fewer sociometric predictions. That is, subjects who were more "conservative” and more "lenient" in making ratings on the JPP (gave their job more nearly the same rating on all characteristics and used the upper end of the scale for most of their ratings) also tended to indicate that they liked more co-workers in making their "Personal Preferences" and to make fewer predictions about which of their co-workers liked or disliked to work with them. But the tendency was slight, as was the relation between the U-Socio-Acc and JPP-Acc scores. In general, the operations used in this disserta tion to measure social perception were not equivalent. The ability to describe a social object (a job) specified by normative methods (JPP-squadron mean profile), the ability to predict individuals1 (supervisors’) job descriptions or individuals’ (co-workers’) responses to an objective test (Soc. Sit. test), and the ability to predict one’s own sociometric standing were not, in terms of the subjects and measures used in this study, equiva lent abilities. Some equivalence was noted when subjects used the same scale (the JPP) for two different prediction or description tasks, but it was possible that this equivalence was due to a "response set" (Cronbach, 19^6). The barely significant relation between the U-Socio-Acc and JPP-Acc scores indicated some overlap (6^ common variance) between the ability to predict a supervisor's job description and ability to predict co-workers' likes and dislikes for the predictor, but this overlap was not considered great enough to warrant further analysis. The Equivalence of Measures Used to Define Assumed Similarity Operationally Pour types of assumed-similarity (AS scores were described in the literature. The experimental measures used in this dissertation were designed to represent three of those types.^ ^The fourth type, the "assumed self-typicality" described by Gage (1952), which was a comparison between a self-description and a "stereotype" description, indicated 93 One type of AS score was that derived from a comparison of a subject’s self-description on some scale with his description of some Other person on that scale or with his prediction of that Other’s self-description. The JPP-AS score (the distance between subjects’ job descrip tions on the JPP and their predicted job descriptions of their supervisors on the JPP) represented an attempt to duplicate this first type of AS score. The U-Socio-AS and W-Socio-AS scores also represented an attempt to measure subjects’ tendencies to assume that Others’ responses would be similar to their own. Tagiuri did not indicate in any of his writings that he tallied the number of his subjects’ "congruent1 1 responses (agreement between a subject’s affective response or relative liking for a co-worker and the affective response the subject predicted the co-worker had made toward the subject), or that he equated congruency with AS, but certainly the number of congruent responses a subject made could be considered a measure of his tendency to assume similarity to his co workers. Accordingly, the U-Socio-AS and W-Socio-AS scores were derived. The relation oetween these two the extent to which subjects assumed they were similar to the average member of some group. Gage found this to be the least valid predictor among his social perception scores, and so it was not Included as an experimental measure in this study* 9k Socio-AS scores was not used In the test of Hypothesis II, however, because the two scores were derived from the same ratings and predictions. The SS-ASp scores measured the tendency of subjects to assume that men they liked to work with would respond to the Soc. Sit. test in the same way they did. This was a close approximation to the kind of ASp score Fiedler used (195^). * The SS-ASo and JPP-ASo scores measured the tendency of subjects to assume that men they liked and men they did not like would respond similarly to one another in taking the Soc. Sit. test or in rating their job on the JPP. Both these scores were close approximations to the ASo scores Fiedler used (195^)* Hypothesis II The tendencies of subjects to assume similarity between (1) the way they perceived their job on the JPP and the way they predicted their supervisors perceived their (the subjects') job on the JPP, (2) the way they felt about their fellow workers (sociometric choice vs. rejection) and the way they predicted their fellow workers felt about them, (3) the way they answered Social Situa tions Test Items and the way they predicted the men they liked to work with would answer those items, and ( I j . ) the way men they liked to work with would answer the Social Situations Test items or describe their job on the JPP and the way men they did not like to work with would answer those items or describe their job are all related to a general tendency to assume similarity between oneself and Others, and so the JPP-AS scores, the U-Socio-AS scores, the W-Socio-AS scores, the SS-ASp scores, the SS-ASo scores, and the JPP-ASo scores will, in one or more of the l l j . acceptable combinations among them, inter correlate positively above the .05 level of significance. Results of the test of Hypothesis II The intercorrelations of the six AS measures are presented in the matrix below: TABLE IV INTERCORRELATIONS AMONG AS MEASURES 1. JPP-AS 2. U-Socio-AS o O . 3. W-Socio-AS • 07 p o 00 k- SS-ASp o o . o o • .11 SS-ASo -.07 o o .16 .31b 6. JPP-ASo 0 • 1 -.17 • o o .1^ • 5' 1 2 3 k $ Significant at the .01 level Not including the relation between the U-Socio-AS and W-Socio-AS scores, only two of the correlation coeffi cients in the above matrix were statistically significant: those between the SS-ASo and JPP-ASo scores (.55) and the SS-ASp and SS-ASo scores (*3l). Both coefficients were significant above the .01 level. Hypothesis II was, therefore, provisionally supported by these correlations, but, generally speaking, the results indicated there was little operational equivalence among the various AS measures used in this study. The tendency for subjects to assume similarity between their own responses and the responses they predicted for Others was generally depend ent on either the nature of the response(s) or on whose responses were being predicted. The two exceptions to this were the tendency to assume similarity between liked and disliked fellow workers (SS-ASo vs. JPP-ASo) and the tendency to assume similarity between the self and a liked fellow worker and between liked and disliked fellow workers on the Social Situations Test (SS-ASp vs. SS-ASo). Discussion of the results of the test of Hypothesis II The relation between SS-ASo and JPP-ASo scores was substantial* There were no artifacts in the relation. The scores were derived from tests administered on separate days, with the time intervening ranging from one day to a week or more, and so the correlation did not represent the operation of a temporary set* The predic tion task varied in two important ways: The Social Situations Test consisted of 25 multiple-choice items concerned with the proper actions to be taken in a variety of "social" situations; the JPP was a 12-trait rating scale, with scale values ranging from "never" to "always," on which a job was rated. On the Social Situations Test, each subject was asked to name the man he most liked to work with and then to predict how that particular man would respond to the tests then he was asked to name the man he least liked to work with and to predict how that man would respond to the test. On the JPP, subjects were merely asked to indicate how a man they could cooperate with best on the Job would describe the job, and then to indicate how a man they could cooperate with least well on the job would describe it. Thus, on the JPP, no specific prediction task was involved. Subjects named no names. All that was required of them was two general descrip tions. The correlation of .55 between the SS-ASo and JPP-ASo scores must be taken, therefore, as evidence of a possible general tendency for subjects to assume similar ity between "opposites," when they are asked to predict or describe how those opposites would respond to test material. The subject who predicted that a man he liked 98 to work with and a man he did not like to work with would respond similarly to the Soc* Sit. test would also generally predict that the kind of man he could cooperate with easily on the job would respond similarly on the JPP to the kind of man he had trouble cooperating with on the job. Conversely, if opposites were predicted to respond differently to the Soc. Sit. test, they were, on the average, predicted to respond differently to the JPP. The SS-ASp scores were derived from the differences between subjects* "personal choices" on the Soc. Sit. test and the choices they predicted the co-worker they liked most would make on the Soc. Sit. test. The SS-ASo scores were derived from the differences between the subjects* predicted-choices for their most-liked co-worker and their predicted-choices for their least-liked co-worker on the Soc. Sit. test. The SS-ASp and SS-ASo scores were D scores that had one element in common, therefore: the profile of the Soc. Sit. test responses predicted for the most-liked co-worker. This overlap between the two scores indicated that the .31 correlation observed between them was probably spurious, but how spurious? And in what direction? This problem of the amount and direction of the spuriousness in the correlation between two D scores computed from a common profile of test responses had not, to the writer*s knowledge, been discussed anywhere in the 99 literature, and so a preliminary analysis was attempted here, A D score is a ’ ’difference" score computed between two "profiles’ 1 of test responses* A profile is a graphic representation of scaled responses to an ordered series of tests or test items. If two D scores are computed from three different profiles, they have one profile in common* The relative size of the differences upon which the two Dfs are based would seem to be partly a function of the elevation of the common profile in relation to the eleva tions of the other two profiles* It would seem that If the common profile were higher or lower on the average than either of the other two profiles (i.e., if the mean of the mean responses that averaged across all subjects make up the common profile were higher or lower than the means of the mean responses that make up the other two profiles), then the differences between the responses that make up the other two profiles and the responses that make up the common profile should, within the limits of chance, covary. In other words, when the common profile Is either higher or lower on the average than either of the other two profiles used In computing the D scores that are to be correlated, there should be a positive correlation between those D scores even if nothing but chance is operating* But if the common profile lies between the other two, then 1 0 0 it would seem that the differences between the common profile and the one above it and between the common profile and the one below it should, within the limits of chance, vary inversely. That is, if the common profile were between the other two and was closer to one than the other on the average, then the D scores computed from it should be negatively correlated. If the common profile were exactly between the other two and nothing but chance were operating, then it would seem that the correlation between the two D scores involved should, on the average, be zero. It is hypothesized here, therefore, that if nothing but chance plus the relative position of a common profile determined the correlation between D scores computed between that common profile and two other profiles, the correlation would be positive if the mean of the common profile were higher or lower than the means of the other two profiles, and the correlation would be negative if the mean of the common profile was between (but not equl- distantly between) the means of the other two. It was beyond the writer’s competence to formulate this analysis mathematically and thus determine, if it were Indeed possible, the amount of spuriousness to be expected in the correlation between two D scores derived from only three profiles of test responses given the mean 1 0 1 differences among the profiles Involved. An empirical demonstration was attempted, however. First, the means of the three Soc. Sit. teat response profiles described above were compared to see if a positive or negative relation would have been expected between the SS-ASp and SS-ASo scores if nothing but chance were determining the relation. The mean for the Soc. Sit. test "personal choices" was ij.1.58. The mean for the "predicted choices for the most liked fellow worker" was The mean for the "predicted choices for the least liked fellow worker" was 55*59* The mean for the common profile (predicted choices for most liked worker), therefore, was between the other two. If nothing but chance and the position of the common profile determined the correlation between the SS-ASp and SS-ASo scores, it would be expected that the correlation would be negative. As it was, the obtained correlation of .31 was positive above the .01 point of significance. This significantly positive correlation indicated either that the above hypothesis was fallacious or that the "true" relation between the two AS scores was greater than .31. That is, if having the common profile between the other two tended to induce a negative correlation a3 hypothesized, then if the mean differences between the profiles were removed the correlation between the SS-ASp 1 0 2 and SS-ASo scores should be greater than *31* To test the hypothesis, mean response values were determined for each subject’s "personal choices" on the Soc. Sit* test and for his "predicted choices" for the man he most liked and the man he least liked to work with. The mean response values were obtained by dividing the total score each subject was credited with on the three administrations of the test by the number of items, 2£, in the test. The value of each item response that each subject made on each of the three administrations of the Soc. Sit. test was then expressed as a deviation from his mean response value for each administration. For example, if a subject’s total score when he made his "personal choices" was lj.0.0, his item mean response value was 1.60. If the value of his response to item 1 (determined by the rank values of each alternative) was 2.^0, then the deviation value of that response was 0.90* D scores were recomputed using the item deviation scores. For example, If the above subject’s total score when he made his "predicted choices" for his most-liked co-worker was then his item mean response value for that administration of the Soc. Sit. test was 1.80. If the value of his predicted choice for item 1 was 1*30, then the deviation value of that prediction was 0.50. The squared difference between the deviation value of his 103 personal choice (0*90) and the deviation value of his predicted choice for his most-liked co-worker (0*5>0) was 0.16. Squared deviations like this summed across the 2$ items of the test were the basis for the recomputed SS-ASp scores. The two sets of profiles from which the D scores were recomputed, which still had one profile in common (predicted choices for most-liked fellow worker), now had the same mean. Thus, according to the hypothesis advanced, no artifactual correlation would be induced between the D scores. When these new D scores, labeled "SS-ASpD" and "SS-ASoD,” where the nB” indicated they were computed from deviation scores, were correlated, the resulting coeffi cient was .1^3. The significance of the difference between the coefficient of *31 obtained between the SS-ASp and SS-ASo scores and this coefficient of . l j - 3 could not be tested by any method known to the writer, but the change was in the predicted direction, and so the hypothesis that the direction of the correlation induced between D scores computed from three profiles was dependent on the elevation of the common profile relative to the other two profiles received some support. Further support was given by results reported later in this chapter. The .31 correlation between the SS-ASp and SS-ASo scores in the face of artifacts that may have reduced the lolf true relation provided some support for Hypothesis II* The relation between these two kinds of AS scores (ASp vs* ASo) was not significantly different from zero when the scores were derived from the responses to different test instruments, however, for the correlation between the SS-ASp and JPP-ASo scores was only .lij.* To sum up the results of the tests of Hypothesis II, there was no equivalence between subjects* tendencies to assume similarity between their description of their job and their supervisors * description of the same job (JPP-AS) and their tendencies to assume that if they liked or disliked to work with another subject that subject would like or dislike to work with them (U-Socio-AS and W-Socio-AS), and there was no equivalence between any of these three AS scores and the other three AS scores (SS-ASp, SS-ASo, and JPP-ASo)* Subjects' tendencies to assume similarity between the responses they predicted for Others and their own responses were not consistent, there fore, from one operation to the next* Only when the predictions were made for ”opposites" (SS-ASo vs* JPP-ASo) was the tendency consistent from operation to operation. When the predictions were made in response to the same items (SS-ASp vs. SS-ASo), there was some indication that the tendency to assume similarity between the responses of a liked co-worker and the self oovarled with the tendency 1 0 £ to assume similarity between the responses of a liked vs. a disliked co-worker, but the overlap in the variance of the two AS scores was no more than 18 per cent. The Relations between Social Perception and Assumed Similarity Measures and Measures of Job Satisfaction and Job Performance The social perception (SP) and AS measures were obtained from men doing the same job as part of their dayfs work. All the measures were based either on attitudes toward the job or toward the fellow workers. It was hypothesized, therefore, that if reliable differences in certain abilities or tendencies were being measured by the SP and AS scores, those scores would then be related to measures that reflected either satisfaction with the job or job proficiency. To measure job satisfaction and performance, four scores were used. A job satisfaction scale (JSS) was assembled from items used in previous Air Force question naires. Following an analysis by Morsh (1957)* the items were divided into three subtests designed to measure "Sense of Ach.," "Attitude to Supervisor," and "Attitude to Co-workers." Only the subtest scores were used (i.e., no "total" 3cores were used), since each proved reliable and the intercorrelations among the three were not 106 high*^ A Job performance criterion (JPC) was assembled as the unweighted composite of three different criteria: combined mean-peer and supervisor ratings of over-all Job proficiency; combined self and supervisor ratings of leadership in performing 27 different tasks; and self ratings corroborated by supervisors of the amount of supervision needed in performing 130 different tasks. Hypothesis III The ability of subjects to describe a social object or predict the response of other persons accurately or the tendency to assume that the responses of others will be similar to their own responses in the same situation are related to their effectiveness or satisfaction as workers (as an indication of social perceptiveness, personal adjustment, or whatever), and so measures of those abilities or tendencies will be correlated positively with measures of job satisfaction and/or Job performance above the .0£ > level of significance. Results of the test of Hypothesis III The Intercorrelations of the SP and AS scores with the job satisfaction and Job performance measures are given in Table V. The TJ-SocIo-AS scores were not included ^Cf. the basic intercorrelation matrix in Appendix A. 107 TABLE V RELATIONS BETWEEN SOCIAL PERCEPTION AND ASSUMED SIMILARITY MEASURES AND MEASURES OF JOB SATISFACTION AND JOB PERFORMANCE Social Perception and Assumed Similarity Measures Job Satisfaction Job Performance Ach Att-S Att-W JPP-raw D .10 *1£ .01; .19 JPP-standard D .22 .16 .11 •33 d JPP-Acc -.08 .07 -.01; .11 U-Socio-Acc -.18 *05 -.09 .03 W-Socio-Ace .09 .08 .08 *43b SS-Acc .09 .00 .22 .18 JPP-AS • 37b ♦3ifb •30b *l;2b W-Socio-AS .22 .02 .23a -.01; SS-ASp *05 .05 ♦ 25a .06 SS-ASo .09 .14 .16 .09 JPP-ASo ♦07 .10 .19 -.19 Significant beyond the .05 point. ^Significant beyond the *01 point. 108 because they correlated *80 with the W-Socio-AS scores. Since both these scores were derived from the same ratings, It was assumed they measured the same thing. "Aeh" stands for "Sense of Achievement," "Att-S" for "Attitude toward Supervisor," and "Att-W" for "Attitude toward Co-workers." Of the l|4 coefficients in the table, 8 were signif icant beyond the .05 point of significance. Three scores correlated above the .01 point of significance with the composite criterion of Job performance, and one of these, the JPP-AS scores (which measured the tendency of subjects to assume that their supervisors would describe their job as they had described it), also correlated with the three JSS subtest soores above the .01 point of significance. Two other AS scores, the W-Socio-AS scores (which measured subjects' tendencies to assume that Others liked or disliked to work with them if they liked or disliked to work with the Others) and the SS-ASp scores (which measured subjects1 tendencies to assume that co-workers they liked to work with would respond to the Soc. Sit. test as they did), were correlated with the "Attitude to Co-workers" scores above the .05 point of significance. Discussion of the results of the test of 'hypothesis Jit The advantage in adjusting the JPP distance scores 1 0 9 and the Soclo-Acc scores for hypothesized response sets that might have reduced their "psychological" meaning was demonstrated by the correlations with the JPG scores, The JPP-standard D and W-Socio-Acc scores both correlated with the JPC scores beyond the ,01 point, but the correlations of the JPP-raw D and TJ-Socio-Acc scores with the JPC scores were not significantly different from zero. The significant correlations of the JPP-AS scores with the JSS subtest scores and the JPC scores indicated that subjects who assumed their supervisors would describe their job in much the same way they had described it had more sense of achievement on their job, had more positive attitudes toward their supervisors and co-workers, and had higher scores on the JPC, Since the three subtests of the JSS all correlated positively with the JPC beyond the ,01 point (,ijlj., *37, and ,3! } . , respectively), it seemed probable that the JPP-AS scores reflected a generally positive attitude toward the job. Two other AS scores, the W-Socio-AS and SS-ASp scores, also correlated signifi cantly with one of the JSS scores, the Att-W scores* These correlations indicated that subjects* tendencies to assume similarity between themselves and their fellow workers was related to positive attitudes toward those 110 fellow workers as measured by the JSS.-^ Since the JPP-standard D scores measured subjects’ differential accuracy in describing the "average percep tion" of their job within their squadrons, their correlation with the JPG scores would seem to be explicable only in terras of some intervening variable. No specific variable was hypothesized, but if there was covariance between ability to approximate a modal profile-description of a job and ability to do that job, then it was possible that some common perceptual variable was involved. No such common perceptual variable needed to be invoked to explain the correlation between the W-Socio-Acc scores and the JPC scores, but a discussion of that correlation will be postponed to the last section of this chapter, which deals with some non-hypothesized relations observed in this study. ^This "positive attitude" aspect of AS scores was suggested by the findings of Fiedler (1952) and Davltz (1953)» who had showed that subjects assumed more similarity between themselves and persons they liked than between themselves and persons they did not like. But the correlations obtained here between the JPP-AS, W-Socio- AS, and SS-ASp scores and various job satisfaction scores indicated in a more conclusive way that subjects’ tend encies to assume similarity between themselves and Others were partially contingent upon positive attitudes or feelings toward those Others, since the AS and SP scores were independent from the JSS scores, whereas Fiedler’s and Davitz’s findings were based on relations that did not represent independent observations. Ill The Predictive Validity of Social Perception and Assumed Similarity Measures for the Job Performance Criterion as Compared with the Predictive Validity of Conventional Predictors for that Criterion In addition to the SP and AS "tests," various conventional aptitude, achievement, and interest tests were administered to the experimental subjects* Since the interrelations of most of these variables with the JPC had been determined (the "basic correlation matrix" of 20 of these variables may be seen in Appendix A), it was possible to compute the partial regression coefficients for the variables in several of their multiple relations with the JPC, and from these beta coefficients to deter mine the multiple correlations of sets of predictor variables and the JPC. For two regression analyses (one reported in this section and one in the section to follow), the Doolittle technique was used* For the other regression analyses, GengerelliTs (19^8) method for approximating multiple regression coefficients and multiple correlation coeffi cients was used* Hypothesis IV If, as Implied in Hypothesis III, the ability to 1 1 2 make accurate social perceptions or the tendency to assume similarity between self and Others were indications of personal adjustment, social perceptiveness, or some other personality or temperament variable, the experimental measures of that ability or tendency would predict a significant amount of the variance in a criterion of job effectiveness over and above that predicted by conven tional cognitive variables (i.e., the multiple R between the criterion and conventional predictors plus experi mental measures of social perception or AS would be significantly larger than the multiple R between the criterion and conventional predictors alone). Results of the test of Hypothesis IV The multiple correlation between six conventional predictors (Pattern Comprehension, Word Knowledge, Visual Attention, Reading Comprehension, Mathematics, and Interest in Electronics) and the JPC was .28?, using Gengerelli*s approximation method. With four SP and AS measures (JPP-standard D, W-Soeio-Acc, SS-ASo, and JPP- ASo) added to the multiple prediction battery, the multiple R with the JPC was .607» The difference between the R»s of .287 and *6o7, when tested with the F ratio as recommended by Guilford (1956* p. ij-00), was significant beyond the .001 point. Thus the hypothesis that SP and AS 113 scores would account for a significant amount of the variance in job performance over and above that accounted for by conventional predictors was given strong support. Having thus established the over-all contribution of SP and AS variables to the prediction of the JPC, the Doolittle technique was employed to determine the differ ential contribution of (1) selected SP and AS measures, (2) the three JSS subtest scores, and (3) the Visual Attention and Social Situations tests. These variables were selected because, In preliminary calculations using the Gengerelli technique, they proved to be the most useful additions to the multiple prediction battery. The beta weights and per cents of the variance in the JPC accounted for by each variable are given in Table VI. If the per cents of variance accounted for by the variables are changed to proportions and summed and the square root of this sum taken, the resulting coefficient gives the multiple R between the 10 predictors and the JPC, .7213* Since the SP and AS predictors alone account for 26.83$ of the variance in the JPC (cf. Table VI), the multiple R between them and the JPC is *5>l80. Discussion of the results of the test of Hypothes'Is TV The superiority of the SP and AS predictors over conventional aptitude, achievement, and interest llt+ TABLE VI BETA WEIGHTS AND PER CENTS OP JPC VARIANCE ACCOUNTED FOR BY SELECTED SP, AS, AND APTITUDE PREDICTORS Predictor Beta Weight Per Cent of JPC Variance Predicted Social Situations • 0089 .Ik Visual Attention • l6l8 3.92 JPP-standard D .1752 5.8$ W-Socio-Acc .3056 13.16 SS-Acc • 06I 4 . 9 l.ll+ SS-ASo -,ll+07 1.28 JPP-ASo .2808 5.1+0 JPP-AS .0271 1.11+ JSS-Ach .2517 11,21 JSS-Att-S .11+21 5.29 JSS-Att-W • 1019 3.14-9 predictors for the criterion of job performance used In this study was quite evident* The experimental subjects were selected for the Job they were doing on the basis of scores on conventional cognitive prediotors, however, and so the range of their scores on the cognitive predictors used in this study was reduced. The means and standard deviations of the scores of subjects In this sample on the tests from Morshfs Factor Reference Battery and on six tests from the Air Force Classification Battery Eire compared in Table VII with the scores of unselected Basic Airmen on the same tests (cf. Morsh, 1957)* In all cases the means of the 5-level K-system mechanics used in this study as subjects were higher than the mean3 of unselected Basic Airmen on the same tests, and the dispersions of the K-mechanics scores were in all cases narrower than those of the Basic Airmen. It was apparent, then, that there was a consider able restriction in range on cognitive variables for subjects used in this study. The correction of correla tion coefficients for restriction in range was not used, however, because none of the three Factor Reference tests used In this study correlated above .09 with the JPC* Further, the generally moderate to substantial correla tions among the conventional predictors Indicated that a restriction in range along could not account for the lack 116 TABLE VII MEANS AND STANDARD DEVIATIONS OP SCORES OP SUBJECTS IN THIS SAMPLE AND UNSELECTED BASIC AIRMEN ON NINE AIR FORCE TESTS Test 5-level K-mechanics Basic is Mean S. D. Mean S, D. Pattern Comprehension 10.91 2.83 10.52 3.19 Word Knowledge (Factor Reference Battery) 16.23 U-90 11.30 6.3^ Mechanical Information 15*13 5.29 9.^8 Word Knowledge (AC-1B) 6.6l 1.71 5.00 1.96 Arithmetic Reasoning 5*95 1.90 5.00 1.96 Dial and Table Reading 6.01; . 1.33 5.00 1.96 Numerical Operations 5.18 1.67 5.00 1.96 Electrical Information 6.80 1.52 5.00 I.96 Mechanical Principles 6.39 1.71 5.00 1.96 117 of relation between the predictors and the JPC* To illustrate this point, the intercorrelations of the 8 conventional aptitude, achievement, and interest tests and the JPC are given in the matrix in Table VIII. All of the intercorrelations among the predictors are positive, and 18 out of 28 are significant above the .05 level; but only one of the predictor correlations with the JPC is significantly different from zero. In short, variance on the JPC simply was not predictable from scores on conventional cognitive or interest tests, but over 25$ of that variance was predict able from scores on SP and AS tests. This result did not indicate what the SP or AS scores measured, but it did indicate they measured something relevant to job perform ance. Since the JSS subtests also measured something relevant to job performance, and since 5 of the 15 correlations between AS scores and JSS scores were significantly different from zero (see Table V), It was possible, as indicated earlier, that the AS scores measured some attitude(s) relevant to job performance. But there was no clue to this point as to what the SP scores measured. For this reason, some nonhypothesized relations were investigated next in an effort to determine something more of the meaning of SP scores. 118 TABLE VIII INTERCORRELATIONS AMONG APTITUDE, ACHIEVEMENT, AND INTEREST TESTS, AND THE JPC Pattern Comp, (PC) Word Knowl. (WK) *39 Mach. Info• (MI) .31 • 67 Visual Attn, (VA) .37 .23 .15 Math, (M) .55 .1*8 .22 .1*0 Read, Comp, (RC) .1*5 .1*0 .1*8 • 35 • 37 Interest (Int) .23 .22 .01* .29 .1*5 .37 Soc, Sit, (SS) .16 • ll* .07 .21 .01 .23 JPC .09 • 00 -.02 .21* .00 .09 PC WK MI VA M RC .13 .16 .16 Int SS 119 Some Nonhypothesized Eolations between the Social Perception and Assumed Similarity Measures, Sociometric Popularity, Job Satisfaction, and Job Performance In this section, an investigation of some non hypothesized relations between SP and AS scores will be reported, both for scores derived from a common profile and scores derived independently* In addition, the Investigation of some nonhypothesized relations between sociometric popularity and job performance will be reported* The research reported to this point wa3 undertaken because there was little or no evidence in the literature that SP measure 1 was equivalent to SP measure 2, or that AS measure A was equivalent to AS measure B* The purpose of this study was to investigate the equivalence of the various operations used to measure SP and the equivalence of the vti; ' operations used to measure AS. Some of the correlates of these measures were also Investigated, since previous studies had established that SP or AS scores frequently did measure something related to nontest behavior variables. But it did not occur to the writer to investigate the relations between SP and AS scores until It became apparent from a final survey of the literature that such relations had not been adequately explored. For 1 2 0 example, Fiedler performed numerous studies (19^9* 1952, 195lj., 195>8, 1959) in which SP as well as AS data were available to him, yet he never made use of the SP data. Gage (1952) actually obtained both SP and AS scores, yet he reported none of their interrelations. Only Norman and Ainsworth (195^)* w^° correlated "projection" (an AS score) with "empathy" (an SP score), and Hastorf and Bender (1953)* were interested in the effect of "projection” on "empathy," investigated relations between AS and SP scores. But CronbachTs critical articles (1955* Gage and Cronbach, 1955) apparently discouraged any further investigations: Norman and Ainsworth report a large number of correlations between Accuracy ("empathy") and Assumed Similarity ("projection"). Since the Accuracy score contains Assumed Similarity components, the two scores would necessarily overlap even if both sets of responses are determined strictly by chance. The correlations have no psychological meaning. (Cronbach, 1955* P* 180J The Bender-Hastorf refined empathy score is equal (the count of) Warranted Assumed Dissimilarity (items) minus Unwarranted Assumed Similarity. Therefore, the refined empathy score has a perfect negative correla tion with Assumed Similarity, when Real Similarity is held constant. (Gage and Cronbach, 1955* P* if-16} The AS components that Cronbach said (1955) were components of the accuracy (or SP) score were the "assumed dispersion in elevation" (ADE) and the "assumed dispersion within items" (ADI). When a subject predicted the responses of two or more Others to two or more items, the 1 2 1 ADE component in either the accuraey-of-predietion or the assumed-similarity scores that were derived from the comparison of the predictions with either the Others* own responses or with the subject’s own responses represented the differences in elevation {mean differences) of the predicted profiles of responses* That is, the ADE compo nent was the variance of the means of the scores predicted for Others* The ADI component represented the differences in responses predicted for specific items after the over-all differences between the responses predicted for Others had been removed. The variance represented by ADI was variance in predicted responses that could not be accounted for by variance in the total scores predicted for Others* Cronbach demonstrated that the ADE and ADI components could be Identified in the accuracy-of- prediction (SP) and assumed-similarity (AS) scores derived from the predictions of the same subjects for the same Others on the same items. If SP and AS scores derived from such predictions were correlated, Cronbach was probably right that part of the correlation observed would be spurious. But his mathematical description (or "model," as he called it) of SP and AS scores derived from predic tions of the same subjects for the same Others on the same items did not fit the data obtained for this dissertation* 1 2 2 The six SP and six AS scores that were the main independ ent variables in this study were not derived from predictions of the same subjects for the same Others on the same items. The JPP-Acc and SS-Acc scores were derived from subjects' predictions of the responses of one Other on 12 and 25 items, respectively. The W-Socio-Acc and U-Socio-Aco scores were derived from subjects' predictions of the responses of from 13 to 19 Others on one item only (the item: predict which of your co-workers has indicated that they like or do not like to work with you). The JPP-raw D and JPP-standard D scores were derived from a comparison of subjects’ job descriptions with a single criterion Job description. The JPP-AS, JPP-ASo, SS-ASp, and SS-ASo scores were derived from predictions of the responses of one Other on 12 or 25 items. The W-Socio-AS and U-Soeio-AS scores were derived from subjects' predictions of the responses of from 13 to 19 Others on one item only. Cronbach's model required that the subjects' predictions be made both for more than one Other and on more than one item, and so his model did not fit the SP and AS scores that were used in this study. Because Cronbach'3 model did not fit the SP and AS scores used in this study, those scores did not have 123 components In common.1^ There were two kinds of SP and AS scores available for analysis in this dissertation, those derived from the predictions of the same subjects for the same Other (or Others) on the same items (or item), and those derived from the predictions of the same subjects for a different Other (or Others) on different items (or item)* The first part of this section will deal with the relations investi gated among the former type of AS and SP scores, and the following part will deal with the relations investigated among the latter type of scores. Relations among social perception and assumed similarity scores derived from a common base To say that the SP and AS scores used in this study that were derived from the responses of the same subjects on the same measures did not have components in common was not to say that they had nothing in common. The JPP-AS and JPP-Acc scores, for example, were D scores computed from the differences between each subject's JPP job-description and his predicted-supervisor JPP job- description, and the subject's predicted-supervisor JPP Job-descrlptlon and the supervisor's JPP job-description, See Appendix B for a more technical discussion of why the SP and AS scores used in this study did not have ADE and ADI components in common. 12k respectively. Thus the D scores were computed from the common base of the subjects predicted-supervisor JPP job-description. The spuriousness of a correlation between D scores computed from a common profile was discussed earlier in connection with the relation between SS-ASp and SS-ASo scores, and it was hypothesized that the direction of the spurious "overlap” could be either positive or negative depending on the position of the common profile relative to the other two. If the common profile were either higher or lower than the other two on the average, the Induced correlation would, other things being equal, be positive; if the common profile lay somewhere between the other two on the average, the induced correlation would, other things being equal, be negative. In the case of the SS-ASp and SS-ASo relation, the latter condition held, yet the obtained correlation was "significantly” positive. It seemed apparent that "other things" were not equal; i.e., there was a definite tendency for subjects who assumed similarity between themselves and persons they liked to assume similarity between persons they liked and persons they did not like. This tendency became more apparent when the mean differences were removed from the three profiles from which the SS-ASp and SS-ASo scores were computed, for the correlation between the SS-ASpD and 125 SS-ASoD scores was .1+3. But the p-value of the correla tion could not be assessed, since the writer could neither formulate a mathematical proof for his hypothesis that; the contribution of spuriouaness due to a common profile to the direction of the correlation between two D scores depended on the average position of that common profile relative to the other two, nor estimate how much spurious ness a eommon profile was likely to add to or subtract from the correlation, except by the empirical method employed• To obtain further empirical evidence with which to evaluate this hypothesis and to get another estimate of the amount of spuriouaness induced in correlations between the type of distance scores used here when the scores were computed from a common profile, t?^ SS-Acc and SS-ASp scores were recomputed. Mean response-values were obtained for the subjects* "personal choices" on the Social Situa tions Test, and for the subjects* "predicted-choice3n for the man they most liked to work with. Then, for the 25 items in the test, each subject's personal-choice item response-values were subtracted from his mean personal- choice response-value, and his predicted-choice item response-values were subtracted from his mean predicted- choice item response-values. This gave the responses a new value: a value that took into account each subject's 126 mean level of performance on the test. These new Item response-values represented deviations from each subject's mean response-value when he made his personal choices and when he made his predicted choices. With the mean differences between the three Social Situation Test profile scores that were used to compute the SS-Acc and SS-ASp scores thus removed, ^ the SS-Acc and SS-ASp scores could be recomputed and compared with the original scores computed between profiles that had different means. The recomputed scores were labeled "SS-AccD" and SS-ASpD." The correlation between the SS-AccD and SS-ASpD scores was . i | . 6. This may be compared to the correlation of *7^ between the SS-Acc and SS-ASp scores. The correlation was thus reduced when the mean differences were removed between the profiles from which the scores were computed. This was further support for the hypothesis that the direction of induced correlation between D scores computed from a common profile was a function of the relative elevation of the common profile, because, in this case, the common profile was higher than either of the other two. The difference between the ^^The Others1 "personal choices" were recomputed in the same operation as the subjects' "personal choices," since in all cases the Others were also subjects. The means of the two profiles did not coincide since some subjects were others as many a3 four times, whereas some were never chosen as preferred fellow workers. 127 original correlation of .7^ and the correlation of J+6 between D scores computed from profiles with the same elevation indicated the amount of spuriousness induced in this particular case, where the mean for the subjects' "personal choices" was i j . 1,58, the mean for the subjects’ "predicted-Other choices" (the common profile) was ij.5^7* and the mean for the Others' "personal choices" was 1^.1.i ^ - 5 * The SS-ASo and SS-Acc scores also shared a common profile, the 25 predicted-choices on the Soc. Sit. test for the most-liked fellow worker. The mean differences had already been removed between that profile and the other two used in computing those scores, and so the SS-ASoD and SS-AccD scores were correlated in a further te3t of the hypothesis that the relative elevation of the common profile determined the direction of the correla tion induced between D scores computed from that profile. The correlation between the SS-ASo and SS-Acc scores was .22, a coefficient that barely missed significance at the • 05 level. The mean score for the subjects’ predicted- choices of their most-liked fellow worker on the Social Situations Test was l^5.1j-8; the mean of the personal-cholces of those most-liked fellow workers was I j . l.lj.5; and the mean for the predicted-choices of the least-liked workers was 55-59» Th© common profile was between the other two, and so it was predicted that a negative relation had been 1 2 8 Induced* The correlation between the SS-ASoD and SS-AccD scores was *37* The prediction was correct. The analyses of the relations between the SS-ASpD and SS-AccD scores, the SS-ASoD and SS-AccD scores, and the SS-ASpD and SS-ASoD scores indicated that AS and SP scores computed from a common profile might profitably be compared if the mean of that profile relative to the other two were known. The SS-ASp vs. SS-Acc relation has already been discussed in part. The correlation between these scores was .7bt but since the common profile was higher than the other two, a positive correlation was induced; it was determined that without this spurious element the correlation was . l j . 6. But this correlation of •I4 . 6 was still significantly different from zero. More than 20% of the variance in the SP (or interpersonal) scores was thus accounted for by the AS (or intrapersonal) adores. The correlation between the SS-ASo and SS-Acc scores was not significantly different from zero, but when the effect of the mean differences between the profiles used to compute the scores was removed, the correlation went up to .37» which was significant beyond the .01 point. The multiple correlation coefficient between the SS-ASpD and SS-ASoD scores and the SS-AccD scores was .£0, and so the two intrapersonal AS scores accounted for 2%% 129 of the variance in the interpergonal SP scores. The correlation between the JPP-AS and JPP-Acc scores was •2$, barely significant at the ,0£ level, but the mean for the common profile, the subjects’ "predicted- supervisor” JPP job description, was between that of the subjects’ JPP job description and the supervisors’ JPP job description, the means being 63,^8, 55*^2, and 68,80, respectively. Thus a negative correlation was probably induced between the JPP-AS and JPP-Acc scores, and so the •25 correlation coefficient obtained was probably an •underestimation of the true relation between the two scores. The correlation between the JPP-AS and JPP-raw D scores was ,f?2, and the correlation between the JPP-AS and JPP-standard D c i-ores was Both these coeffi cients were significant well beyond the .01 point, and since the means of two of the profiles involved in the computation of the scores coincided (the JPP distance scores were both computed from the subjects JPP job descriptions, and so, except for slight between squadron differences, the elevations of the profiles were essen tially the same), no artifact could have been present in these correlations. There was, then, a fairly consistent moderate correlation between AS and SP scores computed from common 130 profiles when artifacts in the relations were either removed or taken into consideration: The JPP-AS scores correlated *52 and 3 with the JPP distance scores and .25 with the JPP-Acc scores, and there was good reason to believe the latter coefficient was an underestimation of the true relation; the SS-ASp scores correlated * 1 } . 6 with the SS-Acc scores when the mean differences between the profiles from which they were computed were removed; and the SS-ASo scores correlated *37 with the SS-Acc scores when the profile means used in their computation were reduced to zero* The consistency of these relations indicated that subjects* tendencies to assume similarity between their responses and the responses of Others had something to do with their ability to predict the responses of those Others when the AS and SP scores were derived from the same measuring instrument* These results were in substantial agreement with those reported by Norman and Ainsworth (195^4-) * Relations among social perception and assumed similarity scores not derived from a common base Twelve of the possible 25 intercorrelations among SP and AS scores not derived from a common profile are reported in Table IX. The other 13 correlations were not computed. The correlation coefficients that would have appeared in the cells marked "X” were presented in the 131 preceding discussion of the relations among SP and AS scores derived from common profiles* TABLE IX INTERCORRELATIONS BETWEEN SELECTED SOCIAL PERCEPTION AND ASSUMED SIMILARITY SCORES NOT DERIVED PROM A COMMON PROFILE Assumed Similarity Scores Scores JPP-AS W-Socio-AS SS-ASo JPP-ASo JPP-standard D X .08 • Oif i . o H JPP-Acc X -.1 11. -.02 -.12 W-Socio-Acc .lj.lb X .0^ .11* SS-Acc .03 .16 X .Ik Significant beyond the .01 level, One of the coefficients computed was significant beyond the .01 level, that between the JPP-AS scores and the W-Socio-Acc scores. This indicated that the subjects* tendencies to assume similarity between the way they described their job and the way they predicted their supervisors would describe their job was related to sub jects1 abilities to predict who liked to work with them. This was the only evidence of any relation between SP and AS scores when those scores were derived from different measuring instruments. 132 Some nonhypothesized relations between sociometrio popularity and .job performance Since conventional sociometrie ratings had been obtained in the course of getting the data needed to compute SP and AS scores from the sociometric predictions, it was possible to compute an index of popularity as a worker (Socio-Pop) for each subject. The W-socio-Acc scores were significantly related to two of the other measures employed in this study, the JPC and the JPP-AS scores, and they acoounted for more of the variance in the JPC (13.16$) than any other single predictor. It seemed advisable to determine the extent to which these relations could be acoounted for simply by knowing the Socio-Pop score of each subject. Accordingly, the effect of Socio-Pop was partialled out of the rela tions between the W-Socio-Acc scores and the JPC and JPP-AS scores, and the Doolittle solution for the partial regression coefficients of the 10 predictors in their relation to the JPC described previously was expanded to include the Socio-Pop scores* The W-Socio-Acc scores correlated .J^3 with the JPC and .67 with the Socio-Pop scores, and the latter corre lated .69 with the JPC scores. When the effect of Socio- Pop was partialled out, the correlation between W-Socio-Ace and JPC dropped to -.06. The fact that Socio-Pop 133 accounted fox* all the variance in the JPC formerly accounted for by W-Socio-Acc was striking evidence in support of Cronbach's (1955) thesis that "variable processes" (or interactions) should be investigated only after the "constant processes" (main effects) had been thoroughly studied* In effect, this meant that as much (or more) evidence about Individual differences in job performance could have been obtained by having the subjects fill out the Personal Preferences blank only as by having them fill that blank out and then try to predict ■shich of their fellow workers had indicated that they liked or did not like to work with them. The effect of Socio-Pop not only on W-Socio-Acc but on other SP measures and on AS measures was demon strated by the results of a Doolittle solution for the partial regression weights of a prediction battery for the JPC that included the Socio-Pop scores. These results, given in Table X, may be compared with the results given in Table VI, which, for convenience, are reprinted in the parentheses of Table X. The results were not strictly comparable, because the JPP-AS scores were not included in this second Doolittle solution; but since the beta weight for the JPP-AS scoreB was only *0271, their inclusion could not have affected the solution appreciably. The beta weights and per cent variance in the JPC TABLE X BETA WEIGHTS AND PER CENTS OP JPC VARIANCE ACCOUNTED FOR BY SELECTED SP, AS, AND APTITUDE SCORES, PLUS THE SOCIOMETRIC POPULARITY SCORES Predictors Beta Weights Per Cents of JPC Variance Predicted Visual Attention .0919 ( .1618) 2.20 3.92) JPP-standard D .1534 ( *1752) 5*06 5.85) W-Socio-Acc -.0160 ( .3056) -O.69 13.16) SS-Acc v .0882 ( ,06i | _ 9) 1.59 l.il*-) SS—ASo -.0815 (-.llj.07) 0.73 1.28) JPP-ASo .1571*. ( .2808) 2.99 5.i*-0) JSS-Sense of Ach .2I 4 . 8I ( .2517) 11.16 11.21) JSS-Attitude to Supervisor .1120 ( .lij.21) 1 +* ll+ 5.29) JSS-Attitude to Co-workers .oltflj. ( .1019) 1.6l 3*^9) Social Situations -.OI4 . 8I ( .OO89) -0.7 2 O.lij.) Socio-Pop .5338 { — ) 36.83 - ) 135 accounted for by the variables used In the first multiple regression analysis all went down in this second analysis, except those for the SS-Acc scores. This wholesale reduction in independent predictive values occurred beoause the Socio-Pop scores were not only highly related to the JPC and W-Socio-Acc scores but were also signifi cantly related to the SS-ASo scores and positively related to all the other predictors. The key relation in these analyses was the .69 correlation between the Socio-Pop scores and the JPC. The criterion, it will be remembered, was made up of five kinds of ratings grouped as three components: (1) the combined mean-peer and supervisor proficiency ratings; (2) the combined self and supervisor task leadership ratings; and (3) the self (corroborated by supervisor) TCL amount of supervision ratings. The Socio-Pop correlations with these three components were .77» *b9* .59* respectively, and the Socio-Pop correlations with the self task leadership ratings was .55* The Socio-Pop scores correlated higher, therefore, with those criteria made up of supervisor and mean-peer ratings and lower with those criteria made up of self ratings.^ It would appear, 13 Wherever there had been an unresolved disagree ment between a subject's TCL amount of supervision rating and his supervisor's rating, the subject's rating was taken as the better estimate of the true value. 136 then, that one-half the JPC was a sort of "popularity rating"; I.e., considering the high relation between popularity as a worker and those criteria made up of mean- peer and supervisor ratings, those ratings were hardly more than an expression of that perceived popularity. But it must be remembered that the experimental subjects were asked on the Personal Preferences blank to list those fj-level mechanics they really liked and did not like to work with. Those mechanics who were preferred, therefore, may have been preferred because they were the best men on the job and not because they were "nice guys," easy to get along with, or whatever. Unfortunately, the reasons for the preferences were not determined, because this was not intended to be a study of sociometric popu larity. Great as the effect of popularity as a worker was, it ended with the relations described above. The . 1 } . 1 correlation between the W-Socio-Acc scores and the JPP-AS scores, for example, was not simply a function of the relation between Socio-Pop and JPP-AS, since the Socio-Pop scores correlated only .II4 with the JPP-AS scores. Thus, when the effect of Socio-Pop on the correlation between W-Socio-Acc and JPP-AS was partialled out, the coefficient was raised to .1 ) 4. Before leaving this chapter on the results of this 137 study, It probably should be observed again that the AS scores were not the product of a dyadic or interpersonal analysis. Rather, they were the product of an lntra- personal analysis: the distance between two (or more) scores provided by the same subject was computed* For this reason, the AS scores were not, strictly speaking, SP scores, for they were obtained independently from evidence of any real social interaction. There were, of course, degrees of ”abstraction" in the AS scores: the SS-ASp and JPP-AS scores reflected the similarity between sub jects and particular persons, whereas the JPP-ASo scores reflected the similarity between two abstract persons. But regardless of their degree of abstraction, AS scores always measured a within-subject tendency; they did not measure a necessarily direct product of a social inter action. CHAPTER IV SUMMARY AND CONCLUSIONS Two kinds of scores, "accuracy-of-prediction of Other's responses” and "assumed-similarity to Other's responses," and four general methods, "predictions of individual Other's responses vs* prediction of group (or stereotype) responses" and "similarity between self and Other vs. similarity between predicted Others (opposites),” have been used to measure interpersonal or social percep tion. When this dissertation was conceived, stereotype and differential accuracy scores were apparently in limbo, and only Fiedler was working with assumed-similarity scores.^-9 The purpose of the study was primarily to "decide the fate" of accuracy-of-prediction scores, and secondarily to Investigate the generality of AS scores. In addition some of the correlates of the two kinds of scores were investigated. An unexpected development was that relations between the accuracy-of-prediction scores and AS scores Invited some Investigation. Six experimental accuracy-of-prediction scores were ■^Elizabeth Alfert (19f?8) did an interesting analytical study of AS scores, which Is mentioned later. 138 139 obtained from a sample of 70 Air Force electronics techni cians: 1* Stereotype accuracy* a. "JPP-raw D": Subjects' accuracy in perceiv ing a job as it was perceived by other members of their work group (comparison of absolute differences in perception). b. "JPP-standard D": Subjects' accuracy in perceiving a job as it was perceived by other members of their work group with individual differences in the elevation and scatter of perceived profiles removed (comparison of relative differences in perception)• 2. Differential accuraoy. a. "JPP-Acc": Subjects' accuracy in predict ing supervisor's perceptions of a job. b. "SS-Acc": Subjects' accuracy in predicting preferred-co-worker's responses to a test. c. "U-Socio-Acc" and "W-Socio-Acc": Subjects' accuracy in predicting co-workers socio metric preferences (both weighted and unweighted scores obtained). Six experimental AS scores were also obtained: 1. "JPP-AS”: Subjects’ tendencies to assume similarity between their perceptions of their job and their supervisor’s perception of their job. 2. "SS-ASp": Subjects' tendencies to assume similarity between their own responses to a test and the responses of their most-liked co-worker. 3. "SS-ASo": Subjects' tendencies to assume similarity between the responses of their most- and lea3t-liked co-workers on a test. iko i j . . "JPP-ASo": SubjectsT tendencies to assume similarity between the perceptions of their job by co-workers they could cooperate with best and least well. £. "U-Socio-AS” and "W-Socio-AS": Subjects* tend encies to assume similarity between their sociometric preferences of Others and the Others * preferences for them (both weighted and unweighted scores were derived). Not counting the relations between scores computed from the same responses (JPP-raw D vs. JPP-standard D, U-Socio-Aoc vs. W-Socio-Acc, and U-Socio-AS vs. W-Socio- AS), nor the intercorrelations among the sociometrio accuracy and AS scores (the latter were uninterpretable), the following intercorrelations of the accuracy-of- prediction and AS scores were significantly different from zero: JPP-raw D vs. JPP-Acc; JPP-Acc vs. TJ-Socio-Acc; JPP-raw D vs. JPP-AS; JPP-standard D vs. JPP-AS; JPP-Acc vs. JPP-AS; W-Socio-Acc vs. JPP-AS; SS-Acc vs. SS-ASp; SS-Acc vs. S S - A S o ;20 SS-ASp vs. SS-ASo; and SS-ASo vs. JPP-ASo. Of the $2 possible correlations, 10 were thu3 significant beyond the .0£ level. The correlation between the JPP-raw D and JPP-Acc scores (J+7) did not necessarily indicate a 22$ overlap 20 The correlation between these scores did not become significant till the mean differences between the profiles from which they were computed had been removed. The uncorrected correlation coefficient was .22 (.23 was required for significance at the .0$ level) and the corrected coefficient was *37» which was significant beyond the .01 level. Ikl between stereotype accuracy and differential accuracy, as measured by JPP ratings. The most "accurate” subjects as measured by both the JPP raw D and JPP-Acc scores tended to make higher JPP ratings on the average and to spread their ratings over a narrower range than did the least accurate subjects, and so it was possible that these rating tendencies alone— and not "accuracy"— could have accounted for the correlation observed. This possibility was strengthened by the fact that the correlation between the JPP-Acc and JPP-3tandard D scores, where the correla tive effect of such rating tendencies was inoperative, was only .15* The correlation (.2f>) between the JPP-Acc and U-Socio-Acc scores, both of which were raw scores, might also have been attributable in part to rating biases. Subjects with the most accurate JPP-Acc scores made higher predictions on the average and confined their predictions PI to a narrower range of the scale and subjects with the most accurate TJ-Socio-Acc scores tended to make fewer predictions than less accurate subjects. Any correlation between "conservatism" in making sociometric predictions and "conservatism" in using a rating scale (if high mean ^As Cronbach (1955) and Crow (1957) pointed out, such rating tendencies Insure greater accuracy of prediction. l i j . 2 ratings and narrow dispersions can be so labeled), there fore, would induce a correlation between "accuracy" in JPP predictions and "accuracy" in sociometric predictions* The correlations between the JPP-AS scores and the JPP-raw D scores (.52), the JPP-standard D scores (J4 . 3), and the JPP-Acc scores (.25) all were presumably affected by the fact that the D scores correlated in each case were computed from a common profile, but it was argued that there was evidently no real spuriousness Involved In the relations between the JPP-AS scores and the JPP-raw D and JPP-standard D scores, since there were no differences between the means of profiles used to compute the latter two scores; and it was further argued that whatever spuriousness there was in the relation between the JPP-Acc and JPP-AS scores had probably lowered the correlation coefficient computed, since the elevation of the profile common to the two scores was midway between the means of the other two profiles involved. The same argument was used with regard to the correlations observed between the SS-Acc and SS-ASo (*22) scores and between the SS-ASo and SS-ASp scores (.31), and it was pointed out that the • 7I 4 - correlation between the Ss-Acc and SS-ASp scores was probably less than this since the mean of the profile common to the two scores was higher than either of the other two profiles involved. ll+3 The ^2 correlations among accuracy-of-prediction and AS scores that were not significantly different from zero, the significantly positive correlations between the JPP-raw D and JPP-Acc scores and between the JPP-Acc and TT-Soclo-Acc scores, which probably could have been partly accounted for by the differential rating tendencies of the "accurate" and "inaccurate" subjects* and the other eight significantly positive correlations pointed to the following tentative conclusions within the limits imposed by the subjects and measures used: 1, Stereotype accuracy and differential accuracy probably do not covary. 2. Differential accuracy scores calculated from one set of predictions probably are not predictable from differential accuracy scores calculated from another set of predictions; i.e., differential accuracy is probably more a function of the measuring instrument used than the subject using it, 3* ASo scores calculated from two sets of predic tions (or descriptions) probably are predictable from ASo scores calculated from two other sets of predictions or descriptions* I j . * ASp and ASo scores calculated from self- descriptive and Other-descriptive responses to the same measuring instrument are probably predictable from one another* 5* AS, ASo, and ASp scores calculated from self- descriptive and Other-descriptive responses to different measuring instruments probably are not predictable from one another*22 PP ^Alfert’s (1958) conclusions were similar. She found that the content in the items used in a measuring 11* In general, then, It was concluded that one set of operations defining either accuracy in social percep tion or the tendency to assume similarity probably was not equivalent-even to a small per cent of overlap— to another set of operations supposedly defining the same trait or ability. But, when the same set of predictions was used as one component in deriving accuracy-of- prediction and AS scores, 1*> to 20% overlap probably existed between the variances of the two scores. Those who study the "process” of social perception should be interested to learn that a probable ingredient of the process Is the tendency to assume that Others are similar to oneself; those who wish to use interpersonal perception scores as predictors of performance should be interested to learn that an appreciable amount of the variance in those scores is predictable from intrapersonal AS scores. It was further demonstrated in this study that those who have used or wish to use social perception scores as predictors of performance probably will find some validity for them. The JPP-standard D, W-Socio-Acc, and JPP-AS scores were all positively correlated with a Instrument determined the extent to which subjects differ entially assumed similarity between their responses and the responses of Others. iij-5 composite criterion of job performance (JPC) above the .01 level of confidence. The multiple correlation between JPP-standard D, W-Socio-Acc, SS-ASo, and JPP-ASo scores and the JPC was .607# as compared to a multiple R of .287 between the JPC and six conventional aptitude and Interest predictors. The difference between these two multiple R fs was significant beyond the .001 level. The predictive validity of the W-Socio-Ace scores for the JPC, however, was shown to be wholly a function of the sociometric popularity of the subjects. When the variance accounted for by popularity was partialled out of the .I4 . 3 correlation between W-Socio*-Acc and the JPC, the correlation dropped to a -.06. Further, when the popu larity scores were included in a multiple prediction battery for the JPC, the beta weight of the W-Socio-Aco scores dropped from .3056 to -*0l60. Not only was all the valid variance in the W-Socio-Acc for the particular criterion used in this study accounted for by the popu larity scores, but those popularity scores accounted for an additional of the variance in the criterion.^3 The .33 correlation between JPP-standard D scores and the JPC, the .1+2 correlation between JPP-AS scores ^Curiously enough, Tagiuri and his co-workers have never (1952, 1953# 195^1-» 1958) reported any relation between their sociometric accuracy scores and sociometric popularity. l i j . 6 and the JPC, and the significantly positive correlations between the JPP-AS, W-Socio-AS, and SS-ASp scores and various job satisfaction indexes did point to the following conclusions, however, within the limitations of this study, regarding the "practical” or empirical meaning of the social perception measures; 1. The more closely subjects’ descriptions or perceptions of their job approximate the average perception by homogeneous groups of that job, the more highly rated will be the performance of those subjects on that job; i.e., "stereotype accuracy” in job perception bears some relation to job performance. 2. Subjects’ tendencies to assume similarity between their perception of their job and their supervisor’s perception of their job is related both to their ability to do the job and their satisfaction with the job. Since satisfaction with the job is also substantially related to job performance (the multiple R between Sense of Achievement and Attitude toward Supervisor for the JPC was .1^79)* it can be concluded that these particular AS scores reflect job satis faction and, through this relation, also reflect job proficiency. 3* The W-Soelo-AS and SS-ASp scores were not related to job performance, but they were significantly correlated with Attitude toward Co-workers; since both AS scores measured the tendency to assume similarity with co-workers, it can tentatively be concluded that the tendency to assume similarity with Others in general is at least partially dependent on satisfaction with or liking for those Others. The complete inclusion of variance in sociometric accuracy-of-prediction scores in the variance of socio metric popularity scores obtained as a by-product of the "relational analysis," and the partial inclusion of the variance of interpersonal aocuracy-of-prediction scores in the variance of intrapersonal assumed-similarity scores derived from the same predictions provided strong support for Cronbachts (1955) Cronbach and Gage*s (1955) the sis that the perceiver rather than the Other should be studied before the interaction of the perceiver and the Other is studied. Otherwise, complex interactional hypotheses may be formulated to explain interpersonal phenomena that might more easily be explained in terms of what each person brings to the interpersonal situation. REFERENCES REFERENCES Alfert, Elizabeth. Two components of assumed similarity. J. abnorm. soc. Psychol.. 1958, i>6» 135-139* Allport, G. W. Personality: a psychological Interpreta tion. New York: Holt, 1937* Bell, G. B., and Hall, H. E., Jr„ The relationship between leadership and empathy. J. abnorm. soc. Psychol.. 195l*-> It2» 1^6-157. Bell, G. B., and Stolper, R. An attempt at validation of The Empathy Test. J. appl. Psychol.. 1955* .19(6), Uk2-kb3. Bender, I. E., and Hastorf, A. H. On measuring general ized empathic ability (social sensitivity). J. abnorm. soc. Psychol.. 1953> IfcS, 503-507. Blake, R. R., and Helson, H. Adaptability screening of flying personnel. Situational and personal factors In conforming behavior. School of Aviation Medicine, TJSAF, September, 1956. Chowdry, Kamla, and Newcomb, T. M. The relative abilities of leaders and non-leaders to estimate opinions of their own groups. J. abnorm. soc. Psychol., 1952, i j . 7 . 51-57. Clark, K. E. Measurement of interest patterns. Annual Report, Project Nbor1-21203, NR 151 2l*.6, November 15, 1953. Cronbach, L. J. Processes affecting scores on "under standing of others" and "assumed similarity." Psychol. Bull.. 1955, 177-191*.. Cronbach, L. J. Proposals leading to analytic treatment of social perception scores. In R. Tagiuri and L. Petrullo (Eds.), Person perception and interpersonal behavior. Stanford, California: Stanford Univ. Press, 1956. Pp. 353-380. 1U9 150 Cronbach, I». J*, and Gleser, Goldine C. Assessing similarity between profiles. Psychol. Bull.. 1953* S o , I4.56-I4.73* Crow, W. J* The effect of training upon accuracy and variability in interpersonal perception. J. abnorm. soc. Psychol., 1957, 355“36o. Davitz, J. Social perception and sociometric choice of children. J. abnorm. soc. Psychol., 1955* jj>0» 173-176. Dymond, Rosalind F. A preliminary Investigation of the relation of insight and empathy. J. consult. Psychol.* 19i4 .8, 12, 228-233. Dymond, Rosalind F. A scale for the measurement of erapathic ability. J. consult. Psychol., 1959* 13* 127-133. Dymond, Rosalind F. Personality and empathy. J. consult. Psychol., 1950, 34, 3J+3-351. Ebel, R, L. Estimation of the reliability of ratings. Psyohometrlka, 1951* , 16, k - ® 7 • Fiedler, F. E. An investigation into the concept of the ideal therapeutic relationship. J. consult. Psychol., 191+9. li, 239-2)+% Fiedler, F. E. A method of objective quantification of certain countertransference attitudes. J. clin. Psychol.. 1951, 2, 101-107. Fiedler, F, E. Assumed similarity measures as predictors of team effectiveness. J. abnorm. soc. Psychol.. 1955* 381-388. Fiedler, F. E. Interpersonal perception and group effectiveness. In R* Tagluri and L. Petrullo (Eds.), Person perception and interpersonal behavior. Stan- ford, California: Stanford iJnlv, Press, 1958. Pp. 2) 43-258. Fiedler, F. E., Hutchins, E. B., and Dodge, Joan. Quasi- therapeutic relations in small college and military groups. Psychol♦ Monogr.. 1959* 23.* 3 (Whole No. 573). 1 * 1 Fiedler, F. E., Warrington, W. G., and Blaisdell, F. J* Unconscious attitudes as correlates of sociometric choice in a social group, J, abnorm, soc* Psychol,, 1952, k z , 790-796. Gage, N. L. Judging interests from expressive behavior. Psychol. Monogr., 1952# 66, No. 18 (Whole No. 3^0). Gage, N. L. Accuracy of social perception and effective ness in interpersonal relationships. J. Pera.. 1953* 22, 128-11* 2. Gage, N* L., and Cronbach, L. J. Conceptual and methodo logical problems in interpersonal perception* Psychol. Nev.. 1955# 62, 1*11-1*23. Gengerelli, J. A. A simplified method for approximate multiple regression coefficients. Psychometrlka, 19I1 . 8* 135-347. Guilford, J. P. Psychometric methods. New York: McGraw- Hill, 19^1*. Guilford, J. P. Fundamental statistics in psychology and education. New York: McGraw-Hill, 1956. Hausman, H. J,, and Strupp, H. H. Non-technlcal factors in the job performance of aircraft mechanics. HFORL Report No. 3b, March, 1953* Hastorf, A. H., and Bender, I. E. A caution respecting the measurement of empathic ability. J. abnorm. soc. Psychol., 1952, kI, 57l|-577. ’ Hastorf, A. H., Bender, I. E., and Weintraub, D. J. The influence of response patterns on the "refined empathy score." J. abnorm. soc. Psychol.. 1955# 34l“3U3* Horwitz, M. The veridical!ty of liking and disliking. In R. Taglurl and L. Petrullo (Eds.), Person perception and interpersonal behavior. Stanford, California: Stanford Univ.' Press, 1958. Jarrard, L. E. Empathy: the concept and industrial applications. Personnel Psychol.. 1956, Q, 157-169. 152 Jones, E. E., and deCharms, R. Changes In social percep tion as a function of the personal relevance of behavior. In Eleanor Maccoby, T. M. Newcomb, and E. L. Hartley (Eds.), Readings In social psychology. New York: Holt, 1958* Keedy, T. C. An experimental investigation of the social Intelligence test In relation to general intelligence and employment success prediction. Unpublished masterfs project, Univ. of Southern California, 1950. Kerr, W. A., and Speroff, B. J. The empathy test. Chicago: Psychometric Affiliates, 19^1• Luce, R. D., Macy, J., Jr., and Tagluri, R. A statistical model for relational analysis. Psychometrlka. 1955i 20, 319-329, Lundy, R. M. Assimilative projection and accuracy of prediction of Interpersonal perceptions. J. abnorm. soc. Psychol., 1956, £2, 33-39* Morsh, J. E. The development of Air Force Factor Refer ence Battery "tl. APPTRC-TN-57-10Zf/ July, 195?. Norman, R. D., and Ainsworth, Patricia. The relationships among projection, empathy, reality, and adjustment, operationally defined. J. consult. Psychol., 195^» 18* 53-59* Schmid, John, Jr., Morsh, J. E., and Detter, H. M. Analysis of job satisfaction. AFPTRC-TN-57-30. March. ^ ---------------- Speroff, B. J. Empathic ability and accident rate among steel workers. Personnel Psychol.. 1953* 297-301. Steinmetz, H. C. Directive psychotherapy: V. measuring psychological understanding. J. clln. Psychol., 19^5* 1, 331-336. Tagluri, R. Relational analysis: an extension of socio metric method with emphasis upon social perception. Soclometry. 1952, 15. 91-lOlj.. Tagiuri, R., Blake, R. R., and Bruner, J. S. Some determinants of the perception of positive and nega tive feelings In others. J. abnorm. soc. Psychol., 1953, M , 585-592. 153 Tagluri, R., Bruner, J. S., and Blake, R. R. On the relation between feelings and perception of feelings among members of small groups. In Eleanor Maccoby, T. M. Newcomb, and E. L. Hartley (Eds,), Readings in social psychology. New York: Holt, 1958* Tagluri, R. Social preference and its perception. In R, Tagluri and L. Petrullo (Eds.), Person perception and Interpersonal behavior. Stanford, California:.. Stanford ITniv. Press, 1958* Tagiuri, R., and Petrullo, L. Person perception and inter personal perception, Stanford, California: Stanford Univ. Press, 1958. Twery, Raymond, Schmid, John, Jr., and Wrigley, Charles. Some factors in job satisfaction: a comparison of three methods of analysis. Educ. Psychol. Measmt. 1958, 18, 189-202. U, S. Department of Labor. Estimates of worker trait requirements for h..000 .lobs. Washington?"£>. C.: U. S. Government Printing 6ffice. Van Zelst, R. H. Empathy test scores of union leaders. J. appl. Psvchol.. 1952, 3&» 293*“295* Van Zelst, R. H. Validation evidence on the Empathy Test. Eduo. Psychol. Measmt., 1953 * 2£> k7k~k-77» APPENDIX A BASIC INTERCORRELATION MATRIX 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 1. 2. 3. 4* 5. 6. 7. APPENDIX A BASIC INTERCORRELATION MATRIX .39 .31 .67 .37 •23 .15 • 55 .48 .22 .40 .45 .40 •48 •35 .57 •23 .22 .04 •29 .45 .37 •05 -.04 -.10 -.23 •00 -.13 -.08 .11 — .25 1 -.25 -.17 • -.18 —.18 •00 •49 -.03 -.16 -.19 .00 -.04 -.05 .25 -.22 -.10 -.18 -.05 r - 4 r - 4 « 1 -.03 -.09 -.01 .20 -.16 -.15 •49 .14 -.10 ■ -.12 .06 -.06 -.23 .07 -.11 -.04 .47 •55 .31 .01 -.12 .03 •26 .18 .14 -.06 -.05 .09 .08 •08 .13 -.10 ■ -.06 - -.05 .11 -.14 •05 -.08 .03 .22 •02 •23 ' -.08 .23 •09 ' -.07 .19 .23 .20 •28 -.09 -.09 •18 .18 •16 .67 —.03 •25 .13 .18 .02 •03 .14 •06 -.01 .06 -.07 ■ -.10 -.19 •14 -.16 •25 •32 .04 .12 -.03 .13 -.04 ■ -.07 •04 .00 -.09 • -.14 to i - i . i -.05 .00 . -.04 .55 .02 -.19 -.24 -.06 ■ -.04 -.20 -.01 •15 .47 .08 . -.07 .04 -.18 .14 -.14 -.12 .16 .14 .07 .21 .01 .23 •13 -.14 -.11 .21 .05 -.11 -.03 -.05 •13 .12 .09 .00 ■ -.02 .24 .00 .09 .16 -.33 -.19 •45 .37 •34 .43 -.04 •69 •19 1 2 3 4 6 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Pattern Comprehension Word Knowledge Meohanioal Information Visual Attention Mathematics Reading Comprehension Interest 8. JPP-standard D 9 • JPP-raw D 10* JSS-Sense of Aoh 11# JSS-Attitude to Supervisor 12* JSS-Attitude to Co-workers 13* W»3ooio-Aco 14. W-Sooio-AS 15. Sooio-Pop 16. JPP-ASo 17. SS-ASo 18. JPP-Aoo 19. Social Situations Test 20. JPC 155 APPENDIX B CRONBACH»S ANALYSIS OF VARIANCE MODEL FOR SOCIAL PERCEPTION SCORES APPENDIX B CRONBACH*3 ANALYSIS OP VARIANCE MODEL FOR SOCIAL PERCEPTION SCORES Cronbach’s (1955) model for accuracy-of-prediction (SP) and assumed-similarity (AS) scores that were derived from the predictions by subjects of the responses of more than one Other on more than one test item was an analysis of variance model. In the case of SP scores, for example, the difference between the actual response of an Other and the response a subject predicted the Other would make was broken down into four components: 1, The difference between the average prediction made by the subject and the average response made by all the Others (the "E" component). 2* The difference between two differences: a. The average prediction made by the subject across all Items minus the average predic tion across both Others and items, b. The average response made by Others across all items minus the average response by Others across both Items and Others (this was the 1 , DEn component, which contained the ADE component common to both SP and AS scores; in analysis of variance terms, the two differences Involved are the variance of the predicted means for the Others around the predicted grand mean, and the variance of the Others1 means around the Others’ grand mean). 157 1 £ 8 3. The difference between two more differences: a* The average prediction made by the subject across all Others minus the average prediction of the subject across both Others and items* b. The average response made by the Others across Others minus the average response made across both Others and items (this was the "SA” component, the variances of predicted and Other item means around their grand means). The difference between the "interaction” score derived from each Other's response (the individual response minus the Other's mean response, minus the mean item response, plus the grand mean for all Others and items) and the interaction score derived from each sub ject’s prediction of the Other's response (this was the ”DA" component, which contained the ADI component). Of the four components Cronbach identified In SP and AS scores, only the S component was present in the SP and AS scores used in this study. The effects of the E component were controlled In the JPP-standard D scores (mean differences between the subjects' job descriptions and their squadron mean job descriptions were removed by the standardization procedure), in the W-Socio-Acc and W-Socio-AS scores (mean differences between subjects' predictions of the sociometric ratings they received and the actual ratings they received were a function of the number of omissions the subject made in his predictions or the number of omissions the Others made in their ratings; both kinds of omissions were controlled by the weighting 159 procedure), and in the SS-ASpD and SS-ASoD scores (the mean differences between the subjects * predictions for the most-liked co-worker and either the subjects1 personal choices or his predictions for the least-liked co-worker on the Soc. Sit. test were not present when these scores were computed). Mean differences were also removed before the SS-AccD scores were computed. The effect of the E component was not controlled in the other SP and AS scores used in this study. To control for the effect of differences in predicted-Other or Other item means (the DE component) and in Other or predicted-Other means (SA component), or for the effect of the interaction between the two components (the DA component), those effects had to be preset;, and they were not present in the SP and AS data obtained for this study. There were no predicted-Other or Other item means in the data. There were no means for predicted- Other s or Others in the data, because subjects made predictions either for one Other only or for one item only. There could be, then, no "main effect" for either Others or items, and neither could there be an "inter action" between Others and items. APPENDIX C DATA COLLECTION INSTRUMENTS THE TASK CHECK LIST 162 INSTRUCTIONS FOR THE TASK CHECK LIST The purpose of this check list is to find out what tasks you perform, how often you perform them, and the conditions under which you perform them. For each ta3k listed down the left side of the page, place a check mark In the column which best describes HOW OFTEN and UNDER WHAT CONDITIONS you perform the task. You will make a check mark for every task under the "How often?" question. If you have never performed the task, you will not check any of the columns under the question "How do you perform this task?" NOTE! The term "replace" when used in a task statement Implies making the necessary alignments and adjustments that go along with replacement. When you have finished the Task Check List, please go on to the Two-Man Task Rating Soale, Instructions for its use appear at the top of the scale. Do you have any questions? TASK 10W OFTEN DO YOU PERFORM THIS TASK? HOW DO FORM T] YOU PER HIS TASK More than once a week Once a week Once a month Every 3 mos. Every 6 mos. I have never per formed it Under tech. super vision Without tech. super vision 1. Adjust phi balance and sensi tivity 2. Replace plug-in amplifier(s) on CAU 3* Troubleshoot in navigation control servo loop li. Replace radar synchronizer (APS 610 5. Conduct check of bomb tone release 6. Replace true heading trans mitter 7. troublesRoot in polar servo loop 8. Adjust theta sensitivity in SAU 9. Replace radar modulator 10. Replace periscope 11. Replace tracking computer 12. Troubleshoot in periscope azi muth data flow servo loop 13. Adjust T-l synchro in antenna lH. Replace safety wire 15. Adjust AR and TiVa-L balance and sensitivity in bombing computer 16. Troubleshoot in azimuth & sighting angle servo loop in LOS position 161) . TASK LEADERSHIP SCALE The tasks listed on the following pages usually require two or more men to perform them. For each of your men listed at the right, please Indicate how often each one takes the lead when assigned to each task in terms of the following scale: 0 - Never performs this task 1 - Never takes the lead in the performance of this task 2 - Seldom takes the lead in the performance of this task 3 - Sometimes takes the lead in the performance of this task i ) . - Usually takes the lead in the performance of this task 5 - Always takes the lead in the performance of this task TASK LEADERSHIP SCALE 1. Adjust phi balance and sensitivity_______ 2. Troubleshoot In navigation control servo loop 3. Troubleshoot In polar servo loop_____________ Replace tracking computer 5* Troubleshoot in periscope azimuth data flow servo loop 6. Adjust AR and TiV^-L balance and sensitivity in bombing computer_________ 7. Conduct complete operational check of the system________________________________________ 8. Adjust Xpj - Xjj balance on CAU 9. Replace bomb release computer 10. Conduct postflight check_____________________ 11- Conduct check for radar pressure leak 12. Troubleshoot in PPI Indicator servo loop 13. Troubleshoot in azimuth computer servo loop l i j . . Replace radar receiver-transmitter___________ 15. Replace stabilization unit l6. Replace polar converter 166 INSTRUCTIONS FOR THE PROFICIENCY RATING SCALE The purpose of this Rating Scale is tc find out how proficient each of the 321£OE mechanics in jour flight is in all aspects of maintaining the entire bomb-navigation system. The scale consists of a series of circles going from the bottom left to the top right of the page. The bottom left circle represents the poorest 5>-level bomb- navigation system mechanic that you have ever worked with. Going up the page, circles represent better and better mechanics with the circle at the center of the page representing the average or ordinary 5“level mechanic, and the circle at the top right representing the best £-level bomb-navigation system mechanic that you h^ve ever worked with. The £-level bomb-navigation system mechanics you work with are listed in the box on the page* Put the number corresponding to each of the mechanics in a circle on the scale corresponding to how well you feel he performs the over-all job. Avoid, if you can, assigning two men to the same circle. Be sure to rate yourself as well as the other mechanics in your flight* In making your ratings consider all of the 5~l®vel bomb-navigation system mechanics you have ever worked with, and rate each manTs present capability for doing the entire job rather than his future promise. Do you have any questions? OVER-ALL RATING SCALE FOR 5-LEVEL BOMB-NAVIGATION MECHANICS (APSC 32150E) The best 5“level bomb-navigation system mechanic I have ever worked with O O o o o o o o o The average or ordinary 5-level bomb-navigation system mechanic O • O o o o o o o o o The poorest 5-level bomb-navigation system mechanic I have ever worked with .••••• NAME 1. 2. ! ) > • 6. 8. 10. 11. 12 • THE APTITUDE TESTS 169 Directions for TEST 1 This is a test to see how quickly you can under stand how a pattern will fold up into an object. 1 PATTERN ______k J i > OBJECT In the sample above there are two drawings. The drawing on the left is a flat pattern* The drawing on the right is the hollow object that is made by folding the pattern. The small arrows appear in exactly the same position on the pattern and its object. The dotted lines show where the pattern is folded to make the object. Each number on the pattern indicates the same edge as one marked with a letter on the object. You are to decide which number on the pattern corresponds to each letter, (A, B,. C, or D), on the object. To answer question 1, find 1 on the pattern, then find where that edge falls on the object. Number 1 falls into the position marked C on the object. Mark C for item 1 on your answer sheet. Now for question 2, find the letter on the object that is in the same position as number 2 on the pattern. Mark the correct letter on your answer sheet. Go ahead with questions 3 and if* Mark the letters on your answer sheet that correspond to edges 3 and If. The right answer for question 2 is A. For question 3 you should have marked D, and for question If, B Is the right answer. 170 Directions for TEST 2 In this test, look at each underlined word, then from the four words beneath., select the word that means almost the same as the underlined words On the answer sheet, opposite the proper number, mark the letter that is the same as the answer you have chosen. You may answer items even when you are not sure your answer is correct. 1. TOXIN means SYNTHETIC means a. document a. manufactured b. poison b. friendly c. alarm-bell c. musical d. illness d. regular 2. INVERT means 5. PERJURY means a. spend a. sincerity b. discover b. probity c. overturn c. candor d. abuse d. lying 3. RETRIEVE means 6. LEGIBLE means a. recover a. colorful b. yield b. clear c. remedy c. outspoken d. revive d. humorous 171 Directions for TEST 3 This test consists of mechanical problems. Each question has i j . answers. Only one of these is right. Choose the correct answer and mark it on the separate answer sheet by thoroughly blackening the proper space. The score is the number right. If you do not know the right answer, guess as well as you can. 1. An acetylene torch is usually used in cutting a. brass. b. copper. c. aluminum. d. iron. 2. A magnet attracts a. iron. b. brass. c. copper. d. platinum. 3. A universal joint is found in the a. crankshaft. b„ transmission. c. drive shaft. d. connecting rod. if. The chief function of the distributer Is to a. distribute oil to moving parts. b. regulate gas and air mixture, c. ignite charge in the cylinder. d* furnish electric current to cylinders at the proper time. 5. The space around a mag net in which there are lines of force is the a. displacement area, b. field. c. polarity area. d. declination. 6. A chemical which aids In making solder stick to metal is called a. ceraenter. b. binder. c. joiner. d. flux. 172 Test Number l \ . - Visual Attention Instructions: This is a test of your ability to keep your eyes on one object as you sweep through a field of many similar objects* Your task is to follow each line from the left side of the page where it begins across to the right side of the page where it ends* You will then write the number of the line by the arrow head to show your answer* For example, in the sample below, lines No* 1 and 2 have already been done to show you how. Go ahead and do 3 and i j . for practice* Remember to write the number of the line where that line comes out on the right* Do not use your pencil to follow the line if you can help it. You will work faster if you depend on your eyes alone* ■ —i Do not turn the page until you are told to. When the examiner say3 "GO," turn the page and work as rapidly and as accurately as you can until time is called* You will be allowed five minutes. THE ACHIEVEMENT TESTS 17k READING COMPREHENSION TEST Directions: This is a test of your ability to read and understand technical paragraphs. A number of questions follow each of the paragraphs. On the basis of Information contained in the paragraph, you are to select the best answer of the five possible one3 and then blacken the space on the answer sheet that has the same number and letter. You may not be sure of the correct answer for all of the questions, but mark the one that seems be3t to you. Fuel from the two-stage gear-type fuel pump is delivered to the adjacently mounted fuel unit power control where it is passed through a screen and is then delivered to the pressure-regulating valve. The pressure- regulating valve permits the passage of fuel at 9Q Psi to the throttle valve. Pressures in excess of 90 psi are returned to the second stage inlet of the pump by means of a by-pass line. The throttle valve meters the fuel according to the position of its contoured valve as dictated by the solenoid-actuated pilot valve. The pro portioning solenoid shifts the pilot valve linearly in a direction and to an extent corresponding to the polarity and the strength of a voltage applied to it by the electronic control. 1. The amount of fuel metered by the throttle valve is a. controlled by the pressure-regulating valve. b. regulated by the fuel unit power control. c. manually controlled by the pilot. d. related to contour valve position. 2. The amount of pilot valve movement is related to a. voltage strength, b* voltage polarity. c. fuel pressure in the by-pa3s line. d. solenoid polarity. 3. The position of the throttle valve is controlled by the a. solenoid contour. b. fuel pressure. c. pilot valve. d. pressure-regulating valve. 17$ MATHEMATICS TEST Directions: This is a test of your knowledge of mathematics. Each question is followed by five possible answers. You are to select the one which be3t answers the question; then blacken the space on the answer sheet that has the same number and letter. You may not be sure of the correct answer for all of the questions, but mark the one that seems best to you. 1. The dots in the expression, 3-Ij.-5*6, mean that these numbers are to be a. multiplied. b. added. c. factored. d. subtracted. e. divided. 2. The expression, {2m)m, is equal to a. 2m2 b. 3m c. 2m <S. 3 e. 2 3. The expression l2, is equal to a. -1 b. -I c. I s d. 1 e. 2 i t - . Which one of the follow ing types of angles is greater than 90° and less than 180°? a. Acute. b. Right. c. Reflex. d. Obtuse. e. Straight. $• If the two sides of a right triangle are 12 and 16 inches, the hypo tenuse is a. 12 inches. b. lk inches. c. 16 inches. d. 18 inches. e. 20 inches. 6. A vector quantity has both a. amplitude and phase. b. amplitude and magni tude, c. magnitude and direc tion. d. direction and phase. e. phase and magnitude. THE VOCATIONAL INTEREST INVENTORY 177 DIRECTIONS FOR PART I You will find many activities listed below. They are arranged in blocks of threes. You must make a choice in each block of the one thing you like to do most, and of the one thing you like to do least. Put your choices on the answer sheet you have been given: Mark the thing you like to do most with a plus- sign (+) Mark the thing you like to do least with a minus- sign (-) That will leave one of three items in each block blank: Example: ( ) a. Write letters. (+) b. Fix a leaky faucet. (-) c. Interview someone for a newspaper story. 1. ( ) a. Catch up on your correspondence. ( )b. Try to fix a kitchen clock. ( )c. Discuss your philosophy of life with someone. 2. { )a. Watch an appendicitis operation. ( )b. Attend a lecture on television. ( )c. Go to an exhibit of recent Inventions. 3- { )a. Write a newspaper column of advice on personal problems• ( )b. Compete in an athletic tournament. ( )c. Take part in a public speaking contest. k- ( ) a. Study chemistry. { )b. Study stenography. ( )c. Study manual training. 178 DIRECTIONS FOR PART II The items "below describe various jobs. For each item you are to make one of five responses: (a) I would like this job very much; (b) I would like this job; (c) I am neutral or don’t know about this job; (d) I would not like this job; or (e) I would dislike this job very much. For example, take the item "Be a steam fitter": If you would like to be a steam fitter very much, you would mark choice (a) on your answer sheet; if you would like to be a steam fitter you would mark choice (b); if you don’t know whether you would like it or not or are neutral toward the job, you would mark choice (c); if you would not like the job, you would mark choice (d); if you would dislike being a steam fitter very much, you would mark choice (e). 1. Drive a heavy truck and trailer on cross-country runs. a. Would like very much b. Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. 2. Announce radio or tele vision commercials. a. Would like very much b. Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. 3* Be a court stenographer and make verbatim recordings. a. Would like very much b. Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. i|. Rally support for civic fund raising campaigns. a. Would like very much b. Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. 5>. Produce a daily radio or television show. a. Would like very much b. Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. 6. Be a night watchman at an airplane plant. a. Would like very much b„ Would like c. Neutral or don’t know d. Would not like e. Would dislike very much. THE JOB SATISFACTION SCALE 180 DIRECTIONS There are a number of statements below concerning you and your job. You are to indicate the extent to which you agree with each statement by selecting any one of the following five alternatives: a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree For each item, blacken the space on your answer sheet that corresponds to the letter of the alternative you select. 1. I am entirely satisfied with my job. a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree 2. My supervisor praises his men for a job well done. a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree 3. My supervisor confidently handles emergency situa tions. a. Strongly agree b. Agree c. Undecided. d. Disagree e. Strongly disagree if. I must get approval for decisions that I should be able to make alone. a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree f?* We have too many bosses. a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree 6. I find my job exciting. a. Strongly agree b. Agree c. Undecided d. Disagree e. Strongly disagree THE JOB PERCEPTION PROFILE RATE THE JOB OF THE 32150E ASSIGNED TO FLIGHT LINE MAINTENANCE The statements down the left side of the page below describe various aspects of jobs. To the right of each statement is a rating scale. You are to rate the extent to which each of the statements DESGRIBES the job of the 3215>OE assigned to flight line maintenance. A FLIGHT LINE 32l£0E»S JOB IS CHARACTERIZED BY: Variety and change. Specific instructions. Need for precision. Independence. Leadership. Routine operations. Danger. Fatiguing work. Pressure. Sociable contact. Recognition and praise. Personal satisfaction. Never Seldom Occasionally Frequently Usually Always 182 THE SOCIOMETRIC RATINGS AND PREDICTIONS 181 } . PERSONAL PREFERENCES In the space below, please write in the names of all the 32l50E's in jour flight that you REALLY LIKE to work with. Do not write in the names of any 150's you simply do not mind working with or do not object to working with; write in only the names of the men you really like to work with. In the space below, please write in the names of all the 32150E»s in your flight that you DON'T LIKE to work with. Write in only the names of the men you really object to working with, not the ones that you simply don't mind working with. 185 PREFERENCES OF OTHERS In the space below, please write in the names of all the men in your flight who you think have indicated in their "personal preferences" that they REALLY LIKE to work with you* Remember, you must consider very carefully everything you know about the men you work with before you can accurately predict what their real preferences are. In the space below, please write in the names of all the men in your flight who you think have indicated that they DON’T LIKE to work with you* THE SOCIAL SITUATIONS TEST 187 In this test some social situations will be out lined briefly* Imagining yourself in these situations, you are to indicate in every case which action among four given alternatives you would consider be3t* It will not be necessary to give a reason for your choice. Look at the following item: You are on a week-end trip with a group of friends. Most of them would prefer spending the day hunting, but you would prefer to go fishing. You should: a. Go hunting with them. b. Tell them to go hunting, while you go fishing. c. Try to convince them that they will have a better time fishing. d. Offer to toss a coin to decide whether the whole group goes hunting or fishing. Going on a week-end trip with a group of friends involves certain obligations. One of them is to abide by the majority opinion on the type of entertainment desired. If you want to be your own boss then you should go on a week-end trip alone. Therefore, a Is the best action to take and so you would choose It s's the best answer. To go your own way, as In alternative b would not sit well with the rest of the group and may even insult them. Similarly, in alternative £, to try to talk the rest of the group into your idea of an enjoyable week-end may well be resented by the rest of the group. D is probably the worst alternative: this action implies that your opinion has as much weight as the opinion of the rest of them put together. In this item, as all through the test, there are no "right" and "wrong" answers. However, in every case there is a best alternative In the sense of leading to the most desirable consequences in the social situation for the group. 188 1. You are late for a business appointment. It would be best tot a. Think up some plausible excuse, b* Give the reason for the delay. c. Say that you forgot the exact time of the appoint ment* d. Don’t mention it at all. 2. You have been appointed to a supervisory position and you find that some of the men under you are quite a bit older than yourself and have been with the company longer. You should: a. Be very sure of yourself to gain respect. b. Ignore the age and seniority factors. c. Take advantage of their experience if It helps your work. d. Ask them what to do rather than tell them. 3. It Is 2 A.M. and you are trying to go to sleep but in the apartment above you a noisy party is going on. You should: a* Show your annoyance by turning up your radio all the way. b. Find the fuse box to the apartment and cut off the electricity. c. Call the police and make a complaint. d. Endure the disturbance and complain to the people the next day. 4* You discover that another member of the office you work in is telling false stories concerning your character. You should: a. Ask him privately what he has against you. b. Tell a few "good” ones about him. c. Ignore the incident. d. Demand than he retract the lies.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Role Playing Ability And Social Adjustment In Children

PDF

On The Relationship Between Anxiety And Aggression In Nine-Year-Old Boys

PDF

A Study Of The Relationship Of Temperament Variables To The Ability To Make Certain Judgments Of Emotional Behavior

PDF

A Study Of The Effects Of Generalized Expectancies Upon Accuracy Of Interpersonal Perception

PDF

Intellect After Lobotomy In Schizophrenia: A Factor-Analytic Study

PDF

Behavioral (Social) Intelligence: A Factor Analysis

PDF

Anxiety Level And The Repression-Sensitization Dimension In Desensitization Therapies

PDF

An Experimental Study Of Aggression Habit Patterns And Sex Role

PDF

The Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat

PDF

Prediction Of Therapeutic And Intellectual Potential In Mentally Retardedchildren

PDF

A Study Of Criteria Of Social Perception And Some Related Variables

PDF

Measures Of Pilot Performance: Comparative Analysis Of Day And Night Carrier Recoveries

PDF

The Influence Of Order Information And Stimulus Display Time On Short-Term Retention

PDF

Delinquency As A Function Of Intrafamily Relationships

PDF

The Effect Of Subject Sophistication On Ratio And Discrimination Scales

PDF

A Factor-Analytic Study Of Problem-Solving Abilities

PDF

A Multidimensional Scaling Of Mood Expressions

PDF

A Preliminary Study Of The Dimensions Of Future Time Perspective

PDF

Relationship Of Achievement Motivation To Perception Of Degree Of Task Difficulty And Estimate Of Performance

PDF

Use Of College Students In A Social Therapy Program With Chronic Schizophrenics To Produce Changes In Mood And Social Responsiveness

Asset Metadata

Creator Strayer, Forrest Keith (author)

Core Title Empathy And Social Perception

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,Psychology, clinical

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Jacobs, Alfred (committee chair), Arnold, Aerol (committee member), Guilford, Joy P. (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-61538

Unique identifier UC11357900

Identifier 6002074.pdf (filename),usctheses-c18-61538 (legacy record id)

Legacy Identifier 6002074.pdf

Dmrecord 61538

Document Type Dissertation

Rights Strayer, Forrest Keith

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA