A STUDY OF THE CONSTRUCT OF FEAR OF SUCCESS“. THE VALIDATION OF ITS MEASURES AND AN EXTENSION OF ITS NOMOLOGICAL NETWORK Dissertation for the Degree of Ph‘ D. MICHIGAN STATE UNIVERSITY ROBERT JAMES GRIFFORE '1976 Ira-4 Q ‘ ' L 73771177 3L\1‘l\ \ Q9 I».' .. IIIIIIIIIIIIIIIIIIII IIIIIII L I '.. EJ'AAIVLEIE‘Aty a.‘ 1 This is to certify that the , ' ’ thesis entitled A STUDY OF THE CONSTRUCT OF FEAR OF SUCCESS: THE VALIDATION OF ITS MEASURES AND AN EXTENSION OF ITS NOMOLOGICAL NETWORK presented by Robert James Griffore has been accepted towards fulfillment of the requirements for Ph. D. degree in Educational PsychoTogy Major professor Date October 13, 1976 0-7 639 AHEC £33: I NINE; Q 4 29132 _.Jillx33 II 9 2%05 ABSTRACT A STUDY OF THE CONSTRUCT 0F FEAR OF SUCCESS: THE VALIDATION OF ITS MEASURES AND AN EXTENSION OF ITS NOMOLOGICAL NETWORK By Robert James Griffore The Problem This study incorporates the use of several methods to critically scrutinize the construct validity of three frequently used measures of fear of success. In addition, the construct of fear of success is, it- self, evaluated in terms of its relevance to performance on an actual classroom test. Achievement motivation is one of several personality attributes which are considered to be instrumental in determining academic behavior, but it is basically a construct which was not theoretically conceived as highly predictive of academic performance. Nor is it a tangible, directly observable property of the individual; it must be inferred from performance or measured as fantasy responses to ambiguous cues. The problem with such constructs is that they are bound into a nomological network of behaviors and other personal attributes and are related in often quite uncertain ways to behaviors outside the boundaries of the extant nomological network. The relationship of achievement motivation to academic achievement is one of these nebulous links to space on the fringe of the nomological network defining achievement motivation. Fear of success is a similar construct which has recently been Robert James Griffore derived from theories of achievement motivation, and some have claimed that it is a particularly powerful trait relative to its hindrance of women's success in competitive situations. What has been inadequately determined is the place of the construct of fear of success in edu- cational tasks. This study seeks to determine some educationally re- lated boundaries of the nomological network surrounding fear of success and to establish the construct validity of three relatively recent techniques of measuring fear of success, including one fantasy-based projective scoring system and two objective questionnaires. Literature Review The literature review is extensive, broad and concerned with the theories of achievement motivation from which fear of success theory was derived. The developmental antecedents of achievement motivation and related sex differences are described. The construct of fear of failure is also described in relation to achievement mo- tivation. With this background, fear of success theory and empirical research are extensively reviewed. Methodology Several methods of determining the validities of instruments which purport to measure a construct are utilized in this study. One of these is to determine whether an instrument measures the construct _in ways consistent with the theory behind the construct. Several pre- dictions which are theoretically consistent are made and tested. Anal- yses are conducted on the influence of fear of success on exam per- fbrmance, to determine whether sex differences exist on the trait, and to determine the degree to which all selected fear of success instruments Robert James Griffore are intercorrelated and negatively correlated with fear of failure. The upper and lower bounds of construct saturation are also estab- lished for each of the fear of success instruments, as an important indicator of construct validity. Combining all analyses, judgments are made about the quality of each instrument and the relevance of the construct in education. Results The major findings can be summarized as follows. 1. Some interactions between fear of success and exam perfor- mance are consistent for the F05 and FOSS fear of success instruments, but the Horner instrument fails to interact with item difficulty, par- ticularly on low difficulty items. 2. No sex differences were found on fear of success. 3. While two fear of success instruments were positively cor- related, a third instrument correlated with neither of these. The existence of a single construct may be questionable. 4. A greater proportion of variance on fear of success scores was accounted for by fear of failure than by behavior theoretically related to fear of success. A STUDY OF THE CONSTRUCT 0F FEAR OF SUCCESS: THE VALIDATION OF ITS MEASURES AND AN EXTENSION OF ITS NOMOLOGICAL NETWORK By Robert James Griffore A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Counseling, Personnel Services and Educational Psychology I976 ACKNOWLEDGEMENTS The investigator wishes to express appreciation to the many individuals who gave assistance during the endeavor of preparing for and conducting this research. Special thanks are due to Dr. Robert L. Green for his assistance in virtually every step of completing the study, and to Dr. Harvey F. Clarizio for his thoughtful and timely suggestions in formulating the research problem. Other members of the guidance committee, Dr. Louise Sause, Dr. Ellen Strommen and Dr. Raymond Hatch, each contributed helpful suggestions in the formulation of the study, which strengthened it in several critical aspects. Without the assistance of Dr. William Mehrens, Dr. Robert Ebel, Eric Gordon and Julie Nyquist, the collection of data would have been either difficult or perhaps impossible. Each made it possible to obtain subjects in both an apprOpriate and an efficient way. Finally, I wish to express special appreciation to my wife, Gaile,for her patience, support and encouragement for the entire dura- tion of the research. ii TABLE OF CONTENTS Page LIST OF TABLES ......................... v LIST OF FIGURES ......................... vi Chapter I STATEMENT OF THE PROBLEM AND REVIEW OF LITERATURE. . . . l Achievement Motivation and Academic Achievement . . . 1 Achievement Motivation-Related Constructs and Instruction .................... 3 McClelland's Theory of Achievement Motivation . . . . 5 Research on McClelland's Theory ........... 7 Atkinson's Expectancy-Value Theory of Achievement-Oriented Activity ........... 8 Research on Atkinson's Theory Relative to Academic Achievement ................ ll Antecedents and Determinants of Achievement Motivation ..................... l5 Sex Differences in Achievement Motivation ...... 22 Devel0pmental Trends in Females' Achievement Motivation ..................... 25 Socialization of Females' Achievement Orientation . . 27 Fear of Failure in Females .............. 29 The Theory of Motive to Avoid Success ........ 30 Research on Motive to Avoid Success ......... 33 Predictive Validity of Measures of the Motive to Avoid Success .................. 4l Reliability of Measures of the Motive to Avoid Success ................... 43 Shortcomings in the Literature and a Statement of the Problem ................... 44 II METHODOLOGY ....................... 48 Design and Background of the Method ......... 48 Sample ........................ 59 Measures ....................... 59 Procedures ...................... 65 Hypotheses ...................... 67 Analysis ....................... 70 Summary ....................... 72 iii Chapter Page III ANALYSES AND RESULTS .................. 74 Introduction ..................... 74 The Effect of Fear of Success on Examination Performance .................... 75 Sex Differences on Fear of Success .......... 87 Concurrent Validity of the Fear of Success Instruments .................... 89 Relationships Between Fear of Success and Fear of Failure .................. 89 Relationships Between Fear of Success and Scores on Low Difficulty Items ........... 92 Summary of the Results ................ 94 IV SUMMARY AND CONCLUSIONS ................. 98 Purpose ....................... 98 Literature Review .................. 98 Design and Analysis ................. 99 Results ....................... IOI Discussion ...................... 103 Limitations of the Study ............... II4 Implications for Education .............. II5 Recommendations for Future Research ......... I18 APPENDICES A STORY CONSTRUCTION TASK ................. I2I B PAPPO'S FEAR OF SUCCESS QUESTIONNAIRE (FOS) ....... I22 C THE ZUCKERMAN AND ALLISON FEAR OF SUCCESS SCALE (FOSS) ..................... I27 D THE ALPERT-HABER DEBILITATING ANXIETY SCALE (DAS) ........................ I30 E BACKGROUND INFORMATION ................. I33 F CODING SYSTEM FOR SOCIAL CLASS ............. I34 G FINAL EXAMINATION INSTRUCTIONS ............. I35 BIBLIOGRAPHY .......................... I36 iv l.("‘l‘lll|ll III..I Table I. LIST OF TABLES Results of Analysis of Variance on Examination Items Using Horner's Empirically Derived System as a Measure of Fear of Success ........... Results of Analysis of Variance on Examination Items Using Pappo's FOS Questionnaire as the Fear of Success Measure .................. Results of Analysis of Variance on Examination Items Using Zuckerman and Allison's FOSS as the Fear of Success Measure ............... Chi-squares for Sex Differences in Fear of Success on Each of the Fear of Success Measures ....... Matrix of Correlations Between All Fear of Success Instruments and the Fear of Failure Measure (DAS) with Levels of Significance ............. Matrix of Correlations Between Scores on All Fear of Success Measures and Number Correct on Low, Medium, and High Difficulty Exam Items and Total Exam Score ................... Aggregation of Information From All Analyses Pertinent to Construct Validity ........... Page 80 90 93 96 LIST OF FIGURES fiHEEQ. Page 1. Age Composition of the Sample by Class Intervals ..... 60 2. Number Correct on Low, Medium and High Difficulty Final Exam Items for Subjects Scoring High and Low on the Horner Instrument .............. 78 3. Number Correct on Low, Medium and High Difficulty Final Exam Items for Subjects Scoring High and Low on the Pappo FOS Questionnaire ........... 82 4. Number Correct on Low, Medium and High Difficulty Final Exam Items for Subjects Scoring High and Low on the Zuckerman and Allison FOSS ......... 86 vi CHAPTER I PROBLEM STATEMENT AND REVIEW OF LITERATURE Achievement Motivation and Academic Achievement One of the most persistent and interesting problems in education is that of identifying the factors which hamper or prevent some students from succeeding in the competitive arena of academic achievement, even though some degree of success would be predicted on the basis of their ability levels. It has frequently been suggested that students' low achievement motivation could account for this phenomenon. Those who have heralded the explanatory power of achievement motivation have wisely chosen a cautiously Optimistic position, however. Their caution stems from the realization that school grades are affected by a very large number of unidentified factors (McClelland, et al., I953). Birney (1968) observes that the relationship between students' achievement motivation and academic performance is not a consistent one. Wide var- iations in the strength of this relationship have been reported. Yet it is also true that aptitude and achievement measures account for as little as 30 percent of the variance in grades. It is quite tempting to believe that achievement motivation contributes a large proportion of the unaccounted-for variance. Indeed, some studies suggest that a significant positive cor- relation does exist between Thematic Apperception Test nAch and average grades. Ricciuti and Sadacca (l955) selected two groups of male high 2 school students, administered the TAT measure of nAch and found corre- lations of +.18 and +.23 in the two groups between nAch and grade averages, based on averages calculated over three terms. Rosen (1955) selected a sample of 120 high school saphomores, stratified on social class according to Hollingshead's Index of Social Position. In a series of within-class analyses, he found that achievement motivation was highly related to grade point average. I College students' academic standing is also reliably related to nAch. High academic standing is associated with a high need for achieve- ment (Morgan, 1951). Parish and Rethlingershafer (1954) conducted re- search similar to Morgan's and found no differences between low and high achievers in nAch. In terms of the usefulness of achievement motivation in accounting for low achievement, this presents a serious challenge, for low-achieving students would not be expected to have a high nAch score. This bit of disconfirming evidence is not uncharacteristic of other studies either. Birney (1968) observes, in fact, that while nAch and academic achievement are highly related in high school samples, selected college samples do not reveal this consistent relationship. In a study of Naval officer candidates and college freshmen, Ricciuti (1954) found zero-order correlations between nAch and academic achieve- ment. Bendig (1959) and Cole, Jacobs and Zubok (1962) found zero order and negative correlations, respectively. One is prompted to ask about the factors which provide such in- consistency between achievement motivation and academic achievement. First of all, as McClelland (1953) pointed out, achievement motivation is but one of the variables which determines achievement. Second, as 3 Broverman, Jordan and Phillips (1960) argue, levels of TAT nAch scores might be highest in individuals who are "unable to express their achievement needs" in real life. A third reason lies in the basic difference between the expression of a transient need state on fantasy measures and the expression of sustained academic achievement in school. And a fourth reason is that the motive to achieve has been erroneously understood as a unitary variable. This ignores the fact that its ex- pression depends upon affiliative tendencies, self-esteem needs and the self-concept. Related to this, McClelland (1961, 1962) has ob- served that scientific productivity is determined by several person- ality characteristics, only one of which is indicated by measured nAch. Achievement Motivation-Related Constructs and Instruction It would seem that, on balance, achievement motivation should not be strongly suggested as a general predictor of academic achieve- ment. It might be more defensible to study the role of this construct as it interacts with the characteristics of instruction. McKeachie (1961) conducted one of the early studies in this tradition and found that nAch was most strongly related to academic achievement in "low achievement cue" classrooms. The role and characteristics of the teacher had much to do with the impact of achievement motivation. This study should be considered as suggestive for other research. It is based on the notion that academic achievement is different from nAch. But to identify the ways in which the situational events in the class- room actually interact with the workings of the construct of achieve- ment motivation remains, at this time, as clearly a task of critical importance. How the construct fits into educational practice and school learning and helps to explain and predict student behavior is a 4 question which remains to be answered. In a more general sense, the time has come to reassess the role of the personality constructs which have been so widely accepted as they relate to education. Ebel (l972) observes that it is not entirely -clear whether these personality traits are causes of behavior or merely names for the behavior. Attempts at measuring personality traits are probably only justified if they are known to determine behavior related to academic achievement. Otherwise, the effort devoted to such measure- ment might be more wisely spent. On the matter of achievement motiva- tion, it might not be appropriate to attribute causative power to this personality construct, but to recognize it as simply a description of an observed result. Even in terms of scores on TAT nAch, this position would suggest that the behavior cued by the projective stimulus is the basis on which the personality construct is merely attributed to the individual. Indeed, the basically inconsistent relationship between achievement motivation and academic achievement might be caused, at least in part, by the failure of measurement techniques to attribute achievement motivation reliably; the problem might be attributional. In this context, the present study may be correctly interpreted as an investigation into the practical educational efficacy of a class of constructs, falling generally under the heading of achievement motivation-related. This study is an immediate challenge to the con- struct of fear of success or motive to avoid success, which was derived from a theory of achievement motivation, and its basic purpose as it pertains to education has recently been to account for different levels of educational achievement in a similar way as has motivation to achieve. That is, when students do not achieve in a way which would be predicted 5 on the basis of their scores on achievement or aptitude measures, it has been suggested that they might have a high motive to avoid success. The psychodynamics behind this hypothesis are somewhat complicated, and their structure and development are best understood given a founda- tion of the theories of achievement motivation which preceeded the theory of motive to avoid success. After a review of achievement moti- vation theories, the rationale and character of fear of success will be reviewed. McClelland's Theory of Achievement Motivation As Brown (1965) observes, the construct of achievement motiva- tion was in danger of extinction due to lack of interest in the early 19505. It was probably sustained by the work of McClelland and his colleagues in such works as The Achievement Motive (1953). McClelland's basic conceptual structure is relatively simple. Achievement motiva- tion is considered to be a function of two factors: 1) the immediate situation, and 2) the relatively constant need for achievement in the individual. It is the latter of these which is McClelland's focal point. Alschuler (1973) points out that McClelland is chiefly inter- ested in the ogerant need for achievement, which tends to occur even when situational cues for it are weak. He is not concerned with what might be referred to as the respondent achievement motive, which would be sensitive to current contextual cues. Basically his theory of the achievement motive is consistent with his general theory of motivation. He asserts that motives determine why pe0ple behave as they do. They are the origins of action and the central constructs of behavior. McClelland, Atkinson, Clark and Lowell (1953) formally define a motive as "the redintegration by a cue of a change in an affective situation." The appearance of meaningful cues which have been associated with strong affective states in the past again arouses those affective states which, in turn, elicit instrumental approach and avoidance behavior. Emotions are not motives, but they are the states from which motives deve10p. Motives themselves appear when there are discrepancies between the pre- sent éffective state and an expected one. The change of the affective state in the direction which is expected is the function of a motive. In other words, McClelland subscribes to an equilibrium or balance model, in which identity between present and expected affective states is the goal. With respect to achievement motivation, an individual as- sociates pleasant feelings of satisfaction with high levels of achieve- ment. When he fails to achieve, such feelings do not occur, thus re- sulting in an affective discrepancy and, in turn, a motive to produce a pleasant affective state through achievement. Fundamentally, then, achievement motivation is expressed as competition with a standard of excellence (McClelland, et al., 1953). McClelland and his colleagues were also interested in placing the achievement motive in an ecological framework incorporating reli- gious, familial, psychological and economic variables. In an early formulation, coinciding with The Achieving_Society (1961), high achieve- ment motivation was associated with a cluster of entrepreneurial char- acter traits in the individual who is drawn to entrepreneurial occupa- tions. But this was only one link in McClelland's hypothesized sequence of events. He suggested that a society which was high on the dimensions of religious Protestantism characteristically subscribes to patterns of early independence training for its children. Consequently, the entre- preneurial character quite commonly deve10ps in these societies, and individuals with this character type are recruited into positions where the role of the entrepreneur is a necessary attribute. If the selection of entrepreneurial characters into these roles is sufficient, the society will enjoy a period of economic growth. This, then, des- cribes the place of the need for achievement in society. Research on McClelland's Theory Within this network of one-directional influences, there are several points at which the causal sequence might be interrupted. And questions have been formally posed regarding the strength of the asso- ciations in this causal sequence. But it would seem that McClelland's notions are not without some supporting evidence. McClelland's careful study of the relationship of achievement-oriented folktales and inde- pendence training in societies to national rates of economic growth has revealed that the association between these factors is significant. Achievement motivation appears to have a causal effect on rate of national economic growth. The degree of achievement motivation present in a society in one generation seems to have a bearing on the actual rate of economic growth enjoyed by the next generation (Clark, 1957). McClelland also computed the economic productivity of all Catholic and Protestant countries in the temperate zone in 1950 and found that there was a decided difference in favor of the Protestant countries; On the other hand, Protestantism was certainly not a part of pre-Incan Peru or ancient Greece, which also had high levels of achievement motivation and economic growth (Alschuler, 1973). And Veroff, Atkinson, Feld and Gurin (1960) have offered Thematic Apperception data showing that when Protestants are compared with Catholics across the United States, Cath- olics are not different from Protestants on nAch. Also, Jews had significantly higher achievement motivation scores than either Catho- lics or Protestants. Brown (1965) suggests that Protestantism per se is not necessarily related to achievement motivation, nor has early independence training been clearly considered in many extant studies. The strongest link in the causal chain seems to be between high achieve- ment motivation and economic growth. McClelland's theory is also concerned with the role of achieve- ment motivation in accounting for the behavior and.thought of the indi- vidual. Taken at this level, achievement motivation is associated with planning or striving for excellence, progress, uniqueness and competi- tion (Alschuler, 1973). Pe0ple with high need for achievement will not work harder when monetary rewards are offered (Atkinson and Reitman, 1956). They prefer to strive for excellence itself. They also prefer to take personal control over events in their lives (Heckhausen, 1967). They like to set their goals carefully, after weighing the probabilities of various alternatives (Alschuler, 1973). Also, they are more con- cerned with the long—range future. Green and Knapp (1959) point out that high achievers perceive time to be passing by rapidly. Atkinson's Expectancy—Value Theory of Achievement-Oriented Activity On both the individual level and the macro-ecological level, McClelland's model emphasizes the stable individual trait of achieve- ment motivation. But Atkinson has constructed a theory which does not assign a central position to this constant trait. Indeed, his approach attempts to account for achievement-oriented activity in the individual rather than the constant motivational trait. Atkinson considers his theory to represent the interaction between personality and environment. It is assumed that the tendency to achieve success (Ts) which is 9 expressed as interest and performance, is equal to Ms (the motive to achieve success) X Ps (the subjective probability that performance will be followed by success) X Is (the incentive value of success). The Is is considered to be 1-Ps (Atkinson, 1974). The tendency to avoid failure is considered to result in a similar way from motive, probability and incentive value. Specifically, T-f = Maf X Pf X If. That is, the tendency to avoid failure equals motive to avoid failure X the probability that the act will lead to failure X the incentive value of failure for the activity. A third tendency in Atkinson's theory is T ext. This is a posi- tive extrinsic tendency to perform an activity. This factor usually incorporates a number of tendencies not directly associated with the motive to succeed. Among these tendencies might be the individual's wish to please others and gain their approval, to conform, or to com- ply with implied or expressed expectations. These factors are consi- dered to be context-specific. Actually, Atkinson only recently has strongly recognized the influence of these context-specific tendencies (Maehr and Sjogren, 1971). This theory is considered to explain the achievement behavior of individuals in "achievement situations.“ In these situations, one is aware that an outcome is at least somewhat uncertain, and that one's achievement is to be compared with a standard of performance. As repre- sented above, the theory clearly suggests that in these situations, the motive to approach success and the motive to avoid failure will be in conflict. The relative strength of these two motives in the individual will determine whether he tends to approach success or whether he tends to avoid failure. The balance of the two basic motives, Ms and Maf is IO thought to be a relatively constant personal trait within the indivi- dual. One of the major predictions Atkinson makes is that in achieve- ment situations, individuals whose M5 is stronger than their Maf will - demonstrate low achievement-oriented activity when P5 = 0.00 or 1.00. They are predicted to exhibit high levels of achievement-oriented acti- vity when P5 = .50. For individuals who have a predominance of Maf over Ms, however, the level of achievement-oriented activity will be lowest when P5 = .50 (Maehr and Sjogren, 1971). It should be noted that since Ps + Pf = 1.00, simultaneously intermediate values of each probability cause high T-f and Ts. Whether Ts and T-f are equal is another matter, and this is determined by Ms and Maf at a given time. Generally Speaking, it can be assumed that one of these is higher than the other, thus resulting in either excitation with the former or inhi- bition with the latter. Because either M5 or Maf will be stronger, the two tendencies to which they contribute will be unequal, given equal probabilities. The difference between Ts and Taf is known as the resultant achievement-oriented tendency. Whether the resultant tendency is expressed as an approach or as an avoidance response de- pends on the relative dominance of Ms and Maf and the subjective pro- babilities Ps and Pf. Either T5 or T-f may dominate (Atkinson, 1974). Atkinson's theory suggests that subsequent to success or failure there is a change in the strength of the tendency to engage in the same activity. Success produces a change in the subjective probability of success on the task, so that level of aspiration is raised. A task which originally had a P5 of only .30 might, after success, have a P5 of .50. After failure, the same task might decline in Ps to .20. II Shifts in expectancies such as those above are typical ones, originally observed to occur in college students. But certain atypical shifts can also be accounted for by the theory. Atkinson (1974) cites the example of the anxiety-prone individual who fails at an easy task. The effect is to lower Ps so that it approaches .50. After a repeat of this failure, the task becomes one of intermediate risk, and this is to be avoided. So the individual shifts from a low to a higher level of aspiration, choosing a task with a very low Ps at the other end of the continuum, where the probability is low. Thus the theory can account for both typical and atypical shifts in expectancy. Atkinson's theory would appear to contradict the Law of Effect, which has generally held a central position in psychology since Thorn- dike's statement of it. There have been several attempts to discredit it, such as McKeachie's (1974) critique. The effect of success is not always to strengthen the probability of a response. Success produces a change in incentive value of success. Sometimes this change results in a stronger tendency to act, and sometimes in a weaker tendency. This depends upon whether the motive to achieve or the motive to avoid failure is dominant in the person and upon the initial strength of P5 of the task. Clearly, the situation is more complicated than the Law of Effect would suggest. Research on Atkinson's Theory Relative to Academic Achievement The implications of Atkinson's theory for education depend on whether its general predictions are supported in research, as well as whether the theory can be characterized as pertinent to academic moti- vation. Many studies which have attempted to test the main predictions of the theory have indicated that it does indeed have explanatory and 12 predictive power. Achievement-oriented individuals, who have a pre- dominance of Ts over T-f, demonstrate a greater tendency to select tasks of intermediate difficulty rather than very easy or very diffi- cult tasks (McClelland, 1958; Atkinson, Bastian, Earl and Litwin, 1960; Litwin, 1958, 1966; Atkinson and Litwin, 1960). This tendency is ob- served in competitive game-like tasks and in curricular (Isaacson, 1964) and job (Mahone, 1960) choices, as well. Some research has called into question the supposition that achievement-oriented individuals prefer tasks which have a difficulty level of exactly .50, however (Heckhausen, 1968). Preferences of slightly less than .50 are quite common. Further- more, failure-threatened individuals, who have a predominance of T-f over Ts, have not always demonstrated the predicted U-shaped function varying around the P5 = .50. It has yet to be demonstrated that failure-threatened individuals actually avoid tasks of moderate diffi- culty. They seem only to have a preference for tasks with either higher or lower levels of difficulty (Maehr and Sjogren, 1971). The relationship between individuals' preferred level of task difficulty and their performance on these preferred tasks is not es- pecially consistent. Achievement-oriented individuals do indeed prefer moderate difficulty tasks, but they do not necessarily show their best performance on these tasks (Smith, 1964; O'Connor, Atkinson and Horner, 1966). Even more challenging to the theory is the finding that failure- threatened individuals rarely show decrements in performance when task difficulty becomes .50 (Karabenick and Youssef, 1968). In short, Atkin- son's theory does not have an impressive ability to predict performance on the basis of levels of task difficulty. Of course, this is one reason why it is appropriate to question the place of Atkinson's theory I3 as a framework for understanding and predicting academic motivation. It is even quite appropriate to question whether it is possible to construct a single theory of academic achievement motivation. At- kinson does not clearly indicate whether he is characterizing indivi- duals who achieve through competition with others, those who like to strive according to internal standards or perhaps both. Of course, a general theory of academic achievement motivation would need to incor- porate both of these types. McClelland et a1. (1953) anticipated the possible existence of different types of achievement motivation along the lines of these types, and Veroff (1969) considered the possibility that the nature of the child's achievement motivation might develop in a sequence of qualitatively different types. Whereas social competi- tion might be characteristic of the child in the middle years, it is quite conceivable that the very young child's achievement motivation is based on competence motivation (White, 1959). Since most of the research has been concerned with competitive performance on normatively defined standards, it would seem, therefore, that as a theory of aca- demic achievement motivation, it would be restricted in power to in- structional-achievement situations in which competition is encouraged, perhaps especially among older children. Another reason to suspect that Atkinson's theory is a rather inadequate attempt at accounting for academic achievement motivation is that it is probably not able to predict the degree of sustained effort students will demonstrate under natural conditions and with actual school learning tasks. In such situations, the student must be contin- uously concerned with revising his estimates of Ps. While Atkinson's theory suggests predictability of change as a function of shifting Ps, 14 the empirical data supporting this theoretical assumption is limited. Maehr and Sjogren (1971) question whether achievement-oriented students would really sustain a continued effort on tasks whose Ps remains at .50. In addition, one would question whether students who are threat- ened by failure will diligently persist at tasks whose Ps value is either 0.00 or 1.00. Because most of the studies concerning the validity of Atkin- son's theory utilized a sample of white middle-class children, and because the average classroom is not entirely populated with similar children, one is prompted to further question whether the theory could represent an account of academic achievement motivation. Katz (1967) observes that it is uncertain whether nAch measures are sensitive to the areas in which disadvantaged children have an interest in achieving. Rosen (1959) and Littig (1968) have found that lower-class black sub- jects and upper— and upper-middle-class black subjects do not report achievement imagery similar to whites' on projective nAch tests. Min- gione (1968), on the other hand, has found no basic difference between the responses of black and white subjects on these measures. Even if there were no basic difference across classes with respect to the ap- plication of Atkinson's theory on distinct achievement situations, one might even more vigorously question the assumption that lower-class, failure-threatened children would in fact persist on tasks with P = 0.00. If all these problems are not sufficient to call into question Atkinson's theory as an account of general academic motivation, there are two final points that add further damaging evidence. In the first place, with the notion of T ext, Atkinson recognizes that there will 15 always be an unidentified number of unknown factors standing ready to qualify the efficacy of the Ts and T—f in explaining academic achieve- ment motivation. Second, when the implications of the theory are con- sidered, one is faced with the conclusion that to operationalize the theory, one would need to measure all students' relative strengths of Ms, M-f, subjective Ps and Pf values, Is and If values and their sub- jective ingredients in the variable T ext. Even if this could be realized, the task of designing instruction on the principle of Achieve- ment Motivation X Treatment Interaction stands as the ultimate challenge. If an instructor is to attend to childrens' levels of achieve- ment motivation in planning instruction, there is a sense in which he might benefit from being aware of all the salient variables which have been strongly associated with the general development of the achieve- ment motive. Many of the variables which have been attended to in the developmental literature are not necessarily related to Atkinson's theory; many of the early studies, in particular, used TAT cues to measure nAch in the context of McClelland's theory. But not only were these studies not directly concerned with Atkinson's theory; they were not always clearly related to academic achievement motivation either. The independent variables which have been strongly related to achieve- ment motivation are typically mediated by family members and other agents outside the school. It would be difficult to apply the findings of most of these studies to education. Antecedents and Determinants of Achievement Motivation One of the classic studies was conducted by Winterbottom (1958). She showed that 8 to 10 year old boys with high achievement motivation I6 had mothers who expected early independent and exploratory behavior. Siss and Wittenborn (1962) have reported that third grade boys with mothers who had expected early independent behavior tend to have higher intelligence and achievement test scores. Rosen and D'Andrade (1959) made observations of parents whose 9 to 11 year old sons were given tower-building tasks. They found that the fathers of sons with high achievement motivation allowed their sons more independence in those tasks, while mothers tended to permit less independence. Adding to this inconsistent picture, Crandall, Preston and Rabson (1960) fOund that early independence training was quite unrelated to children's achievement. Crandall (1967) observes, however, that part of this discrepancy may be inherent in the differences between retrospective responses in some studies and current observations of independence training techniques in other studies. There appears to be a sex difference in the influence of par- ents on children's achievement motivation. Crandall, Katkovsky and Preston (1960) found that children's mechanical skill achievements are related to influences of parents on children of the same sex, but that achievement of physical skills seems to be related to whether parents of both sexes had participated in sports and had given encouragement. Rosen and D'Andrade (1959) found another difference in treatment based on the sex of the parent. In the tower-building task, mothers, but not fathers, tended to reward sons with approval and to punish with dis- approval and rejection for failure. It seems, based on data from Feld (1960) that there is interaction effect between independence training and age of the child. Contrary to Winterbottom's (1958) findings, achievement motivation is increased by a lack of independence training 17 in 14 to 16 year old boys. In addition, Rosen (1959) found that Winterbottom's findings were generalizable only to a middle-class p0pulation. Moss and Kagan (1961) determined that "maternal acceleration," brought to bear on children's achievement motivation during the first 3 years, relates to increased achievement behavior in children aged 6-10 and to achievement in adulthood. Heckhausen (1967) proposes that the conclusion one might draw from all such research is that achieve- ment is not necessarily related to independence training, but that it is related to children's lower concern for conformity. Indeed, too harsh independence training and pressure to achieve may be effectively equivalent to cold rejection, which Rosen (1959) found negatively re- lated to achievement motivation. Morrow and Wilson (1961) who were interested in over- and under-achievement in school, found that the parents of high achievers were more approving, trusting and affectionate with their children. McClelland (1961) found that sons with highest nAch are likely to have mothers with moderate levels of nAch. One of the most interesting findings is that the parental be- haviors related to children's high achievement motivation have often been described as deleterious in a more general sense. For example, Crandall, Dewey, Katkovsky and Preston (1964) found that mothers of high-achieving girls were less affectionate and less nurturant. Hoff- man, Rosen and Lippitt (1958) found that mothers of high-achieving boys were more coercive, and Barwick and Arbuckle (1962) found them to be less accepting. A general modeling effect appears to Operate for both boys and I8 girls. Norman (1966) found that fathers of high-achieving boys and mothers of high-achieving girls were higher in independence and lower in conformity than the parents of low-achieving children. But there are other determinants of achievement motivation over which those who govern in the educational setting would seem to have unknown degrees of control. Among these variables are the child's in- ternalization of standards of excellence, prediSposing levels of cog- nitive deve10pment, the influences of various subcultures and the in- fluence of the child's sex on achievement motivation. Crandall, Kat- kovsky and Preston (1960) have asserted that the function of achieve- ment behavior is to attain approval or to reduce disapproval origin- ating in oneself or in others in situations where there are clear standards of excellence. This is consistent with McClelland's position, of course, and it recognizes that the individual must associate plea- sant feelings with achievement, and he must anticipate that higher levels of skill will lead to feelings of pleasure. Further, perfor- mance standards are in continuous need of revision, according to the demands of social norns. A degree of persistence is required to attain progressively more acceptable levels of achievement, and Crandall, Preston and Rabson (1960) have shown that individual differences exist among 3, 4, and 5 year olds in persistence of exercise of skill and effort. Differences in achievement depend upon these differences in persistence. Early individual differences in ad0ption of standards have also been demonstrated by Crandall (1961). But apparently no study exists relating emergence of achievement behavior to a specific age or to Specific conditions. It would seem, however, that a certain level of cognitive I9 development might be required before adoption of standards of perfor- mance and their pursuance are possible. It is also conceivable that attainment of this cognitive state leads, as a matter of course, to deve10pment of such standards. Heckhausen (1967) maintains that achievement motivation first becomes possible when the child "struc- tures the situation within an achievement-related and person-related frame of reference, and that this takes place no earlier than ages 3 to 3 1/2, when success or failure of activity is associated with plea- sure or disappointment. There is a step in cognitive development which permits this structuring of the person-environment frame of reference. The lower threshold of three years for this cognitive step has also been observed in adult retardates. At a mental age of about 3 years, success and failure feelings are quite apparent in these individuals (Heckhausen, 1967). In addition, by age 4 or 4 1/2, normal children appear to select either a success orientation or a failure orientation, and individual differences in achievement motivation are apparent, such as those reported by McClelland (1958) in preferences for level of dif- ficulty in a ring-toss game. Further, their preferred levels of diffi- culty correlate with achievement motivation to the same degree as found in adults. In summary, then, it seems likely that the first manifesta- tion of achievement motivation depends on cognitive development. Several determinants of achievement motivation are associated with the social organizations of which individuals are part, in the form of cultures and subcultures. Child, Storm and Veroff (1958) found that in 52 preliterate cultures, achievement-related content of folk- tales was associated with high levels of achievement in children and with the dominant child rearing practices. Hayashi and Yamaushi (1964) 20 found that in Japan, insistence on self-reliance did not relate to children's achievement motivation. This is, of course, contrary to Winterbottom's conclusions. And Rosen (1962) has found that levels of achievement motivation vary culturally. Brazilian children, for example, tend to be lower in achievement motivation than North American children. Cole (1972) has also observed the existence of subcultural differences on the dimension of learning to learn. It is possible that some sub- cultures rely on learning and memory on a concrete basis, while others learn to learn in the abstract and with the aid of rich conceptual structure. In summary, then, within each cultural context, unique con- tributions are made on children's learning and achievement. Indeed, Greenfield and Bruner (1969) have evidence that sociocultural exper- ience, particularly in school, may influence the child's progress through Piagetian cognitive stages. One general implication of these studies on variables gener- ally removed from the control of educators might be that educators must simply take what they are given; it may well be out of their hands. Another possible implication might be that researchers have been con- cerned with the wrong variables. They have failed to direct attention to educational ecological variables which are at the disposal of edu- cators, and all that remains to be done is to identify these variables and learn how to successfully manipulate them. Or all this might mean that, in fact, researchers have been on the right track all along, recognizing that theories of achievement motivation are not likely to prove extremely useful to education. One final interpretation of this group of studies is that we have perhaps made the task of managing achievement motivation too 21 difficult, too involved and inclusive of too many determining variables. McClelland (1969) has demonstrated another and generally less compli- cated method of enhancing achievement motivation. The variables mani- pulated in McClelland's approach are quite simple. Individuals are taught to think, talk and act like persons with high nAch, and they are even trained to give TAT responses similar to those with high nAch. But since the relationship of nAch with academic achievement is not high, one might question whether direct training for achievement moti- vation holds real promise for education. As it turns out, achievement motivation training conducted by McClelland and his colleagues has not produced generally significant effects on school performance. The only differences in performance were observed in boys' performance, but not in girls'. Moreover, even the boys' performance increment was short- lived, being sustained for only one year (McClelland, 1972; Alschuler, 1971). Ryals (1969) has reported rather marked success, however, of a similar achievement motivation development project at Washington Uni- versity, under the direction of Richard deCharms. The difference be- tween the two projects was basically that in the deCharms program, students were given the achievement motive—related treatment by their teachers, whereas in the McClelland approach, regular classroom teachers did not play this important role. It seems, upon closer analysis, however, that these training techniques really might have little to do with enhancing students' achievement motivation. Yet, deCharms' technique did bring about in- crements in performance. The most obvious point is that the construct of achievement motivation has offered little in accounting for this 22 change in performance. And it is possible that the construct offers little more in any academic situation. McClelland (1972) suggests that the efficiency of the training program is largely due to its effects of teaching students to manage their lives more effectively and of teaching teachers to better manage their classrooms. But he tends to explain this finding by claiming that it is merely a matter of seman- tics. While nAch, the formal construct discussed by psychologists, is not enhanced, he observes, achievement thinking, achievement planning and achievement consciousness are all stimulated. But if the formal construct is not altered in ways predicted by theory as a consequence of prescribed education treatments, then it is questionable whether the construct has validity in the classroom. Rather than rely on the construct in any sense, it would be better to speak in terms of think- ing, planning and, if need be, consciousness. Sex Differences in Achievement Motivation A final limitation of achievement motivation, with respect to its educational implications, lies in the apparent dearth of evidence that the construct applies to females in the ways it has typically ap- plied to males. Indeed, females appear to differ from males on sev- eral aspects of achievement-oriented activity, as well. Females' in- terest in achievement is related specifically to the subject matter in school, for example. The value girls place on English, verbal skills, social skills, and artistic accomplishment is often higher than on natural sciences, athletic and mechanical skills (Battle, 1965, 1966; Stein, 1971). Sixth grade girls have also been known to have higher attainment values for tasks labeled "feminine" and "neutral" than on 23 "masculine" tasks (Stein, Pohly and Mueller, 1971). It appears that whereas males' achievement motivation is aroused by a challenge to intelligence or leadership, females' achievement mo- tivation is aroused by stress on social acceptability. Field (1953) fbund that when college females were informed of either social accep- tability or unacceptability, their achievement motivation scores were higher than when they received neutral information about their social acceptability. French and Lesser (1964) found that in colleges where most students valued traditional women's roles, social skills arousal produced greater achievement motivation. Generally speaking, women may be more ready and willing to express achievement motivation when the behavioral context is defined in terms of traditional women's roles. 0n the other hand, Frankel (1974) has found that non-traditional atti- tudes toward femininity may be positively related to achievement- oriented behavior, while a traditional orientation to femininity may be associated with non-goal oriented behavior. Equally suggestive of contradictory trends is the finding by Heilbrun, Kleemeier and Piccola (1974) that females with strong male gender identification and who per- ceive strong similarities between themselves and their fathers tend to have stronger achievement patterns. I One of the traditionally recognized and quite well accepted differences in the needs of males and females has been a predominance of affiliation need in females and achievement need in males. Crandall (1963) suggested that initially both boys' and girls' achievement- oriented behavior is aimed at obtaining social approval and rewards, but that boys learn standards of excellence and come to reward their own achievement, while girls fail to do so. Indeed, females are more 24 likely to be concerned with social desirability than males, but this concern is not consistently related to actual achievement efforts (Crandall, 1966). Although one would predict that females would ex- press relatively high effort when praise is given, such praise is not necessarily more efficacious than the absence of social reward in ex- perimental situations (Cotler and Palmer, 1971). In short, the effect Of social approval on females' achievement and achievement-oriented be- havior is not a consistent one. Stein and Bailey (1973) have summarized the literature on females' affiliation need, desire for social approval and achievement by suggesting that it is quite doubtful that females have a higher need for social approval or affiliation than males. While social skills might be a central area Of achievement concern for females, their actual achievement efforts are not always determined by the need for affiliation or social approval. Crandall, et a1. (1964) prOposed that females may not really be more sensitive to social approval; they nay simply receive more social approval for social patterns of achieve- ment. Females' sex role definitions may be quite uniform, yet some fe- males tend toward the Opinion that achievement is not necessarily mas- culine but is perhaps a quite feminine behavior. Lesser, Krawitz and Packard (1963) Observed that high school females who were high achievers considered achievement to be more sex-appropriate than did female under- achievers. Similarly, Lipman-Blumen (1972) found that women with non- traditional sex rOle concepts were more likely to consider their achieve- ments to be important. The meaning of these studies concerning women's achievement is that there are numerous variables which are not influential in males, 25 but which are powerful in determining women's achievement-oriented be- havior and in influencing women's achievement motivation. With such basic differences in optimal arousal conditions, sex-role orientations and patterns of reinforcements, it should not be surprising that, as O'Leary (1974) has noted, investigations of women's achievement has produced puzzling results. Not only are extant theories of achievement motivation derived from research on male samples, the projective instru- ments typically used to measure achievement motiVation in males are not highly correlated with actual achievement behaviors in females (Ent- wisle, 1972). This may be because females do respond differently from males when exposed to the cue of the competitive, competent and achiev- ing male. Developmental Trends in Females' Achievement Motivation Not only are there differences between males and females, but there also are clear developmental trends in females' achievement. Stein and Bailey (1973) point out that adolescence is the time of greatest change in achievement motivation. While many girls have av- erage or high levels of achievement motivation in junior high, at the senior high level, many apparently conclude that achievement is anti- thetical to their efforts at achieving sex role conformity. Douvan and Adelson (1966) suggest that females are left with a rather uncertain perSpective on adult career goals as a result of this basic conflict. Horner (1972) has Observed that college women are confronted with a similar conflict. Those who have studied traditionally masculine sub- jects Often move into fields which are more traditionally feminine. Even into adulthood, there are patterns of declining achievement 26 motivation, associated particularly with the years devoted to child rearing. Although declines are common during these years, increasing patterns of achievement motivation typically follow for well-educated women (Baruch, 1967). Expectancy of success in females shows other marked differences from males. Crandall (1969) found that females have lower expectancies of success than males even when their performance was superior. But this difference could not be accounted for by different reinforcement histories, parental differences in expectations or any other variable. It is possible that one determinant of this lower expectancy is the generalized sex role stereotype that females are simply less competent than males. Of most importance is the strength of the relationship between females' expectancies and their achievement patterns. Since these expectancies are a major component in the theory Of achievement motivation, one would inquire about whether the same theoretical rela- tionship between expectancies and achievement motivation exists for females as for males. While similar relationships have been Observed for adolescents and college females and males, data on female samples can best be described as showing a low to moderate relationship (Cran- dall, 1969). In another study, however, Crandall et al. (1962) found that elementary school females' expectancies were not positively re- lated to achievement behavior. This uncertain and apparently age- functional relationship calls into question the general applicability of an expectancy-value formulation of achievement motivation. There are other male-female differences which also place addi- tional demands on the explanatory power of expectancy-value motivational theory. Females have been occasionally seen as more anxious in academic 27 settings (Hill and Sarason, 1966). They also may tend toward setting very low levels of aspirations (Stein and Bailey, 1973). Socialization of Females' Achievement Orientation There are several studies which describe existing differences in male and female achievement patterns, but the socialization of these different patterns is another and equally important matter. Two of the most powerful factors in parental behavior are the warmth-hostility and the permissiveness-restrictiveness dimensions (Becker, 1964). There are some studies of mother-daughter interaction involving these factors which are suggestive of socialization antecedents of females' achieve- ment motivation, but most of these studies are focused on adolescents. A general conclusion pertaining to the socialization of achievement for young girls is that maternal nurturance is positively related to fem- ininity and negatively related to the emotional independence, asser- tiveness and competitiveness underlying achievement-oriented behavior (Mischel, 1970). Stein and Bailey (1973) caution, however, that the development of these traits requires sufficient affection and support so that a girl's confidence and security are not jeopardized. Adolescent females who express low achievement motivation in- deed tend to have mothers who were very affectionate during the early years (Kagan and Freeman, 1963). Concurrent maternal nurturance is also detrimental to achievement-oriented behavior (Bronfenbrenner, 1961). Parental permissiveness is generally positively related to fe- males' achievement motivation. Baumrind (1971) found that paternal permissiveness was also positively related to the achievement-oriented 28 behavior Of preschool girls. But Douvan and Adelson (1966) found that general parental restrictiveness is associated with adolescent females' achievement aspirations. Independence training is another parental factor which has Of- ten been associated with childrens' achievement motivation, but its relationship to achievement-oriented behavior is quite uncertain. This could be because it can take the fOrm of either encouragement or rejec- tion. Hatfield, Ferguson and Alpert (1967) found that the achievement standards of preschool girls were negatively related to either restric- tions on independence or to pressure to be independent. Another typical finding is that early maternal attempts at in- duction are associated with moderate to high levels Of aSpiration in middle-class preschool age girls (Collard, 1968). Mothers' and fathers' encouragement and instigation to achievement are related to high achievement activity in elementary school girls (Crandall, 1963). Stein and Bailey (1973) point out that early maternal acceleration, direct reinforcement Of achievement, parental restrictiveness and encouraging dependence also lead to female achievement behavior. Douvan and Adel- son (1966) Observe that some adolescent females with high achievement motivation name individuals outside the family as their most important models. Lansky et al. (1961) found that some adolescent females with high achievement motivation were highly critical of their mothers. While some socialization antecedents of achievement are shared by males and females, there are basic differences. Stein and Bailey (1973) observe that the child rearing practices most conducive to feminine sex typing are not those which usually lead to achievement. The female is most likely to develop achievement-oriented behavior when 29 her parents are moderately permissive and warm. If her parents encour- age independence, she is likely tO be independent, but this is not necessarily related to high achievement motivation. There does appear to be a contrast between the optimal child rearing behavior Of males and females, with respect to achievement mo- tivation. A high degree Of warmth tends to be associated with males' achievement, but not with females' (Crandall, 1963). Also, males seem to benefit most from slightly less permissiveness than is Optimal for females (Baumrind, 1971). Fear of Failure in Females Because there appear to be some sex differences in the actual goodness of fit of Atkinson's theory of achievement motivation, devel- opmental trends in achievement motivation and socialization antecedents of achievement motivation, some theorists have turned their attention to accentuating concepts which can adequately account for the nature and development of achievement motivation in women. One attempt at this task has been to heavily emphasize the fear of failure, or motive to avoid failure, which comes from Atkinson's theory. O'Leary (1974) suggested, for example, that fear of failure might be a factor in women's reluctance to aspire to higher level positions. It is proposed that the male must learn how to cape with failure as a part of his so- cialization experience. But females might not do this due to the ab- sence of social expectations that they will compete in the world of work. Kagan and Moss (1962) found a significant relationship in females between childhood and adulthood fear of failure, but the relationship was not found in males. 3O Females' high fear Of failure may be related to their unreal- istically high or cautiously low career goals (Mahone, 1960). It may also be related to females' often lower level of aspiration (Burnstein, 1963) and greater tendency to avoid the uncertainties and effort asso- ciated with a prestigious career (Heckhausen, 1967). The Theory of Motive to Avoid Success Other researchers remain unconvinced that an emphasis on fear Of failure can account for sex differences in achievement motivation. Matina Horner introduced the notion of motive to avoid success in order to account for otherwise inexplicable sex differences. The theory of motive to avoid success is directly derived from Atkinson's expectancy- value theory of achievement motivation. The motive is conceptualized as a stable, latent personality disposition which is acquired early in life along with standards of sex role identity. According to this formulation of motivation, the two powerful categories Of influence on one's behavior are the expectations one has about the nature and the likelihood of the consequences Of one's behavior and the value of these consequences to the individual. When one expects that the consequences of one's actions will be negative, then anxiety is elicited. This anxiety consequently inhibits the behavior which one predicts will bring about the negative consequences. The course one's behavior will not take can thus be predicted by expectancy-value theory (Atkinson and Feather, 1966). Horner (1968) argued that fear Of success is an individual's disposition to become anxious about achieving success because of his expectation Of negative consequences as a result Of succeeding. This 3I is not to say that he has a "will to fail" or a motive to approach failure. A motive to avoid success will not be aroused unless the individual expects negative consequences. Such an expectation will occur most Often in competitive situations where an individual must attend either to a standard of excellence or where his performance must be compared with the performance of others. Horner identified the need for this construct after reviewing sOme shortcomings of the achievement motivation literature, which have been identified above. Specifically, she observed that Veroff, Wilcox and Atkinson(l953) early identified scarcities on research pertinent to females. In The AchievingSociety_(l96l), Atkinson makes no mention of achievement motivation in females, even though he discusses other minorities. Several studies especially encouraged Horner that this new construct was needed. McClelland et a1. (1949), for example, found that women showed no increase in nAch imagery, as did men, when their intelligence and leadership ability were challenged. This result is still common today. Lesser et a1. (1963) compared groups of high- achieving and under-achieving high school girls. There was no signi- ficant difference between groups in nAch scores. But nAch scores of the high achievers did increase when they were asked to produce stories to pictures of females rather than Of males. In addition, the nAch scores of under-achieving girls declined when they were exposed to the female cues. But these results raise additional questions concerning the dynamics of such scores. Horner contended that extant theory was inadequate to the task of explaining these phenomena. Horner asserts that when these results are interpreted in terms of fear of success 32 theory, the dynamics can be identified. One would postulate that suc- cess would lead to expectation of negative consequences, and it is im- portant to identify what these negative consequences might be. Horner proposed that there might be two: 1) loss Of one's sense of femininity and one's self-esteem, and 2) social rejection. Without success, how- ever, involvement in competitive situations would not imply unfeminine behavior. In short, in achievement-oriented situations where the em- phasis is on intellectual and leadership skills, females' achievement motivation might be inhibited by the arousal of fear of success. And the Atkinson theory of achievement-oriented behavior, since it does not consider this construct, will not be able to differentially predict the behavior of males and females. Horner (1974) enumerates several of the notions surrounding the construct of M-s: 1. The stable trait of motive to avoid success is a disposition to feel uncomfortable when successful in competition, since this is inconsistent with femininity, and females expect negative consequences, including social rejection, to re- sult. 2. The M-s is more common in females than males. This is pre- dicted because competition is more consistent with mascu- linity than with femininity. 3. Fear of success is probably stronger in women whose achieve- ment motivation is high or whose ability is high. Other women probably do not consider success to be an important goal, so they have no reason to feel anxious about the pos- sibility that they might succeed. 33 4. Fear of success will be aroused more strongly in competi- tive situations, either with internal standards Of excel- lence or with competition against others. 5. The tendency to avoid success functions to inhibit the positive tendency to achieve success [TA = (Ts - T-f) - T-s + T ext]. 6. The strength of the tendency to avoid success is a multi- plicative function of motive strength, the incentive and the probability of success (T-s = M-s X Ps X Ias). 7. The negative incentive value of success (-Ias) will be higher in competitive than in non-competitive achievement situations, when women are competing against males rather than females. 8. The relationship between (Ias) and (P5) is an uncertain one. Several factors in addition to P5 might be Operative. Research on Motive to Avoid Success Recognition Of the construct of motive to avoid success came following Horner's doctoral dissertation (1968). In this study, male and female college students at the University of Michigan first re- sponded to several verbal cues of the Thematic Apperception Test and performed several tasks in mixed sex groups. Motive to avoid success was inferred from reSponses to the single cue: "After the first—term finals Anne (John) finds herself (himself) at the top Of her (his) medical school class." Female subjects responded only to the Anne cue and male subjects responded only to the John cue. Fear Of success was judged to be present if subjects' responses evidenced negative 34 consequences due to success, avoidance of future success, expressions Of conflict over success, denial Of the responsibility fOr succeeding or bizarre responses. Fear of success was scored as either present or absent. The results Of this part of the experiment were that 65.5% of females, but only 9.1% of males, evidenced fear of success responses. Social rejection, concern for normality and femininity and denial or bizarre responses were most common in females' reSponses. The second part of the study required random assignment of sub- jects to non-competitive or competitive conditions. All conditions re— quired the subjects to perform the same tasks. Females whose projective stories showed fear of success performed better in a non-competitive situation than in the competitive situation. Low fear Of success fe- males performed better in a competitive situation. It is on the basis of these results that Horner concluded that females have a higher level Of fear of success than males, and that this motive interferes with competitive task performance. Zuckerman and Wheeler (1975) have identified several weaknesses in Horner's study. Since the Anne cue refers to success in a profes- sion currently dominated by males, it is possible that females' re- sponses are specific indicators of anxiety about competition with men, rather than a general competition anxiety. Horner's instructions to subjects were somewhat questionable as well. Indeed, it is not certain whether they really implied a competitive task, since subjects in com- petitive groups were not told their performance would be compared with the performance of others. Further, the non-competitive instructions were quite similar to the competitive instructions. On these grounds alone, Zuckerman and Wheeler (1975) may be correct in questioning 35 whether Horner's conclusions were warranted. Other studies have also scrutinized the fear of success con- struct, with the general result of disconfirming most Of Horner's con- clusions and effectively challenging the theory of fear Of success. These studies, addressed to several factors relevant to the construct, are of particular interest. One of the predictions which might derive from fear of success theory is that the trait increases with age. Horner and Rhoem (1968) found that 47% of 7th grade girls reported fear of success imagery, but 88% Of undergraduate students evidenced such imagery. Baruch (1973) found that 10th grade girls constructed stories with much more fear of success imagery than 5th grade girls. But there are some contradictory results reported in other studies. Monahan, Kuhn and Shaver (1974) Observed that in a sample of 10 to 16 year old girls, fear Of success declined with age, while in a sample of 10 to 16 year Old boys, fear of success was unrelated to age. In a study of 18 to 50 year Old males and females, Moore (1974) found a negative relationship between fear of success and age. In short, Horner's prediction that fear Of success is positively associated with age is not supported. Since fear of success is not positively associa- ted with age, one could legitimately question whether it is a learned disposition. Another critical conclusion Of Horner's, that fear of success is more common to females than to males, has also been disconfirmed (Feather and Simon, 1973; Robbins and Robbins, 1973; Morgan and Mausner, 1973; Hoffman, 1974; Jackaway, 1974). Zuckerman and Wheeler (1975) ob- serve that the two explanations Of such discrepancies are that Horner's high fear of success results in women might have been due to the highly 36 competitive climate at the University of Michigan or that the women's movement might have recently attenuated fear of success. But on bal- ance, fear Of success does not appear to be related to trends in socio- cultural movements (Tresemer, 1974). Another prediction deriving from fear of success theory is that it would be more common in women who subscribe to more traditional be- liefs about women's roles. Indeed, Schwenn (1970) found that women with high fear Of success tend to study the humanities, which are usually considered to be more feminine. And Makosky (1972) reported that females with high fear of success believe that home and family life are more important than their career success. They also consi- dered themselves to be more feminine than was typical of low fear Of success females. But several studies show that such relationships are not widely generalizable (Peplau, 1973; Zanna, 1973; Moore, 1974). In fact, Heil- brun, Kleemeier and Piccola (1974) found that high levels of fear of success were related to a masculine orientation in female graduates. One would also predict a negative relationship between women's activism, related to the promotion Of non-traditional roles for women. Yet Unger and Krooth (1974) found no difference in fear of success among women classified as activists and those classified as non- activists. Similarly, Moore (1972) found no difference on fear of success between women who belonged to non-traditional academic groups and others who belonged to more traditional groups. Tangri (1974) has even reported a positive relationship between fear Of success and women's selection of non-sex-typical occupations. In summary, there seems to be little consistent evidence in support of the prediction that 37 women's fear of success is related to a traditional feminine role- orientation. The theory Of fear Of success suggests that the construct is an intrapsychic phenomenon, cutting across specific situations and repre- senting a conflict resulting in anxiety. But others have pointed out that it is quite plausible that fear Of success is merely a reflection of learning a predominant cultural stereotype that women do not in fact succeed as highly as men. The "cultural hypothesis" would predict that both males and females would have high fear Of success imagery concern- ing the Horner Anne cue. This is something which Horner did not in- vestigate, probably due to her assumption that the phenomenon derives solely from the eXpectancy-value theory of motivation. Several studies do, in fact, report that both males and females construct stories high in fear Of success imagery to the Anne cue, and the cultural hypothesis is given some degree Of support (Monahan, Kuhn and Shaver, 1974). Feather and Raphelson (1974) conducted research on American and Australian males and females. Half of the males and half Of the females were exposed to male cues, while the other subjects were shown female cues. Both male and female Australian subjects and Ameri- can males wrote a much higher proportion of fear of success stories to the female cue than to the male cue. American females, however, wrote a much lower proportion of fear Of success stories to the female cue than did the Australian females. Moreover, the proportion Of fear of success stories to the female cue was much lower for both American and Australian subjects than was noted in Horner's (1968) study. Feather and Raphelson (1974) attribute at least a portion of this decline in fear of success to the changing definition of women's roles associated 38 with the women's movement in our society. Another facet of the cultural hypothesis suggests that making Anne's success in the female cue less culturally atypical would reduce the frequency of negative reSponses to it. Katz (1971) tested this hypothesis by adding to Horner's Anne cue either one of the following sentences: "All Anne's classmates in medical school are men" Or "Half of Anne's classmates in medical school are women." The findings were that males evidenced more fear of success imagery to the deviant cue than to the nondeviant cue, but there was no difference in females' in- cidence Of fear Of success imagery related to the degree of deviance. Grainger, Kostick and Staley (1970) modified the Anne cue to read, "After first-term finals, Anne finds herself at the tOp of her nursing school." This is another way of ostensibly making the cue less cultur- ally deviant. The subjects responded with less fear Of success to this cue. 0n the other hand, some research suggests the nature of the cue has little to do with the amount of fear of success it elicits (Hoffman, 1974). Another prediction consistent with fear of success theory would be that males' fear of success stories would differ from females' in content. Several studies have found this to be the case (Hoffman, 1974; Morgan and Mausner, 1973; Zuckerman and Allison, 1973; Krusell, 1973). Males who question the value Of achieving tend to construct stories with high fear of success imagery, and these men are likely to build their stories around bizarre and/or hostile themes. This is quite different from the predominance in females' fear of success stories of the loss of femininity or the threat of social rejection. Among the other predictions one would make on the basis of fear 39 of success theory is that achievement motivation and fear of success are positively related. Horner's (1968) research revealed such a pos- itive'relationship between fear of success and achievement motivation. Yet, Horner (1972) has elsewhere suggested that womens' achievement- oriented activities are inhibited by their inconsistency with appro- priate sex-role standards, which causes anxiety. Zuckerman and Wheeler (1972) point out that on the basis of these views, it is not easy to predict whether women might be successful and anxious because of this success or passive and non-achieving because anxiety is sufficiently strong to inhibit achievement. In other words, it is not theoretically clear whether fear Of success constitutes a reaction to success or a factor which inhibits it. From a rather large number of studies on the relationship be- tween fear of success and achievement-oriented behavior, no clear pat- tern seems to be apparent. Horner (1968) reported a high level of fear of success among the honors students in her original sample, and Hoff- man (1974) supported this finding. Sorrentino and Short (1974) found strong relationships between IQ, grade point average and fear of suc- cess. While these three studies seem to suggest a consistent relation- ship, other studies have not been able to lend support to the findings (Zanna, 1973; Baruch, 1973; Zuckerman and Allison, 1973). While the use of correlations between fear of success and achievement-oriented behavior is an indirect way of identifying how fear of success is related to achievement motivation, another way of investigating this is through direct correlations with achievement mo- tivation. But the general conclusions of such studies are inconclusive as well. Horner (1968) found a non-significant difference between high 40 and low fear of success subjects on resultant achievement motivation, while in other studies, zero-order correlations have been found (Tangri, 1974; Sorrentino and Short, 1974). If this inconclusiveness is not enough, it should be noted that using Mehrabian's (1968) measure Of re- sultant achievement motivation, Zuckerman and Allison (1973) found a significant negative correlation with fear Of success for females, but not for males. It is this extremely inconclusive group Of studies which ap- pears tO call into question much of the foundation of the construct of motive to avoid success. After all, this construct was introduced ini- tially to explain why achievement motivation is not correlated with other measures of performance in women in the same ways it is in men. The original hypothesis was that women with high achievement motivation would also have a high level of fear of success. But obviously this prediction has not been supported. One is prompted to question whether this disconfirmation of such a central hypothesis does not do substan- tial damage to the theory. This evidence is particularly weighty when taken in combination with the other conclusions, that high fear of suc- cess females do not perform poorly under competitive conditions, that there are no reliable age or sex differences on the construct, and fear Of success and sex-role orientation appear to be unrelated, and that it is not clear whether fear Of success stems from cultural stereotypes or from intraphysic conflicts. In short, many of the notions which are easily derived from fear of success theory appear to be without sub- stance. 4I Predictive Validity_of Measures Of the Motive to Avoid Succe§§_ NO matter what else may be said about fear Of success, the most important evidence in favor Of this construct would be that in compe- titive, success-producing experiences, individuals who have high levels of fear of success, and who expect to succeed would perform poorly. The evidence has been inconclusive on this important prediction for a good reason. The situation in which the motive to avoid success is pre- sumed to be aroused has not been defined in specific terms. Therefore, different investigators have selected quite different dimensions of situations to evaluate, hoping to identify those which are indeed re- lated to motive to avoid success. Some have presented subjects with tasks inconsistent with their sex roles, some have made salient the value of success by reporting success on previous tasks, and some have manipulated both variables. But, taken all together, these studies have produced most inconsistent results. Three studies have recently been reported concerning the masu- linity/femininity of the task and the sex of the competitor. Makosky (1972) found that high fear of success females performed best in compe- tition against another female, while low fear of success females per- formed best in competition against a male. Also, high fear of success females competed more effectively on the feminine task than on a mascu- line task, while the conditions were reversed for low fear of success females. But Sorrentino and Short (1974) found, quite to the contrary, that high fear of success females performed best when the task was mas- culine rather than feminine. Karabenick has similarly reported that with a sample of 88 undergraduate females, following success experiences there was an improvement in the performance of these females who 42 competed with men rather than those who competed with women. Some studies have sought a relationship between fear of success and performance in a large group setting. Heilbrun, Kleemeier and Piccola (1974) found that females with high fear of success competed more effectively against males than did those with low fear of success. One restriction in this study, though, is that the performance task may have been judged to be unimportant by the subjects. Feather and Simon (1973) and Zuckerman and Allison (1973) found no relationship between performance and fear Of success for either males or females. There are two particularly interesting designs which have been used to manipulate the salience of success for subjects. Subjects may either be informed about their "success" on previous tasks, or they may be told that tasks are either "easy" or "difficult." These would appear to be functionally equivalent designs. 0n the former, subjects judge the probability of their succeeding on the task according to their pre- vious record of success. They construct a probability statement re- garding future success on the basis of their previous success. On the latter design, when subjects are told the difficulty level of the task, they are essentially told how everyone else has performed and will per- form on the task. Here again, they are left with what they need to ar- rive at a judgment of the probability of their success on the task. A task that has a high level of difficulty would indicate to the student that he has a small probability Of succeeding on it. More precisely, the student knows that, on the average, few students will succeed on the task and therefore his chances Of being in the group which succeeds are small. At least two studies have Obtained the effect of students' past 43 success on performance. Karabenick and Marshall (1974) found that low fear of success subjects who had a low fear of failure as well improved their performance more with a competitor present than when the compe- titor was absent. These subjects also improved more when competing against a male than against a female. Another study which manipulated students' previous success was conducted by Zaro (1972), producing findings which were not entirely consistent with the theory on fear Of success. TWO other studies have informed subjects about the difficulty of the task. Patty (1974) showed that high fear of success females performed better following directions that the tasks were easy, while low fear of success females had higher performance when the tasks were described as difficult. Zanna (1973) found, to the contrary, that fe- neles working on either easy or difficult tasks Opposite either males or females provided no evidence of a relationship between fear of suc- cess and performance. Although the results reported in these two experiments are contradictory, it would appear that the design Of these two studies is quite applicable to research in actual educational settings. The var- iables manipulated, previous success and in particular difficulty of the task, would be worthy of attention in future studies, and the lat- ter of these two variables will be manipulated in the present study. Reliability of Measures of the Motive to Avoid Success Zuckerman and Wheeler (1975) point out that the reliability of projective instruments such as Horner's original fantasy-based measure of fear of success are evaluated by three criteria of reliability: 1) 44 homogeneity reliability, or the variation of scores over different cues, 2) intertester reliability, and 3) test-retest reliability. Horner's original measure cannot be evaluated by the first criterion, since it has only one cue. Several studies concerning the administration of several cues to the same subjects, however, did not report correspon— dence in the ratings of fear of success across all cues (Weston and Madwick, 1970; Karabenick and Marshall, 1974). Entwisle (1972) sug- gests that the best estimate we have of homogeneity reliability of fear of success measures is .30-.40. While intertester reliabilities of .80 or .90 have been reported fOr Horner's original measure, this may be expected since a present- absent scoring system is used with this measure. 0n the other hand, the scoring procedures might have been quite variable from study to study, due to the absence Of clear-cut scoring rules. SO, while relia- bility might be high within one scoring system, reliability across raters who are accustomed to unique scoring approaches might drop markedly. Indeed, one is prompted to question whether the wide differ- ences in observed fear of success between males and females, from study to study, could be at least partly due to differences in scoring pro- cedures. Percentages of fear of success imagery have ranged from 20% to 88% among females and from 9% to 76% among males (Zuckerman and Wheeler, 1975). Shortcomings in the Literature and a Statement Of the Problem The construct of motive to avoid success was derived directly from the expectancy-value theory of achievement motivation, and it has been investigated with regard to the ways in which it is influenced by 45 the variables usually associated with achievement motivation. There- fore, it is limited not only in the degree Of understanding and pre- dictability it offers, but it probably is only a marginally useful edu- cational construct. But the current status of the research on fear of success does not permit any stronger assertion about the utility of the construct in education. There are several gaps in the research at those junctures which would associate the fear of success construct with edu- cational variables. For one thing, no study has asked about the effect Of students' fear of success as it might bear upon their actual compe- titive performance in extant, naturally assembled academic classrooms. Nor has any study inquired about the influence Of students' fear Of success on the meaningful evaluation criteria commonly encountered by all students in their classes. That is, while several studies have used such convenient but artificial competitive tasks as the generation anagrams task, there are no studies which have considered actual exam- inations as criteria. In short, this area of weakness in the literature means that there is no way to presently evaluate the validity of fear of success in educational practice. Since it is not sufficient to generalize from controlled studies using systematically matched pairs of subjects work- ing on artificial tasks to real educational situations, there is a need for research to bridge this gap. Another significant gap in the research has recently been created by the construction of several new instruments which purport to measure fear of success. Because the Horner fantasy-based technique was described as ambiguous and low in reliability, these new techniques were introduced to Offer researchers and clinicians ostensibly better 46 techniques. One Of these new techniques is the empirically-derived fantasy- based scoring system by Horner, Tresemer, Berens and Watson (1973). This technique utilizes three or four unspecified verbal TAT cues rather than the single cue used in Horner's fantasy-based system (see Appendix A). Each story is scored on six categories, which are reported to explain 45% of the variance Of performance decrements when subjects proceed from neutral to competitive conditions. Another new measure Of fear of success is Marice Pappo's (1972) objective questionnaire, which measures self-doubt, preoccupation with competition, preoccupation with evaluation, repudiation of competence and self-sabotage behavior (see Appendix B). Pappo (1972) reports that males and females who score high on this instrument significantly lower their performance on a digit symbol test when they are led to believe they have been successful on previous tests. But Curtis, Zanna and Campbell (1973) found that among law school students, Pappo's measure did not correlate with the original Horner measure. Moreover, it may be that the instrument in part measures will to fail. A third new instrument purporting to measure fear of success is the Zuckerman and Allison (1973) Fear of Success Scale (FOSS) (see Appendix C). Zuckerman and Allison have Obtained significant low cor- relations between FOSS and Horner's original method. Furthermore, Zuckerman (1975) has found female college students to have higher scores on the measure than males. Other research has shown that subjects with higher FOSS scores perform poorly on anagrams tasks under achievement- oriented instructions, when compared with subjects who score low on the measure. These findings would support the predictive validity of the 47 FOSS. The gap in the literature on fear Of success which was created by these new techniques is infOrmation about the reliability and val- idity of the measures. While some researchers appear ready to specu- late about the quality of these measures, there appears to be very little research on the indices of realiability and validity of each of them. Perhaps most urgent of all is the need to determine whether these techniques actually measure the same construct of fear of suc- cess. Hoffman (1976) suggests that the new Horner scoring system might measure fear of failure. Unless it can be determined that all these extant measures of fear of success really measure the same variable and that this variable is in fact the construct Of fear of success, the only outcome of further research which utilizes these instruments will be the addition of more gaps and inconsistencies in the literature. In a general sense, then, what seems to be called for is a general examination of the construct of fear of success and its indi- cants. It has come from the expectancy—value theory Of achievement motivation in an abstract manifestation, and it has yet to be applied to actual educational settings and to meaningful, competitive educa- tional tasks. Moreover, the relevance Of the construct stands ready to be made even more uncertain if a critical construct validation study is not conducted on its several ostensive measures. This study will provide evidence adequate for determining whether each of these tech- niques measures the construct of fear of success and whether these scales are valid for accounting for students' performance in academic settings, consistent with the theory on fear Of success. CHAPTER II METHODOLOGY Design and Background of the Method I TO assert that this study involves construct validation of fear of success scales does not necessarily define the methodology of the study. What it does suggest, however, is that fear of success has no single available criterion or no set of Operations which is considered fully valid alone. This is a crucial fact which differentiates between criterion-related validity and construct validity (Bechtold, 1951). Similarly, the Technical Recommendations (1954) expressed the original conception of construct validation as follows: . .The psychologist interested in construct validity for clinical devices is concerned with making an estimate of a hypothetical internal process, factor, system, structure or state and cannot expect to find a clear unitary behavioral criterion. An attempt to identify any one cri- terion measure or any composite as the criterion aimed at is, however, usually unwarranted (pp. 14, 15). Anastasi (1950) Observed even earlier that tests can only be prOperly interpreted if we "know the relationship between the tested behavior and other behavior samples, none Of these behavior samples necessarily occupying the preeminent position of a criterion" (p. 75). Cronbach and Meehl (1955) emphasize the value of clarifying constructs which are indeed indicated by performance, but are quite distinct from such per- formance. 48 49 Although there is no single acceptable criterion for the vali- dation Of a construct, it is Often possible to identify several var- iables which, according to the theory behind the construct, will be positively or negatively associated with a measure of the construct. (One of the most effective ways Of assessing construct validity is to determine whether an instrument measures the construct in ways predicted by its theory. Selection Of two theoretically different groups or man- ipulation of certain relevant conditions sets the occasion for deter- mining the instrument's ability to measure differences in the construct in ways predicted by the theory. A second method of assessing the construct validity Of an in- strument is to study the stability of test scores over time either with or without controlled experimental intervention. One would be inter- ested in knowing the degree to which changes in the individuals or the environment would bring about changes in test results. The meaning of either stability or change of test scores would depend on how they are accounted for by the theory. A third approach to construct validation is to correlate to- gether several instruments which all purport to measure the construct. If any two tests are presumed to measure the same construct, one would expect them to be positively correlated. If one uses this method of construct validation, one must be careful that high Obtained correla- tions do not result from irrelevant similarities of the tests rather than measurement of the construct. In other words, the dimensionality of the tests must be specified if this is to be the only method of con- ducting construct validation. Another problem associated with the use of this as a sole method is that assuming low correlations are Obtained 50 between tests 1 and 2, unless other evidence is available, one could never be certain whether this fault was in the construct itself rather than the tests. A fourth method of conducting a construct validation study is to Obtain correlations between single items or separate parts of the test. Assuming the construct is in fact conceptualized as a unitary variable, one would expect that the item-test and interitem correla- tions would be positive. While this kind of study Of the internal structure of the test is quite apprOpriate, another approach toward studying internal structure can be equally definitive: factor analysis. A classic article by Guilford (1946) explains the ways in which factor analysis provides a description of a test which is pure, stable and economical. While some argue that factors have no psychological reality, this criticism is not entirely true. Indeed, the value Of factors is that they are authentically derived from empirically measured indivi- dual differences. The intercorrelations of the individual differences determine the factors, and the name attributed to these individual fac- tors depend On the perceived commonalities in these groups Of inter~ correlated differences. When personality is defined in terms Of indi- vidual differences, then it would seem that factors do have real psy- chological existence. These factors are also real in the sense that they probably also correlate with determining variables in the environ- ment or in the biological makeup of the individual. In the realm Of construct validation, factors help in understanding why a test can pre- dict some criterion consistent with the theory behind the construct but not predict others. The degree to which the specific ways a test has construct validity can be described by a factor analysis of the scores 51 on the test. Finally, through the isolation of factors which are im- portant to the construct, the test constructor can discard items which contribute to irrelevant factors and create a univocal test, free of such factors. The deve10pment of a test which is a measure of a construct must begin with a theory Of the construct, or what Cronbach and Meehl (1955) describe as a "nomological network." This network specifies the relationship between the construct and other concepts and indicators of behavior. Moreover, the same nomological network, which defines the construct must be accepted widely so that this view of the construct is a predominant and a consensual one. Of course, in the early stages of development of the construct, the nomological network defining it is likely to be abbreviated and in- complete. The consequences Of this relative incompleteness Of para- meters is that the construct is not able to account for or predict a wide variety Of factors. As successively elaborating studies are con- ducted on the construct in areas just beyond the boundaries of the ex- tant nomological network, the boundaries are expanded, and the entire network becomes more richly elaborated. In time, the construct can be called upon to relate systematically to other constructs and events. Cronbach and Meehl (1955) Observe that the hypothesized network char- acterizes events which are part of an incompletely sampled cluster Of concepts; hence it is used to make predictions relative to yet unsampled areas in the space represented by the network. With any attempt at construct validation, one of two outcomes might result: 1) the test appears to have construct validity, or 2) the test does not appear to provide evidence consistent with the theory on 52 the construct. In the latter case, the meaning might be that: l) the test does not measure the construct 2) the theoretical or nomological network representing the construct is incorrect 3) a fault in the experiment prevented accurate testing of construct validity. But which of these interpretations will a researcher accept? If there are no obvious flaws in the experiment, upon reexamination Of the study, then he is faced with a difficult task. The theoretician will be tempted to conclude the fault lies in the instrument, while the psychometrician might be predisposed toward labeling the construct as faulty. Perhaps the strongest assertion which can be made is that the bridge between the test and the construct network is weak. It must be realized that this bridge needs to be strengthened, and that it is un— safe unless both ends are planted on firm soil. The most effective way to anchor the theory Of the construct is to test the accuracy with which it can make predictions relative to phenotypically diverse cri- terion variables. That is, the nomological network must be continuously extended to include variables other than those already incorporated in it. If all predictions made on the basis of a theoretical construct are pertinent only to a restricted range Of criteria, then the validity of the utilized measure is also restricted. Validity is strengthened most effectively when the instrument is tested with a new and relatively uncharted segment of the nomological network. In doing so, the pheno- typic space is enriched. One last point concerning construct validity on a general level is that most methods of construct validation will not produce a 53 coefficient of validity. Perhaps the best one can strive to attain is a more general notion about the construct saturation of a measure. Even on this index, it is practical to simply establish the upper and lower bounds of the reliable test variance which can be attributed to the construct. Assuming, for example, an ostensible measure Of selected variable A correlates .50 with a theoretically unrelated measure of variable B, the coefficient of determination of .25 indicates that 25% of the variance of the scores on selected variable A are accounted for by the other variable. This variance is irrelevant to variable A, and it could be considered an upper bound on the saturation of the test on variable A. If the measure Of variable A correlates with other and more theoretically defensible and consistent variables at the level of .70, the 49% Of the variance on the ostensive measure of variable A is apparently accounted for by the selected construct itself. This is the lower bound of construct saturation on the measure. In short, at least 49% and possibly as much as 75% of the variance on the test is accounted for by the selected variable. The establishment of such bounds Of con- struct saturation would appear to be quite a useful objective. With measures Of fear of success there is a strong need for establishing construct validity. MacFarlane (1942), in a pioneering article on the construct validation of projective techniques, called attention to the especially tenuous validity most Of these measures can boast. She particularly criticized the degree to which these tech- niques are interpreted in subjective ways. The three criteria for validation which she identified are: 1) that the concepts used must be explicitly identified, 2) that generalizations from these concepts should be limited to the samples originally used for validation, and 54 3) that appropriate validation techniques should be used in establish- ing the construct validity of these projective techniques. Fear of success has been theoretically defined to some degree, but its meaning is still uncertain due to the failure to explicitly delimit the meaning of "competitive situation" and due to some remain- ing confusion on how it is related to tasks which are not conveniently "feminine" or "masculine." On the matter Of generalization, fear of success has posed two problems. The nomological network has not been enriched by other interesting loci where fear Ofsuccess might assume an influential role. Indeed, as this study points out, it has not been effectively validated in the classroom with real classroom tests under authentic competitive conditions. Until now, it has merely been as- sumed that its applicability extended to such tasks. And that is the other problem concerning generalization: it has been too freely as- sumed that fear of success is generalizable from restricted, controlled experiments to virtually all other situations which could be loosely construed as similar to those constructed in the extant studies. Spence (1974) has questioned the assumption that fear of success is the gen- eral single disposition which would be required if such a degree Of generalizability were to be Obtained. Indeed, one is led to suSpect that Horner's selection of subjects, her selection of content for the single projective cue and others' emulations of these techniques might have identified a more specific phenomenon. Hoffman (1976) and Zucker- man and Wheeler (1975) have called for immediate attention to the qual- ity of the scales which purport to measure fear of success. With these considerations in mind, one of the ways this study attempts to establish the construct validity of each fear of success 55 instrument is to identify the degree to which the test can predict stu- dents' behavior on a final examination, consistent with the theory on fear Of success. This method is recommended by Cronbach and Meehl (1955). This important point here is that the criterion is one which has not been used before. It is not an "artificial" or contrived per- formance task imposed upon the academic setting. Unlike the perform- ance tasks frequently used in fear of success studies, such as gener- ation anagrams tasks, students are not likely to perceive this task as an unimportant one. IHeilbrun, Kleemeier and Piccola's (1974) subjects did not believe that successful performance on a digit-symbol task was important. In general, this aspect of the study establishes the strength Of the relationship between fear of success and a segment of the nomological network relating to classroom performance. The study is also designed to determine whether the three new measures of fear of success yield scores related to other character- istics Of the tasks and the subjects as predicted by the theory Of fear of success. The other variables selected here are considered to be the central concepts of the nomological network of fear of success: sub- jects' perceptions of the probability of succeeding on the task and the sex Of the subjects. The most straightforward way in which stu- dents are able to judge the probability of success on a task is to be informed of its difficulty level. Knowing that a test item is correctly answered by most other students, or in other words, that the item is easy, permits them to judge their own probability Of success on the item. Patty (1974) and Zanna (1973) used this method of manipulating students' perceptions of task difficulty but on less realistic perform- ance tasks. Subjects were manipulated into competition with a selected 56 group of subjects in ways which did not represent actual competitive conditions in classrooms. Also, the instructions suggested to the sub- jects that tasks were either difficult or easy, with no medium diffi- culty level to produce uncertainty. A final examination in any class is probably the single most important evaluation tool, and students are acutely aware of this. They also know that some items are more difficult than others. Another advantage of using the final examination as a performance criterion measure is that it can be conceptualized as several independent and equally important tasks. For each test item, students would probably find a description as either "high," "medium," or "low" difficulty as helpful in predicting their success. Extending the nomological net- work Of fear of success theory, one would hypothesize that overall exam scores would be related to level of fear of success and that this ef- fect might be dependent on the item difficulty levels as well. This would be the basis of an important research question. The sex of the subject is also of interest here. Some Of the literature reviewed in Chapter I reveals that sex was originally the key variable accounting for differences in fear of success, and some studies find persistent patterns Of evidence for this assertion. But these findings are far from consistent. The research question would ask whether the sex differences predicted by the theory would be ob- tained. In addition to the predictive power of fear of success, as measured by the selected instruments, another test of construct validity is affOrded by correlating together all the measures Of fear of success. This study is designed to correlate all the selected measures together 57 in order to investigate the research question of whether each of the instruments measures the same construct. As Alpert and Haber (1960) argue with respect to this method as they applied it to anxiety scales, high intercorrelations among the scales would suggest that there is in- deed a single underlying state of fear Of success, so that general measures are appropriate. Low correlations would call into question the existence of such a general trait of fear Of success and suggest that measures of such a general trait are misguided. An especially important research question would be whether the scales which purport to measure fear Of success actually correlate positively with a measure of fear Of failure. If this occurs, the con- struct of fear of success as a distinct and pure construct is placed in jeopardy. A significant correlation of any of the fear of success mea— sures with the fear Of failure measure means that a certain proportion Of the variance Of fear Of success can be accounted for by fear of failure. It is this proportion of variance which establishes the upper bound of the construct saturation on the measure of fear of success. Just as the correlation between fear of success scores and fear Of failure scores establishes the upper bound of an instrument's con- struct saturation on fear of success, the correlation between fear of success scores and another variable related to fear of success theory can be considered the lower bound of an instrument's construct satura- tion. Although research findings are mixed, it would seem theoretically logical that lower scores on low difficulty exam items ought to be Ob- tained by subjects with high fear of success than with low fear of suc- cess. High fear of success subjects would likely perceive the prospect Of success as a cue to avoid doing well. Although it may be premature 58 to make directional predictions of self-sabotage on the basis Of theory and research findings, the research question would be whether there is a negative correlation between fear Of success on each of the instru- ments and students' scores on the group of items labeled as “low" in difficulty. If the data Obtained in establishing the upper and lower bounds of construct saturation are combined, one Obtains a range of construct saturation. This is the most precise figure available for the construct saturation of each test, according to this analysis. But if all data from the several methods utilized in this study are intuitively aggre- gated, it would appear that a fairly accurate index of the general construct validity Of each instrument will be obtained. It is this general index, combining outcomes Of all analyses performed here, that is most useful for answering another crucial research question asking which end of the "bridge" Of construct validity, either the tests or the theory, should be revised or whether either end of the bridge needs strengthening. The design and instrumentation Of this study suggest other re- search questions. One is whether the scorer reliabilities on the pro- jective technique Of fear of success is apprOpriately high for this study. The other research question asks whether one of the measures Of fear Of success stands out as superior to the others on the basis of an aggregation of all the evidence, and what this apparent superiority im- plies for the use of the instrument. 59 Sample The subjects included 68 graduate students enrolled in courses in Standardized Testing, and Growth and Behavior as well as 14 from other classes, all in the College of Education at Michigan State Uni- versity during the summer term, 1976. The total sample size was, therefore, 82. Of these, 57 were females and 25 were males. On the basis of combined self-report indices of mother's and father's occupa- tion and mother's and father's educational attainment, coded as in Appendix F, 2 were classified in the lower-upper class, 10 were in the upper-middle class, 52 were of lower-middle class origins, and 18 were found to have upper-lower class origins. NO subjects could be des- cribed as members of either the upper-upper or the lower-lower classes. The mean age of the subjects was 28.79. See Figure l for a description of the age characteristics of the sample. Racially, the sample con- sisted of 78 whites and 4 blacks. Measures The scales with which the variables of interest were measured included a questionnaire designed to assess the background variables: social class, race, and age. But the questionnaire consisted largely of three purported new measures of fear of success and one measure of fear of failure. The fear of success techniques are Horner's empiri- cally derived fantasy-based instrument, Pappo's Fear of Success Ques- tionnaire (FOS), and Zuckerman and Allison's Fear of Success Scale (FOSS). Appendices A through C display these measures. Appendix E shows the segment of the questionnaire which measures the background variables alluded to above, including instructions to subjects for re- Sponding to the questionnaire. Number Of Subjects 50 45 40 35 30 25 20 15 IO 60 10 IO ..__1. 3 212m 22 23-27 28-32 33-37 38-42 43-47 48-52 53+ Age Figure l.--Age Composition Of the Sample by Class Intervals 61 As indicated in Chapter I, Horner's original fantasy-based measure Of fear of success was intended to fill the need for a measure of a dynamic trait which inhibited capable individuals, particularly women, from performing as their ability levels would otherwise permit. A verbal projective cue was selected rather than a picture because Horner believed it permitted the testing of males and females with equal effectiveness. It was alleged to do this because of the rela- tive ease of finding cues with similar values for males and females. Whether this is necessarily true is not certain, as indicated in the literature reviewed in Chapter 1, some Of which treated the inherent masculinity or femininity of the cues as the primary variable of in- terest. Horner cited such other successful attempts at using verbal cues as Winterbottom (1953), French (1955), and Atkinson and Litwin (1960). She selected six verbal cues altogether, the first four Of which were used to measure achievement motivation, while the other two were intended to explore sex differences in achievement motive. Only the sixth cue was intentionally included to measure motive to avoid success. The scoring system adOpted for the cue was a simple present- absent system. If there was any negative imagery about success, the stories were scored to contain fear of success. Any of the following negative responses was considered to indicate fear of success. 1) Negative consequences because of success. 2) Anticipation Of negative consequences because Of success. 3) Negative affect because Of success. 4) Instrumental activity away from present or future success, including leaving the field for more traditional female work 62 such as nursing, school teaching or social work. 5. Any direct expression of conflict about success. 6. Denial of the situation described by the cue. 7. Bizarre, inappropriate, unrealistic, or nonadaptive re- sponses to the situation described by the cue. For the cue, "After first-term finals, John (Anne) finds him- self (herself) at the top Of his (her) medical school class," Horner obtained 96% rescore reliabilities on 40 protocols and 90% reliability with an independent scorer on 90 protocols (Horner, 1974). But, as Observed in Chapter I, while typical scorer reliabilities are .80-.90, the wide range of Obtained fear of success imagery across studies raises a question about the possible existence of unique scoring systems across these experiments. And, of course, the validity of this measure is, as Zuckerman and Wheeler (1975) have Observed, uncertain. Possibly due to the weaknesses inherent in the original system, a new empirically derived fantasy-based scoring system was developed by Horner, Tresemer, Berens and Watson (1973). It assigns a fear of suc- cess score to individuals according to six categories of responses. For each of three or four undefined, unspecified cues, scoring is con- ducted as follows. (a) (+2) is scored when negative consequences are caused by the person. (b) (+2) is scored when negative consequences are caused by external factors. (c) (+2) is scored when two or more persons are involved with each other. (d) (+1) is scored when tension or deprivation is alleviated. 63 (e) (+1) is scored when there is no activity toward attaining a goal. (f) (-2) is scored when there are no characters in the story except for the person specified in the cue. NO reliability data have been reported on the new scoring sys- tem, but the authors recommend that an intertester reliability of .85 should be obtained before using the measure in research. One of the new Objective measures of fear of success is the Zuckerman and Allison (1973) Fear of Success Scale (FOSS). This is a 27 item questionnaire which asks subjects to either agree or disagree with items considered to be indicative of fear of success. The items describe either the benefits of success, the cost Of success or the reSpondent's attitudes toward success compared to other alternatives. About one-half of the items are scored positively in the direction of high fear of success, so that subjects' agreement is indicative of the construct. For the other items, disagreement is indicative of fear of success. The authors Of the test report low (.18) but significant correlation between the scale and Horner's original fantasy-based sys- tem for females. With both males and females in the sample, correla- tions of .19 (p«(.05) and .25 (p (.05) are reported with Horner's ori- ginal measure. NO reliability coefficients are reported for the measure, and Zuckerman and Wheeler (1975) conclude that it is too early to pass judgment on this or on the other new fear of success scales. The other new Objective fear of success scale is the 83 item yes-no questionnaire developed by Pappo (1972) as part of a doctoral dissertation. The Fear of Success Questionnaire (FOS) is suitable 64 either for administration in groups or individually. The reported re— liability Of the questionnaire is .90. The measure is constructed to tap the following five aspects of fear Of success: self-doubt, pre- occupation with competition, preoccupation with evaluation, repudia- tion of competence and self-sabotage behavior. Scoring Of the Pappo scale is according to the number Of questions which are responded to in the direction keyed to theory of fear of success. The measure of fear of failure used in this study is one which has been used for this purpose quite frequently (Horner, 1974): the Alpert-Haber (1960) Achievement Anxiety Test (AAT). When the test is used to measure fear Of failure, common practice is to use only half of the instrument: the Debilitating Anxiety Scale. This is a 10 item questionnaire which is sensitive to the anxiety which interferes with students' performance on academic tasks. All the items are based on the single factor of inhibiting students' performance. Students may possess a large amount of debilitating anxiety but little facilita- ting anxiety, which is measured by the Facilitating Anxiety Scale. The Debilitating Anxiety Scale has a test-retest reliability Of .87 after a 10 week interval and reliability Of .76 after an 8 month inter- val. When given as a whole, the AAT mixes the items from the FAS and the DAS along with some neutral buffer items. In this study, the DAS items will be given alone, without the other itens, in the interest of saving respondents' time in completing a rather lengthy questionnaire. The validation Of the Debilitating Anxiety Scale was carried out by Alpert and Haber (1960). They correlated the scale with the Mandler-Sarason Test Anxiety Scale (1952), a widely accepted specific anxiety measure, and Obtained a correlation of .64 (pr<.Ol). The 65 Debilitating Anxiety Scale is shown in Appendix D. The measure of performance under competitive conditions was the regular final examination. The only alteration made on this mea- sure was the addition Of individual item difficulties, indicated by the descriptors "High," "Medium," or "Low." Procedures One week prior to the final examination, 68 subjects in Stan- dardized Testing and Growth and Behavior were asked to complete a questionnaire outside of class to Obtain data on the background var- iables of socioeconomic status, race and age. In addition, these stu- dents' fear Of success was measured on each of the three fear of suc- cess instruments and fear Of failure was measured on the Alpert-Haber Debilitating Anxiety Scale. Fourteen additional subjects responded only to the background questions, fear of success and fear of failure measures. Race and social class data were gathered only to permit the description of the sample used in this validation of the selected scales. These variables were not built into the design of the study. In addition, subjects' sex was recorded. The total time required to complete the questionnaire was about 50 minutes. Appendix E displays the background variable questions. The second testing session was the final examination in the classes. In addition to the specifications of item difficulties on the exam, the only other special conditions incorporated into the ex- amination consisted of the instructions to subjects, shown in Appendix G. 66 Before distributing the final exam, each item was, of course, described according to its difficulty level as "High," "Medium," or "Low" difficulty. These descriptions accurately represent the exam items, based on the proportions of correct responses to the items when they were used on previous final examinations. That is, items in the upper third of item difficulties were described as "High" in diffi- culty; those in the middle third were labeled "Medium" in difficulty; and those in the lowest third were described as "Low" difficulty items. After data had been gathered on all measures incorporated into the questionnaire, the answer sheets of all subjects were scored by the experimenter for the background variables and the F05 and FOSS measures. In order to perform a test Of scorer reliability on the Horner empirically-derived measure, two raters, the experimenter plus another graduate student familiar with the scoring system, independently scored the stories, Obtaining a scorer reliability of .87, computed as a Pearson product-moment correlation. When the final examination scores were Obtained, subjects' scores on each group of exam items, "High," "Medium," and “Low" diffi- culty, were also recorded on each subject's questionnaire. To summar- ize, for each subject, data collected included: (a) Sex (b) SES (c) Race (d) Age (e) Score on Horner's instrument (f) FOS score (9) F055 score 67 (h) Alpert-Haber Debilitating Anxiety Scale score (i) Total score on "High" difficulty exam items (j) Total score on "Medium" difficulty exam items (k) Total score on "Low" difficulty exam items Students in all Standardized Testing classes were given a common final exam, and this test as well as the Growth and Behavior final exam were both specifically planned to consist of 54 items, evenly divided into low, medium and high difficulty items. Hypotheses A major part of the construct validation of the three fear of success measures selected here concerns the extent to which the scores on these instruments are related to overall final exam scores and to the scores on subsets of high, medium and low difficulty items. There is probably inadequate research evidence and insufficient theoretical background to predict the directional influence Of the interaction of fear Of success with item difficulty. And it may be that fear of suc- cess does not affect test performance unless it is aroused. In this study, the arousal condition is item difficulty level. Nevertheless, if a directional hypothesis concerning the total test score is appro- priate, one would predict that since fear of success is conceptualized as a stable trait of the individual, it might have a general inhibiting effect on test items, regardless of their difficulties. By the same token, one would expect subjects' scOres to be determined by item dif- ficulty levels, regardless Of the subjects' level of fear of success. For each fear of success instrument, several specific hypo- theses were tested. 68 Hypothesis I The overall final exam score will be lower for subjects who score high on the Horner projective instrument than for sub- jects who score low. Hypothesis II Obtained scores will be higher on items of low difficulty than on items of medium difficulty, and scores on medium difficulty items will be higher than scores on high difficulty items, re- gardless Of subjects' scores on the Horner instrument. Hypothesis III The scores of high fear of success subjects on the Horner in- strument will be unequal to those of low fear of success sub- jects at the low, medium and high item difficulty levels. Hypothesis IV The overall final exam score will be lower for subjects who score high on the Pappo FOS instrument than for subjects who score low. Hypothesis V Obtained scores will be higher on items of low difficulty than on items of medium difficulty, and scores on medium difficulty items will be higher than scores on high difficulty items, re- gardless of subjects' scores on the Pappo FOS instrument. Hypothesis VI The scores Of high fear of success subjects on the Pappo FOS instrument will be unequal to those of low fear of success sub- jects at the low, medium and high item difficulty levels. 69 Hypothesis VII The overall final exam score will be lower for subjects who score high on the Zuckerman and Allison FOSS than for subjects who score low. Hypothesis VIII Obtained scores will be higher on items of low difficulty than on items of medium difficulty, and scores on medium difficulty items will be higher than scores on high difficulty items, re- gardless of subjects' scores on the Zuckerman and Allison FOSS instrument. Hypothesis IX The scores of high fear Of success subjects on the Zuckerman and Allison FOSS instrument will be unequal to those Of low fear of success subjects at the low, medium and high item dif-- ficulty levels. Hypothesis X On the Horner fear of success instrument, a significantly higher proportion of females than males will obtain high fear of suc- cess scores. Hypothesis XI On the Pappo fear of success scale (FOS), a significantly higher proportion of females than males will Obtain high fear Of success scores. Hyppthesis XII On the Zuckerman and Allison fear of success scale (FOSS), a significantly higher proportion of females than males will ob- tain high fear of success scores. 7O Hypothesis XIII Obtained intercorrelations among scores on the three fear of success measures will be significantly positive. Hypothesis XIV Correlations between scores on each of the fear Of success measures and scores on a measure of fear of failure will be significantly negative. Hyppthesis XV Correlations between scores on each Of the fear of success measures and subjects' scores on the low difficulty final exam items will be significantly negative. The final research question shall remain formally unstated as a hypo- thesis to be tested statistically. This question is whether an aggre- gation of the data from all analyses conducted will indicate that the construct validity of each fear Of success measure is adequate. Analysis Several analytic procedures were utilized to test these hypo- theses. Hypothesis I through IX were tested in a series of three 2x3 univariate repeated measures ANOVAS. Scores on each fear of success measure were dichotomized as high or low at the median score, and these dichotomized scores served as one Of the main effects. A separate ANOVA was conducted for each measure of fear of success. The main ef- fects of level of fear of success and item difficulty and the_inter— action effect Of item difficulty X fear Of success all were tested for significance. Hypotheses X, XI and XII required a determination Of whether, 71 on each measure of fear Of success, significantly more females than males obtained high fear Of success scores. Chi-square tests were conducted in which the variables were fear of success measures and sex. Proportional differences between males and females on each of the fear Of success scales were tested at the .05 level of significance. Hypotheses XIII and XIV required that all the fear Of success scales and the measure of fear of success be intercorrelated. This re- quired the calculation of Pearson product-moment correlations. The Pappo FOS, Zuckerman and Allison's FOSS, Horner's empirically derived instrument and the Alpert-Haber DAS were all apprOpriately correlated together using the Pearson Product Moment. In Hypothesis XV the concern is with correlating scores on each Of the fear Of success scales with scores on the low difficulty items on the final examination. For this test, which is crucial in esta- blishing the lower bounds of the range of each scale's construct sat- uration, it is once again appropriate to calculate Pearson product- moment correlations between the low difficulty item scores and the FOSS, F05, and Horner's fantasy-based measure. The resulting corre- lation coefficients are necessary for the calculations of r2, the co- efficients Of determination, which are used as the lower bounds of the construct saturations. The last research question asks whether the evidence from all hypothesis tests leads to the conclusion that each measure Of fear of success has demonstrable construct validity. .This involves the final analytic step of combining data Obtained from the item difficulty and sex differences analyses, the intercorrelations of the fear of success measures, the correlations of the fear of success measures with the 72 fear of failure measure and the degree Of construct saturation Of each measure. This combination of evidence is used to judge whether there is support for the assertion that each instrument has construct vali- dity. Finally, this aggregation of data from all hypothesis tests per- mits general conclusions concerning the ability of the nomological net- work Of construct validity to encompass the effect on performance on academic tasks. Summar The sample for this study consisted of 82 students enrolled in graduate courses in the College of Education at Michigan State Univer- sity, including 25 males and 57 females. Data were collected on race, age, SES, three measures of fear of success and one measure of fear of failure. In addition, scores or numbers correct were recorded for final examination items in three groups: "high" difficulty, "medium" difficulty, and "low" difficulty. The study was done in order to test fifteen research hypotheses which, taken separately provide evidence for several indices of con- struct validity, and taken together permit making an assertion about the degree of construct validity of each of the measures. In testing these hypotheses, several analytic techniques were used. Repeated measures ANOVA was used to test the extent to which theoretically pre- dicted differences on the basis of item difficulties had been obtained, and chi-square tests were used to determine whether proportional sex differences consistent with fear of success theory had been Obtained. To determine the strengths Of the correlations among the three fear Of success scales, the fear of failure scales and the scores on the "low" 73 difficulty items, Pearson product-homent correlations were calculated. While the correlations among fear of success scales were analyzed in reaching a conclusion about the construct validity of each scale, the correlations between the fear Of success scales and the fear of fail- ure scale were used for this reason and additionally to establish the upper bound of the construct saturation of each of the fear of success measures. This was calculated as the prOportion of fear of success variance accounted for by scores on the fear of failure measure. TO obtain this proportion, the coefficient of determination, r2, was cal- culated. To establish the lower bound of construct saturation, corre- lations were obtained between the fear of success scores and scores on the "low" difficulty exam items. This performance criterion was sel- ected to establish the lower bound because it would appear to be an ideal variable in the educational setting. Moreover, it might be predictable from fear of success theory that students who obtain high fear Of success scores will Obtain low scores for those test items on which they have a good chance of succeeding. Pearson product-moment correlations were calculated between these score distributions. Finally, all evidence from the several analyses were aggregated to provide a basis on which to judge the magnitude of the construct validity of each of the scales. Interpretations and conclusions were derived from these data regarding the construct validity of each Of the measures as well as the strength of the construct itself. ' CHAPTER III ANALYSES AND RESULTS Introduction In this chapter the results Of the several analyses conducted will be presented, together with brief explanations Of the rationales and/or justifications for selection Of the methods Of analysis. All analyses were conducted using the CDC 6500 computer at the Michigan State University Computer Center. It will be both appropriate and efficient to restate each research hypothesis, discuss the outcome of the test of the hypothesis and describe the analysis which permitted the outcome to be obtained. In the course of this description, three different analyses will be discussed. The first is a set of repeated measures analysis of variance tests, each with three repeated measures on two groups. Second is a series Of chi-square tests, with one test for each fear of success instrument. The third analysis is an investi- gation of the correlations between the three measures of fear Of suc- cess and the measure of fear of failure. To conduct these analyses, two computer programs were utilized by the CDC 6500 computer. The first was Statistical Package for the Social Sciences (SPSS). This program calculated the Pearson product- moment correlations, and the chi-squares and, in addition, provided complete distributions and a full complement of statistics for all variables. The second program was PROFILE. This program is 74 75 specifically written to do split-plot or repeated measures ANOVA with any number of repeated measures and with any number of groups. It is also written to accommodate unequal cell sizes in the groups. In short, it accomplishes precisely the type Of analysis required in this study. It furnishes, in addition to the ANOVA table, a calculation Of conser- vative degrees of freedom required for the Greenhouse-Geisser test. This conservative test is necessary in the event that the Box Chi-square test Of homogeneity in the variance-covariance matrix is significant (Kirk, 1968). Following are results of all fifteen hypothesis tests. The Effect of Fear of Success on Examination Performance Hypothesis I The overall final exam score will be lower for subjects who score high on the Horner projective instrument than for sub- jects who score low. Hyppthesis II Obtained scores will be higher on items of low difficulty than on items of medium difficulty, and scores on medium difficulty items will be higher than scores on high difficulty items, re- gardless of subjects' scores on the Horner instrument. Hypothesis III The scores of high fear of success subjects on the Horner in- strument will be unequal to those Of low fear of success sub- jects at the low, medium and high item difficulty levels. These three hypotheses are grouped together, of course, because the repeated measures ANOVA provides a test Of all three simultaneously. These three hypotheses represent the tests of the two main affects of 76 fear of success and item difficulty and the interaction effect of fear of success by item difficulty. The results of each test are shown in Table l. Hypothesis III concerns the critical interaction effect. Sig- nificance of this interaction would mean that the subjects' performance on items of different difficulty levels was dependent upon their level of Horner fear of success. It is clear, however, that this interaction is not significant. It appears that there is no significant discrep- ancy between the scores Of high and low scoring subjects on the Horner measure at any level of item difficulty. Those small but nonsignifi- cant differences that were found are shown in Figure 2. It would ap- pear that the Obtained differences are not only statistically nonsig- nificant, but they may also be insignificant in terms of raw score dis- crepancies. Of special interest in the understanding of this nonsignificant interaction effect is the nature of the differences which do exist. Subjects who score high on Horner fear of success score somewhat higher on low and on high difficulty items, than do low Horner subjects. This relationship is reversed on medium difficulty items, however, where the scores of low difficulty items becomes relatively higher. Since the interaction effect is rejected, it is appropriate to proceed to tests Of the main effects: Horner fear of success and item difficulty. Hypothesis I addresses the former of these while Hypothesis II is concerned with the latter. Hypothesis I is rejected, since there is no significant differ- ence between high and low Horner fear Of success subjects on the total exam score. The other main effect, item difficulty, is clearly 77 Table l.--Results of Analysis Of Variance on Examination Items Using Horner's Empirically Derived System as a Measure Of Fear Of Success , S' .f._ Source SS df MS F caOSQILevel Horner Fear of Success (Groups) 2.121 1 2.121 .174 NS Subjects Within Groups 803.305 66 12.171 Item Difficulty (Repeated Measures) 849 . 098 2 424 .549 108. 757 p<.OOl Item Difficulty X Horner Fear of Success 14.955 2 7.477 1.916 NS Item Difficulty X Subjects Within Groups 515.280 132 3.903 Total 2184.759 203 78 18+ High Horner: __ __ __ Low Horner: 16 T- 14*- 12 — IO _ Number Correct 1 1 1 H M L Difficulty Level Figure 2.--Nunber Correct on Low, Medium and High Difficulty Final Exam Items for Subjects Scoring High and Low on the Horner Instrument. 79 significant (F = 108.7, p