CHILDRENS’ PERCEPTION 0F COMPETENCE AND OF SUCCESS CRITERION AND THEIR MATCHING FAMILIAR FIGURES TEST PERFORMANCE Thesis for the Degree of M. A. MICHIGAN STATE UNIVERSITY RENATE MAHLER 1977 .- _ .- .. -'. ,F‘de-c‘ A1. L I "i 7') a —.'_ ‘,' if , I 4“... " ' v ' .,I.,~..,,._ ( '4‘.“ ‘LI‘IJLI._- ‘ ‘. " 'v 51“»- L University ABSTRACT CHILDRENS' PERCEPTION OF COMPETENCE AND OF SUCCESS CRITERION AND THEIR MATCHING FAMILIAR FIGURES TEST PERFORMANCE By Renate Mahler Kagan, Rosman, Day, Albert, and Phillips (1964) devised the Matching Familiar Figures Test (MFFT) to measure individual differences in cognitive tempo that they labeled as "reflection-impulsivity." The MFFT has recently been criticized by Block, Block, and Harrington (1974) and others on conceptual and methodological bases. One criticism was that MFFT response latency is confounded with MFFT response accuracy in the operationalization of "reflective” and "impulsive" response styles. The purpose of the present study was to determine whether MFFT response style is a "predispositional” tendency of the child as proposed by Kagan et a1. (1964) or whether variables associated with MFFT latency and accuracy, respectively, combine to influence the MFFT response style of the child. On the basis of the literature review, it was expected that the child's perception of the task success criterion would influence MFFT latency and that the child's perception of his MFFT competence would affect his MFFT accuracy. It was predicted that children who believed speed plus accuracy to be the success criterion would show shorter MFFT latencies than would children who believed only accuracy to be the success criterion. It was also predicted that children who believed themselves to be competent on the MFFT, because they were told so by the test administrator, would make fewer MFFT errors than children who were told nothing about their MFFT competence. Furthermore, it was predicted Renate Mahler that the task success criterion would not affect MFFT error scores and that competence perception would not affect MFFT latency. One hundred twenty-eight second- and third-grade boys were administered the MFFT in two test sessions. The task success criterion and competence manipulations were introduced at the beginning of the posttest session, four weeks after subjects had been given the MFFT for pretest classification purposes. The results were analyzed with analyses of variance and Scheffe post hoc comparisons. The results indicated that the children who believed the task success criterion to be speed plus accuracy showed shorter latency than did children in the accuracy only and control conditions. The MFFT error scores, however, also increased for the children in the speed plus accuracy condition, which indicated that MFFT errors were indirectly related to the success criterion for these children. Competence perception, as defined in the present study, was unrelated to MFFT performance. Overall, the MFFT response classification categories to which children had been assigned were unaltered after the introduction of the experimental manipulations. Results were discussed in terms of a speed—accuracy tradeoff strategy. Future research should establish whether intellectual and personality factors are related to information processing strategies on the MFFT. CHILDRENS' PERCEPTION OF COMPETENCE AND OF SUCCESS CRITERION AND THEIR MATCHING FAMILIAR FIGURES TEST PERFORMANCE BY Renate Mahler A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1977 To my true friends, who transcend time and space. ii ACKNOWLEDGMENTS I express my appreciation to Dr. Ellen Strommen, thesis chairperson, for her advice and guidance throughout my work on this thesis. The completion of this work is credited to her encouragement, patience, and sincerity. Thanks are extended to Dr. Patricia Busk for her expertise and accessibility during the statistical analysis of the data as well as for her constructive criticisms of the writing of the thesis. Dr. Lauren Harris also criticized the writing of the thesis, and was especially . helpful on questions of methodology and design. iii TABLE OF CONTENTS pag§_ LIST OF TABLES v LIST OF FIGURES vi INTRODUCTION 1 Overview of Literature Review ..................... ....... ........ 2 Correlates and Experimental Antecedents of MFFT Performance ...... 3 Modifiability and Generality of MFFT Response Style ........ . ..... 8 Present Study............ ...................................... .13 METHOD 16 Subjects........................................................16 Apparatus .......... .... ....................... ....... ........ ...16 Procedure........... ................................... . ........ 18 RESULTS 23 Analyses......... ...................................... . ........ 23 Latency Scores .......................................... . ....... 24 Error Scores ...... . ............................................. 32 Summary of Statistical Analyses ................................. 45 DISCUSSION 47 APPENDIX A: STANDARD INSTRUCTIONS FOR THE MFFT 54 APPENDIX B: POSTEXPERIMENTAL INTERVIEW SS REFERENCE NOTES 56 LIST OF REFERENCES 57 iv LIST OF TABLES Table page 1. MFFT Posttest Latency Means and Standard Deviations for ' Competence, Success Criterion, and Pretest Classification Groups (2 x 2 x 4 ANOVA) .............. . .......................... 25 2. Summary of 2 x 2 x 4 Least Squares Analysis of Variance of MFFT Posttest Latency (Seconds to First Response) Scores ......... 26 3. Mean Posttest Latency Differences (in Seconds) for Scheffe Post Hoc Pairwise Comparisons of Classification Levels (2 x 2 x 4 ANOVA) ....... . ...................................... . ............ 27 4. Latency to First Response Score Means (in Seconds) and Standard Deviations for Measures, Success Criterion, and Pretest Classification Groups (2 x 3 ANOVA) .................. ....29 5. Summary of 2 x 3 Least Squares Repeated Measures Analysis of Variance for MFFT Seconds to First Response Scores ............ 3O 6. Mean Total Posttest Error Scores Per Test, Standard Deviations, and Frequencies of Competence, Success Criterion, and Pretest Classification Groups (2 x 2 x 4 ANOVA) ............ . ..... . ....... 35 7. Summary of 2 x 2 x 4 Least Squares Analysis of Variance of MFFT Posttest Total Errors ....................................... 36 8. Mean Posttest Error Score Differences for Scheffe Post Hoc Pairwise Comparisons of Classification Levels (2 x 2 x 4 ANOVA) ....... . ..... . ............................................. 38 9. Total Error Score Means and Standard Deviations for Measures, Success Criterion, and Pretest Classification Groups (2 x 3 ANOVA) ........ . .......................... . ................ 41 10. Summary of 2 x 3 Least Squares Repeated Measures Analysis of Variance for MFFT Error Scores ................................ 42 LIST OF FIGURES Figure age 1. Mean MFFT Latency Scores Per Item for Pre- and Posttest Sessions as a Function of Classification Group ...... .. ........... 33 2. Mean MFFT Latency Scores Per Item for Pre- and Posttest Sessions as a Function of Criterion Group ........... . ........ ....34 3. Mean Total Errors Per Test for Pre- and Posttest Sessions as a Function of Criterion Group ................ . ..... ...........37 4. Mean Total Errors Per Test for Pre- and Posttest Sessions as a Function of Classification Group ............................ 43 vi INTRODUCTION Kagan, Rosman, Day, Albert, and Phillips (1964) devised the Matching Familiar Figures Test (MFFT) to measure individual differences in the problem-solving strategy of children. The MFFT involves a high degree of response uncertainty, since the task of the child is to select, from an array of several alternatives that subtly vary from a standard line drawing, the line drawing of a familiar object that is identical to the standard (Kagan and Kogan, 1970). Children who responded slowly and accurately on the MFFT were labeled "reflective," whereas those who responded quickly and inaccurately were labeled "impulsive." As these labels imply, Kagan et a1. (1964) conceptualized the MFFT as a measure of cognitive tempo. Block, Block, and Harrington (1974), however, recently noted that both MFFT latency and MFFT accuracy are used in the opera-. tionalization of ”reflection-impulsivity." Since they found that personality factors correlated with MFFT accuracy, but not with MFFT latency, they suggested that the MFFT is a confounded measure of cognieg_ tive tempo, and it is best to consider MFFT latency and MFFT accuracy as two independent dimensions. The present study examined MFFT latency and accuracy as independent factors. The literature review suggests that MFFT latency appears to be affected by task factors, whereas MFFT accuracy appears to be related to stable characteristics of the individual. The purpose of the present study was to determine whether this general trend of the literature 1 operates at the individual level to affect the outcome of a child's MFFT classification. In this study, the task factor of the MFFT success criterion perception (i. e., whether speed or accuracy was important) and the personal characteristic of MFFT competence perception (i. e., being told or not told that prior MFFT performance was good) were mani- pulated. If MFFT latency scores of the child can be explained by task variables and MFFT error scores by the stable characteristics of the individual, then perception of the MFFT success criterion should influ- ence MFFT latency performance and perception of MFFT competence should affect MFFT accuracy performance. Overview of Literature Review An abundance of literature that addresses cognitive style as it is defined by the MFFT has appeared since the topic was introduced by Kagan et a1. (1964). The literature deals with three basic topics, the trends of which suggest and support the predictions of the present study. The first topic is concerned with factors associated with cognitive tempo, including correlates such as personality factors, attitude, intelligence, and sex and experimental antecedents of MFFT performance such as task structure factors and performance anxiety. Consistent with the predic- tions of this study, the correlates of MFFT performance (i. e., the stable characteristics of the individual) are primarily related to MFFT error scores, and the experimental antecedents (i. e., those dealing with task structure) affect MFFT latency scores. The second topic deals with the modification of "reflective" and "impulsive" cognitive styles through modeling and verbal training techniques and with the generality of "reflection-impulsivity" as a behavioral measure. Subtopics of the generality issue are motor inhibition, correlation among latency indices, 3 visual scanning and information processing strategies, risk-taking behavior, and the ability to delay gratification. The modification literature reveals that MFFT latency scores are readily changed through instructions, but that MFFT error scores are not. This trend supports the predictions that MFFT latency is affected by task structure vari- ables, whereas MFFT error scores are more stable, perhaps related to characteristics of the individual. Since MFFT latency has generality as a behavioral measure to the degree that the task to which MFFT latency is compared is similar in structure to the MFFT, the trend of this litera- ture review is congruent with the present prediction that MFFT latency is related to task variables. MFFT latency is modifiable and generalizable, whereas MFFT accuracy is not. The third issue that is addressed below, in conjunction with the other two t0pics, deals with the adequacy of the MFFT as a measure of cognitive tempo. Correlates and Experimental Antecedents of MFFT Performance Correlates. Generally, personality factors and intelligence correlate with MFFT accuracy. Studies, however, have produced incon- sistant results, in part because of methodological inadequacies. Several investigators have used teacher ratings of personality; such ratings may be methodologically unsound and biased. Furthermore, most investigators did not examine MFFT latency and accuracy performance separately (Ault, Crawford, and Jeffrey, 1972; Bjorklund and Butter, 1973; Nadeau, 1968). Nadeau (1968) found no relationship between teachers' ratings of personality and MFFT performance. Bjorklund and Butter (1973) found no correlation between MFFT response style and the Impulsivity Scale for Children nor between MFFT response style and teacher ratings of preschool-age children. Ault et a1. (1972) examined 4 the correlation between teacher ratings of attention, hyperactivity, and motivation of school-age children and MFFT latency and error scores in a design that included the traditionally excluded "fast-accurate" and "slow-inaccurate" MFFT performance groups. Teachers, unaware of the children's MFFT scores, rated "slow-accurate" children as significantly higher on attention than "fast-inaccurate," "fast-accurate," and "slow- inaccurate" children. No significant correlations were obtained for motivation ratings. Ratings of hyperactivity showed that children having high MFFT error scores were rated as more hyperactive than children with low MFFT error scores, irrespective of MFFT latency to first response classification. It is impossible, however, to determine the actual correlation between personality traits and MFFT latency and accuracy scores in the Ault et a1. (1972) study, since no boys were classified as "fast-accurate," and thus, teacher ratings were confounded with the sex of the child. Block et al. (1974) avoided both of the methodological errors mentioned above. They used a standardized personality test, a modified version of the California Q set, in place of teacher ratings. They also defined four MFFT response style quadrants by pairing median splits of MFFT error scores and MFFT latency to first response scores, thus uncon- founding the latency and accuracy components of the MFFT. Only two personality attributes were significantly related to MFFT response latency, but thirty-two attributes were significantly related to MFFT response accuracy. Block et a1. (1974) commented that it is possible that the "predispositional" differences between "reflective" and "impulsive" response styles of children referred to by Kagan et a1. (1964) may be primarily associated with MFFT errors because stable 5 personality traits were found to be associated with MFFT errors rather than with MFFT latency and error—latency interactions. Generally, the children who displayed few errors on the MFFT showed the following traits: competence, intelligence, perceptiveness, and interpersonal attractiveness. Those children who made many errors (i. e., who were "inaccurate") on the MFFT showed rigidity, over- sensitivity, and a lack of self-confidence (Block et al., 1974). Attitudes also have been found to correlate with MFFT response style. It is impossible, however, to ascertain whether the correlation is stronger for MFFT accuracy or latency. Adams (1972) and Schack and Massari (1971) found that children categorized as "slow-accurate" on the MFFT had a greater expectation for success than children categorized as "fast-inaccurate." Campbell and Douglas (1972) reported that "slow- accurate" children were more optimistic in situations of frustration and potential failure than were "fast-inaccurate" children. These findings may reflect the association between MFFT accuracy and ”competence" found by Block et a1. (1974), although the confounding of speed and accuracy in these attitude studies makes it impossible to determine whether this is the case. Whether the sex of the child is correlated with MFFT performance is unresolved, since research results are inconsistent (Lewis et al., 1968; Messer, Note 1). As Egeland and Weinberg (1976) have suggested, the effect of sex differences on MFFT performance is probably best examined with a developmental paradigm. In order to avoid problems which could arise if there are sex differences in MFFT performance, the present study was limited to one sex-~males, who show a stronger negative correlation than females between MFFT error and latency scores (Lewis et al., 1968). Experimental antecedents. Both task structure variables and performance anxiety are experimental antecedents to MFFT latency. Several investigators (Bush and Dweck, 1975; Rhetts, 1974; Ward, 1968; Weiner and Adams, 1974) have suggested that "reflection-impulsivity" may be due to situational task variables. Bush and Dweck (1975) found that children of "slow-accurate" cognitive style exhibited long latency on the MFFT and short latency on speeded tasks according to the situational requirements of the task. They suggest that "fast-inaccurate" children did not attend to or use the situational cues that indicated what con- stituted an appropriate response to a task. Likewise, Rhetts (1974) suggested that it is possible that learner and task characteristics interact and result in differences in MFFT latency. Weiner and Adams (1974) proposed that MFFT response style may be related to the rein- forcement history of the child, since they found that consistent feedback in a failure condition resulted in a latency increase, but inconsistent feedback in a frustration condition did not. Speed task instructions and accuracy task instructions may also interact with an "evaluative" or "permissive" test situation, affecting MFFT response style (Ward, 1968). A conclusion of these studies is that the task structure may interact with the characteristics of the individual to affect MFFT latency. Task performance anxiety appears to be related to MFFT latency. Kagan et a1. (1964) first suggested that anxiety over failure could lead to either "reflective" or "impulsive" MFFT responses. On the one hand, the child who was anxious about possible rejection by the test adminis- trator might behave "reflectively" on the MFFT because he lengthened his time to enhance the possibility of providing a correct answer. 0n the other hand, anxiety could produce "impulsive" MFFT performance because the child presumably found the silence between himself and the test administrator impossible to bear during problem solving. Kagan et a1. (1964, Study 6) found no significant differences between children tested by an "impersonal" silent versus a "reassuring" talkative test adminis- trator, so the performance differences of children on the MFFT were interpreted as reflecting ”predispositional" fundamental response tendencies of the child rather than situational anxiety factors. Since Kagan et al.'s (1964) Study 6, however, other investigators have found anxiety to have a significant affect as a situational ante- cedent to conceptual tempo. Ward (1968) investigated the effect of failure feedback that followed the individual MFFT test item and found that failure feedback was associated with longer latency on MFFT items that immediately followed such feedback. If it is assumed that the failure feedback provoked anxiety, then Ward's (1968) finding supported Kagan et al.'s (1964) conjecture that anxiety leads to "reflective” responses because the child attempted to investigate all solution hypotheses to avoid failure. Messer (1970) and Weiner and Adams (1974) used failure on an anagrams task as an anxiety provoker with third- and fourth-grade children, and confirmed Ward's (1968) results that anxiety significantly increased MFFT response latency, but did not affect MFFT errors. Bush and Dweck (1975) administered the Test Anxiety Scale for Children and Lie Scale for Children (Sarason et al., 1960) to fourth- grade children. They found that high-anxious "reflective" children and low-anxious "reflective" children behaved similarly on speeded tasks of increased difficulty and showed longer latency and greater accuracy on the MFFT than did "impulsive" children. The results of the Bush and Dweck (1975) study can be interpreted in two ways. Instruments used to measure anxiety may have been unrelated to the anxiety of the MFFT task. This possibility was suggested by Kagan and Messer (1975) when they pointed out the importance of Operationally defining anxiety. The other explanation is that anxiety had no effect on MFFT performance in the Bush and Dweck (1975) study. It is possible that a psychological characteristic other than anxiety, such as the personality factors suggested by Block et a1. (1974),.differentiated the MFFT performance of children classified as "reflective" and "impulsive." This review suggests that MFFT accuracy is related to the compe- tence, perceptiveness, and the intelligence of the individual. Optimism and success expectancy, found to be characteristic of "slow-accurates," may also be associated with the competence characteristic that Block et. a1. (1974) found correlated with MFFT accuracy. Performance anxiety, feedback conditions, and the perception of task requirements, however, affect MFFT latency independent of MFFT accuracy. Thus in this study, it was expected that manipulated competence perception would affect MFFT error scores, whereas the perception of the task success criterion would affect MFFT latency scores. Modifiability and Generality of MFFT Response Style Modifiability of MFFT response style. Research that deals with the modifiability of "impulsive" and "reflective" cognitive styles evolved primarily from the assumptions that "reflective" cognitive style is associated with superior problem-solving ability and that "impulsive" cognitive style is characteristic of children of low socioeconomic background (Kagan, 1967). As noted by Block et a1. (1974), MFFT latency can be significantly modified by modeling procedures (Cohen and Przycien, 1974; Debus, 1970; Meichenbaum and Goodman, 1971); errors, however, are not significantly decreased unless special training procedures are employed. The self-instruction of covert and overt ver- balization employed by Meichenbaum and Goodman (1971) was a training procedure that resulted in a significant decrease in MFFT errors. Since MFFT response latency is signficantly altered through modeling and verbal training procedures, the indication is that MFFT latency is affected by task structure variables. The finding, however, that MFFT errors are not significantly decreased unless the subject gave himself self-instruction as compared to the same instruction provided by another individual (Meichenbaum and Goodman, 1971) tends to support the possibility that such self-instruction results in a heightened sense of competence, usually only characteristic of children making few MFFT errors, and that this heightened sense of competence significantly affected MFFT accuracy. Block et a1. (1974) noted that it is difficult to determine the long term effects of self-instruction since no follow-up studies have been conducted. Generality of MFFT response style. An issue seperate from although related to MFFT response modifiability is whether the MFFT "reflective" and "impulsive" response style of an individual generalizes across situations, which is suggested by ascribing characteristics such as restlessness, distractibility, and hyperactivity to the "impulsive" child as was done by Kagan et a1. (1964) and Kagan and Kogan (1970). Variables explored in the study of MFFT response style generalizability are: motor inhibition, correlation among latency indices, visual scanning and information processing strategies, risk-taking behavior, ability to delay gratification, and personality correlates. All but the 10 last of these, which has already been discussed, appear to be unrelated to "reflection-impulsivity" as defined by the MFFT. Literature that deals with each of these variables is presented below. Motor inhibition. If "reflection-impulsivity" has generality as a behavioral dimension, then it can be expected that there is a correlation between motor activity and cognitive style; studies, however, have not indicated such a correlation. Inhibition of motor response on the Motor Inhibition Test failed to correlate with "reflectivity" on the MFFT (Shipman, 1971). Harrison and Nadelman (1972) reported a negative correlation between motor inhibition on the Motor Inhibition Test and error-per-time scores on the MFFT, though this result appears to be attributable to a confound with intelligence, as noted by Block et a1. (1974). Constantini, Corsini, and Davis (1973) found that motor inhibi- tion is inconsistent across age because preschoolsage boys labeled as "impulsive" on the MFFT had more difficulty inhibiting motor movement on the Wald-a—Board, Reel-up, and Finger Tap Tests than school-age boys labeled as "reflective" on the MFFT. No relationship was found between cognitive tempo and motor inhibition for girls at any age. Generally, these results do not support the concept that'RreflectionéimpulsiVity" is a general measure as indicated by motor response indices. Latency indices. Block et a1. (1974) provided an extensive review of the various latency indices that have been investigated as possible correlates of MFFT performance. Block et a1. (1974) noted that correla- tions are high between MFFT latency to first response scores and other measures when the cognitive tasks are highly similar in structure, require similar motor skills, similar intellectual competence, and elicit similar anxiety reactions. This positive correlation between MFFT 11 latency scores and other indices of latency that are similar in task structure tends to indicate that task structure and situational variables, rather than a stable attribute of the individual, determine MFFT latency. Visual scanningand information processing strategies. Variables that have received attention in an effort to understand the perceptual and cognitive components associated with conceptual tempo are visual scanning strategy and information processing strategy. Siegelman (1969) compared children classified as "reflective" and "impulsive" on the MFFT and found that the former group deployed rela- tively less looking time and less frequent looks to the standard figure of the MFFT. Drake (1970) found, however, just the opposite effect. Ault et al. (1972) found that all children employed a visual scanning strategy of pair comparisons, but that those children with low MFFT error scores employed the strategy more systematically than those children that provided many inaccurate responses on the MFFT, regardless of MFFT latency to first response scores. Thus no consistent relationship was found to exist between cognitive tempo and visual looking time. Recently, however, Zelniker and Jeffrey (1976) have suggested that "slow-accurate" children attend to the details of visual stimuli while "fast-inaccurate" children attend to the "global" characteristics of visual stimuli. What determines these visual scanning styles was not addressed by Zelniker and Jeffrey (1976). Jones and McIntyre (1976) also have indicated that different responses on the MFFT may be the result of information processing differences. In particular, they suggest that latency and accuracy classification are attributable to "game plans" and speed—accuracy tradeoffs. 12 Delay ofgratification and risk-taking_behavior. Other measures that should relate to the "globality" of cognitive style as measured by the MFFT are ability to delay gratification and also risk-taking beha- vior. These measures, however, fail to correlate with MFFT classifica- tion. No relationship was observed between MFFT latency or error scores and the Mischel Delay of Gratification task (Hess, Shipman, Brophy, and Baer, 1969; Shipman, 1971). Risk-taking behavior also was found to be uncorrelated with MFFT latency and error scores (Kopfstein, 1973; Shipman, 1971). All motor and behavioral variables except personality traits appear to be unrelated to either MFFT error or latency scores. There are three possible explanations for this finding. One explanation is that there is no correlation between the cognitive factors tested by the MFFT and actual behavioral measures. A second explanation is that the correlation between behavioral measures and MFFT scores is low because of a different task structure. This explanation agrees with the general trend of the literature that shows MFFT latency to be related to task structure variables. A third explanation is that the MFFT is scored so that latency is confounded with accuracy and that the MFFT is, therefore, not so simple or direct a measure of latency as are other latency indices. Since it is impossible to determine which of these explanations for the nongenerality of the MFFT to behavioral measures is the case, this area must be disregarded in the attempt to determine the variables related to MFFT latency and accuracy. 13 Present Study From the review of the literature, MFFT latency appears to be influenced by task structure variables that are unrelated to MFFT accuracy. Anxiety about performance, consistent feedback about task requirements, and modeling and verbal-training techniques all signifi- cantly increased MFFT latency scores, but not MFFT error scores. Thus it seems that MFFT latency is readily influenced by a range of task struc- ture variables. Since the correlation among latency indices depends on a similar task structure, the indication is that MFFT latency is not determined by a "predispostional" response tendency of the child as suggested by Kagan et a1. (1964). The child, instead, may be "pre- disposed" to a MFFT accuracy style. The literature review indicated that MFFT accuracy is related to personality attributes, to competence, and to optimism, and that such factors are independent of MFFT latency. The question raised in the present study is whether MFFT accuracy and latency scores of individual children are affected by different variables, as the literature review suggests. This possibility seems to have been overlooked because in its traditional conceptualization the MFFT was not considered to be a two-factor classification scheme. Instead the MFFT was conceptualized as a measure of two response styles, "reflectivity" and "impulsivity." That accuracy and latency factors were confounded in this conceptualization was disregarded until it was pointed out by Block et a1. (1974). Most studies that have defined MFFT response style by this traditional method have examined the effect of one variable on the MFFT response style. The present study examined whether variables that were found in previous research to affect MFFT latency and accuracy combine to determine the MFFT response classification of the individual child. 14 Since the MFFT has both a latency and an accuracy component, it seemed reasonable that experimental manipulations directly related to these components would be most influential on them. One attribute of task structure is the success criterion presented to the child, that is, whether accuracy alone or speed plus accuracy is stressed as important for good performance scores. Ward (1968) suggested that instructions about MFFT speed and accuracy could affect MFFT latency. Bush and Dweck's (1975) suggestion that "fast-inaccurate" children were incapable of discerning task requirement cues was supported by Block et al.'s (1974) finding that "accurate" children were more "perceptive" than "inaccurate" children. Thus in the present study, the child's percep-~ tion of the MFFT success criterion was manipulated by verbally labeling the criterion as either accuracy or speed plus accuracy. In accord with the results found in the literature review, this task structure manipu- lation was expected to influence MFFT latency scores, but not MFFT error scores. The following predictions were made about the MFFT success criterion manipulation: Prediction 1: Children in the speed plus accuracy criterion conditin will have significantly shorter MFFT posttest latencies to the first response than will children in the accuracy criterion condition. Prediction 2: The MFFT task success criterion will not significantly influence MFFT posttest error scores. Block et a1. (1974) showed competence, intelligence, and other personality attributes to be correlated with MFFT accuracy. Of these attributes, it seemed that competence could be most readily experimen- tally manipulated; therefore, perceived task competence was selected for study. On the basis of these results, manipulations of perceived 15 task competence were expected to affect MFFT error scores, but not MFFT latency scores. In the present study, the test administrator told children in the competence condition that their MFFT pretest performance was outstanding. Children in a control condition were told nothing about their MFFT pretest performance. Predictions about the effect of the competence manipulation were: Prediction 3: Children in the MFFT competence per- ception condition will make significantly fewer MFFT posttest errors than children in the control condition. Prediction 4: The MFFT competence perception of the child will not significantly affect his MFFT posttest latency to first response scores. An alternative measure of competence is the child's perception of his own competence in relation to that of other children his age. Prediction 5 is that there is a significant relationship between the child's own evaluation of his MFFT competence in relation to that of other children his age, based on an ordinal rating, and his MFFT posttest error SCOI‘BS . METHOD Subjects The subjects were one hundred forty—eight second- and third-grade boys, mean age eight years and two months, who were enrolled at the East Lansing Elementary Schools in East Lansing, Michigan. All were adminis- tered the initial MFFT. Of these boys, 59 were "fast-inaccurate," 47 were "slow-accurate," 32 were "fast-accurate," and 10 were "slow- inaccurate" on MFFT pretest classification. One hundred twenty-eight (70 second-grade, 58 third-grade) boys were assigned to experimental success criterion and competence perception groups on the basis of their performance on the initial administration of the MFFT. Of these boys, 48 were "fast-inaccurate," 38 were "slow-accurate," 32 were "fast- accurate," and 10 were "slow-inaccurate" on the MFFT pretest classifica- tion. Twenty additional boys from the original population, all classified as either "slow-accurate" or ”fast-inaccurate" on the MFFT pretest, constituted a test-retest reliability control group. Apparatus The test instrument was Kagan's (1965) Form F Elementary version of the MFFT, consisting of two practice items and twelve test items. Each test item consists of one standard familiar figure and six alternatives-- five similar to the standard figure and one identical to it. The child's task was to select the alternative that matched the standard. When the 16 17 subject selected an incorrect alternative, he was asked to choose again until the correct match was made. The maximum number of errors possible per test item was five, since each incorrect alternative was recorded as an error only the first time it was selected. The child's mean number of errors and the mean number of seconds (to the nearest half second) to the first response on each item, the standard measure of MFFT latency, are the two dependent measures of the MFFT. On the basis of these two dependent measures, the boys were classified as "slow-accurate" (SA; usually referred to as "reflective"), "fast-inaccurate" (Fl; usually referred to as "impulsive"), "fast- accurate" (FA), and "slow-inaccurate" (SI). The boys classified as ”slow" and "fast" were those who had scored respectively above and below the subject population mean latency to first response scores (20.15 seconds). "Accurate" and "inaccurate" classifications were based on a mean-split of the total error scores per test of the subject population (8.74 errors). Egeland and Weinberg (1976) have called this classifica- tion procedure the "composite standard score" and have suggested that it should replace Kagan's (1965) standard median-split procedure of subject classification in order to avoid the misclassification of subjects. Since the size of the present population sample is large--N=l48-- it was assumed that the MFFT scores of the sample were normally distributed rather than skewed, and that the mean was the more stable measure of central tendency on which to split subjects into MFFT pretest classification groups, as compared to the median. Even though the mean- split procedure was more likely to result in unequal pretest group sizes, precedence was given to this procedure because of the greater 18 statistical power of the mean and also for the purpose of establishing "normative" MFFT data. Since no reliable alternative form of the Elementary version of the MFFT exists (Messer, personal communication, May, 1976), the Kagan (1965) Form F Elementary version of the MFFT was used in both pretest and post- test sessions. Pearson product-moment correlation coefficients, based on the test-retest reliability control group (n=20), indicate that the test- retest reliability of the instrument for the present population was .83 for MFFT error scores and .74 for MFFT latency to first response scores. Procedure Each child was individually tested in two sessions seperated by approximately four weeks. The standard MFFT test procedure was used at the pretest session, providing the scores for MFFT pretest clasSifica-fi tion. Verbal comments about the boys' MFFT competence and the MFFT success criterion were the manipulations presented in the MFFT posttest session: In both sessions, the MFFT was administered by a young woman in a room free of distractions. The child sat at a table in a right-angle to the test administrator. The MFFT test booklet was placed on a stand in front of the subject so that the standard figure, on one page, and the six alternatives, on the facing page, were nearly at right angles to one another and clearly visible to the child. Since two test administrators were employed, care was taken so that the same woman administered the MFFT to a particular child in both sessions. Both the test administrators tested children in all the posttest conditions and were unaware of the MFFT pretest classification of the subjects. Interrater reliability for consistency in scoring was 19 .99 for MFFT error scores and .94 for MFFT latency to first response scores, as established by Pearson product-moment correlation coefficients. MFFT pretest session. In the MFFT pretest session, the child was told that the task involved looking at pictures and that the test admin- istrator would write down his score. In order to avoid distractions, the scoresheet was below the table surface, on the test administrator's lap. A quiet stopwatch was held beneath the table surface, away from the child's sight.' Once the child was seated the test administrator gave the child the standard MFFT instructions (see Appendix A). When the child selected an incorrect alternative, the test adminis- trator commented, "Good try. Find the one that is just like this one." A correct match received the comment ”good." At the end of the session, the child was not given any information about his pretest performance. He was told only that he would be doing a similar task again in a few weeks. MFFT posttest session. In the posttest session, the experimental manipulations were introduced. The manipulations consisted of verbal comments by the test administrator at the beginning of the session about the boy's MFFT competence and the MFFT task success criterion. The four experimental conditions were: competence and accuracy comments, compe— tence and speed plus accuracy comments, accuracy comments, and speed plus accuracy comments. The speed condition stressed speed plus accuracy so that the child would not believe speed to be the criterion to the exclu- sion of accuracy. The visibility of the scoresheet and the verbal comment about the dual criteridn.befdre each test item, described below, were precautions taken to assure that speed was not significantly 20 increased because the subject thought accuracy to be an irrelevant criterion. After the boys had been classified using MFFT pretest error and latency to first response scores, and grade-level, they were matched and assigned randomly to one of the four experimental groups with the constraint that a proportional number of subjects from each MFFT classi- fication were assigned to each group. The following verbal comments were given to the boys in each of the respective experimental conditions. Competence and Accuracy Comments I've come back to watch you do this again because you were so good on it the last time I was here. You got most of the pictures right. So this time I will keep your score on my scoresheet and I'd like you to try to find the correct matching picture each time. Competence and Speed Plus Accuracy Comments I've come back to watch you do this task again because you were so good on it the last time I was here. You chose the matching pictures really fast. So this time I brought a stopwatch and I'll keep your score on paper. You are to try to find the correct matching picture as fast as you can each time. Accuracy Comments The reason I've come back to watch you do this task again is because this time I want to keep your score on my scoresheet. You are to try to find the correct matching picture each time. Speed Plus Accuragy Comments The reason I've come back to watch you do this task again is because this time I brought a stopwatch and I want to keep your score on paper. You are to try to find the correct matching picture as fast as you can each time. Before each of the test items, the test administrator said, ”Try to find the correct matching picture as fast as you can" in the speed plus accuracy condition, or "Try to find the correct matching picture" in the 21 accuracy only-condition. In the accuracy conditions, the scoresheet was on the tabletop in front of the test administrator. In the speed plus accuracy conditions, the stopwatch and scoresheet were on the tabletop. The stopwatch could not be seen in the accuracy-only conditions. Test-retest reliability control group. The posttest session for the test-retest reliability group was identical to the standard MFFT proce-. dure used in the pretest session (see Appendix A). No experimental manipulations were introduced in the posttest session because the major purpose for including the group in the present study was to establish the test-retest reliability coefficient under standard administration con- ditions on the Form F Elementary version of the MFFT for the present population. The group also served as a control group in a statistical analysis involving change scores over test sessions. The test-retest reliability control group, and thus the analysis of change scores over sessions, was . limited to the SA and F1 MFFT pretest classification groups in order not to decrease the sparsely represented FA and SI groups that were included in the major analysis of the predictions. Postexperimental interview. Following the MFFT posttest, a brief postexperimental interview was conducted in order to determine the boys' perception of their MFFT competence and the experimental manipulations (see Appendix B). The child evaluated his MFFT posttest performance as "better than," "the same as," or "not as good as" that of his peers. These ranked data were correlated with MFFT error and latency scores and also with the experimental group to which the child had been assigned. Since it was possible that some children in the competence control condition may have interpreted the test administrator's lack of verbal 22 comment about the previous MFFT performance as a negative evaluation, care was taken to assure all children, at the end of the postexperi- mental interview, that their MFFT performance was adequate. RESULTS Analyses MFFT error and latency to first response scores were analyzed seperately. The major analysis consisted of a 2 x 2 x 4 least squares analysis of variance with Competence, Criterion, and Classification as the factors. Competence refers to whether the child was told he had performed well on the MFFT pretest or whether the child was in a control condition in which he was not given information about the MFFT pretest performance. The two levels of the Criterion factor were accuracy, and speed plus accuracy. The Classification factor had four levels of MFFT pretest classification: slow-accurate (SA), fast-inaccurate (Fl), fast- accurate (FA), and slow-inaccurate (81). To assess the magnitude and direction of MFFT performance change over test sessions, a 2 x 3 least squares repeated measures analysis of variance was performed with Classification and Criterion as the factors. The Classification factor consisted of SA and F1 prettest classification groups. In addition to accuracy, and speed plus accuracy criterion groups, a third group, the test-retest reliability control group, was included in the analysis of the Criterion factor in order to examine whether subjects exposed to experimental manipulations differed signifi- cantly from those not exposed to experimental manipulations. Since the means were based on unequal cell frequencies, all post hoc comparisons among the means were analyzed with the Scheffe test. The .05 level of 23 24 statistical significance was selected for all statistical analyses in the present study. Latency Scores The mean MFFT posttest seconds to first response, standard deviaei tions, and the number of observations for the 32 experimental conditions of the 2 x 2 x 4 factorial design analyzed by least squares analysis of variance are presented in Table l. Unequal cell frequencies were obtained because of the mean—split procedure by which speed and accuracy groups were defined on the pretest. In particular, relatively few children were classified as SI. Since unequal observations were caused by the mean-split procedure, a least squares analysis was performed accordingly (Kirk, 1968, p. 204). The least squares analysis adjusts the error sum of squares as small as possible so that a true test of the effects is obtained. A summary of the 2 x 2 x 4 least squares analysis of variance is presented in Table 2. Highly significant main effects for Criterion (5(1,112) = 66.81, 2 ‘E001) and for Classification (f(3,112) = 25.63, p_‘€001) were obtained. Neither the main effect of Competence nor any of the interactions reached statistical significance. The main effect of Criterion supported Prediction 1, since the speed plus accuracy group showed a shorter mean latency to first response (i = 9.78 seconds) than did the accuracy group (i = 19.24 seconds). Since the pretest mean latency for the two groups is 20.4 seconds for the speed plus accuracy group and 20.6 seconds for the accuracy group, respectively, it appears that children who were told that speed plus accuracy was the MFFT success criterion decreased in MFFT seconds to the first response of a test item compared to the children in the accuracy 25 Table l. MFFT Posttest Latency Means and Standard Deviations for Competence, Success Criterion, and Pretest Classification Groups (2 x 2 x 4 ANOVA) Experimental Pretest Classification Success Criterion Condition SA FA SI FI x 15.07 7.84 9.18 5.96 Speed 8 SD. 9.52 3.25 5.53 2.65 - Speed 8 Accuracy Accuracy n 9 8 3 12 Competence i ' 29.46 15.51 15.13 15.50, i 9.77 Accuracy .§Q 9.60 3.75 11.38 5.94 SD_ 7.07 n 10 8 2 11 n 65 . X 16.10 10.76 11.41 5.21 Speed 8 SD. 9.94 3.96 6.97 1.74. Accuracy Accuracy n 10 7 3 13 Control x 26.27 16.11 31.15 12.42 x 19.24 Accuracy SD_ 7.10 6.30 14.83 6.00 §Q_ 9.38 n 9 9 2 12 n 63 Classification 2 21.78 12.72 15.43 9.56 SD 10.84 5.59 11.39 6.10 n 38 32 IO 48 Note. Data are based on 128 experimental subjects. 26 Table 2. Summary of 2 x 2 x 4 Least Squares Analysis of Variance of ' MFFT Posttest Latency (Seconds to First Response) Scores ‘- Source 88 d_f_ MS 5 Competence (A) 1.67 l 1.67 .04 Criterion (B) 2863.20 1 2863.20 66.81** Classification (C) 3294.74 3 1098.25 25.63** A x B 26.63 1 26.63 .62 A x C 220.24 3 73.41 1.71 B x C 186.00 3 62.00 1.45 A x B x C 160.80 3 53.60 '1.25 Within 4799.77 112 42.86 ------ **g <. 001. 27 condition who were unaware that speed was a relevant criterion of success. Scheffe post hoc pairwise comparisons of Classification group latency scores revealed tht the SA group differed significantly from the FA and F1 groups in posttest latency (see Table 3). The SA group was significantly slower than the other pretest classification groups (R = 21.78; see Table 1). Since the MFFT competence perception of the child did not signifi- cantly affect MFFT latency to first response, the null hypothesis must be accepted with respect to Prediction 4. In accord with Prediction 4, a nonsignificant correlation (r = .05, ns) was found between MFFT posttest seconds to first response on an ordinal rating in which the child compared his MFFT performance to that of his peers. Table 3. Mean Posttest Latency Differences-(in Seconds) for Scheffe Post Hoc Pairwise Comparisons of Classification Levels (2 x 2 x 4 ANOVA) Latency SA FA 81 F1 SA ------- 9.06* 6.35 12.22* FA -------------- 2.71 3.16 81 -------------- , ------- 5.87 FI —-----— --------------------- *p_< .05. 28 It was also of interest to determine the effect of experimental factors on the magnitude and direction of MFFT performance change over test sessions. Of particular interest was whether the experimental groups differed significantly from the test~retest control group. Since SA and PI were the only classification groups assigned to the test-retest condition, these were the only MFFT pretest classification groups included in the 2 x 3 (Classification x Criterion) least squares analysis of variance. (Note that SA and F1 experimental groups in the original 2 x 2 x 4 analysis, however, constituted 67.3 percent of the population; therefore, the samples of the 2 x 2 x 4 and 2 x 3 analyses were fairly similar). The analysis of main effects of the 2 x 3 least squares repeated measures analysis of variance was based on pre-posttest composite scores, and the effects across sessions were based on pre—posttest differences. MFFT pretest and posttest latency to first response means and standard deviations for Classification and Criterion factors are presented in Table 4. A summary of the 2 x 3 least squares repeated measures analysis of variance is presented in Table 5. The analysis of main effects, based on composite pretest-posttest means, revealed highly significant effects of Classification (§(l,lOO) = 196.01, p_<.001) and of Criterion (§(2,100) = 9.01, p_<.001).' The results of the analysis also demonstrated highly significant effects of Measures (E(l,100) = 35.77, p <.001), Classification x Measures (F(l,lOO) = 32.48, p_<.001), and Criterion x Measures (F(2,100) = 28.52, p_<.001). The Classification factor showed the same trend as in the 2 x 2 x 4 analySis. .The F1 group had a shorter composite latency (R = 21.53) than 29 Table 4. Latency to First Response Score Means (in Seconds) and Standarereviations for Measures, Success Criterion, and Pretest Classification Groups (2 x 3 ANOVA) Pretest Classification SuCCess Criterion Speed 8 Accuracy Control Accuracy 2 32.53 31.95 32.22 32.24 SA §p_ 12.31 8.84 6.37. n 19 19 9 Pretest X 11.17 11.26 12.37 . 11.43 Fl §p_ 3.91 3.46 3.37 25 23 11 ’ 20.39 20.62 21.30 x 15.61 27.95 29.93 23.34 SA §p_ 9.48 8.43 9.27 n 19 19 9 Posttest i 5.57 13.89 12.49 10.10 Fl §p_ 2.21 6.04 5.98 n 25'- 23 11 R 9:91 20.25 20.34 Criterion x .15.15 20.44 20.82 §p_ 12.29 9.45 11.18 n , 44 42 20 30 Table 5. Summary of 2 x 3 Least Squares Repeated Measures Analysis of Variance for MFFT Seconds to First Response Scores **R<~001- Source S_S_ if M_S_ 5 Between subjects Classification (A) 29954.39 1 29954.39 l96.01** Criterion (B) 2753.72 2 1376.86 9.01** A x B 131.22 2 65.61 .43 Subj. w. groups 15282.00 100 152.82 ------- Within subjects Measures (C) 1737.64 1 1737.64 35.77** A x C 1578.02 1 1578.02 32.48** B x C 2770.60 2 1385.30 28.52** A x B x C 292.66 2 146.33 3.01* Subj. w. measures x . 4858.00 100 48.58 ------- groups *2 < . 0537 . 31. the SA group (X: 55.58). Scheffe post hoc pairwise comparisons revealed that the speed plus accuracy criterion group mean (15.15 seconds) differed significantly from the accuracy criterion group mean and the test—retest control group mean (2 = 20.14 and 20.82 seconds, respec- tively). The accuracy and test-retest control groups, however, did not differ significantly. Subjects who were told, with emphasis, that it was important to select the correct matching picture did not spend more time selecting a correct match than did subjects who were given a brief description of the MFFT objective on correct matching. Posttest latency (R = 10.38) was less than pretest latency (i = 20.65); overall, children responded faster at the second MFFT session than the first. Possibly this is an indication of a practice effect. However, since the Measures factor significantly interacted with the Classification factor and also with the Criterion factor, such a possibility must be qualified. With regard to the Classification x Measures interaction, Scheffe post hoc comparisons indicated that SA pretest latency scores differed significantly from the FI posttest latency scores and that the FI pretest latency scores differed significantly from the SA posttest scores. The difference between the means for the two comparisons were 22.14 seconds and 11.91 seconds, respectively. It was found that the SA pre- and posttest means were involved in all the significant pairwise comparisons. Figure 1 shows that the SA and F1 groups differed substantially on MFFT pretest classification, but that the SA group showed a significant decrease in posttest latency, whereas the FI group did not. It seems that "fast" subjects could not improve on response time because of a 'ceiling effect.‘ This possibility is supported by the pre-posttest 32 trends for FA and 81 groups, depicted in Figure 1 (though not included in the 2 x 3 analysis), that are similar to F1 and SA groups, respectively. Figure 2 shows that the speed plus accuracy group differed signifi- cantly on MFFT posttest latency to first response compared to the accuracy and test-retest groups. The mean latency differences of the Criterion x Measures groups of the Scheffe pairwise post hoc comparisons that reached statistical significance involved the speed plus accuracy posttest condition. The mean latency difference between the accuracy group pretest and the speed plus accuracy group posttest was 10.17 seconds and the mean latency difference between the test-retest group pretest and the speed plus accuracy group posttest was 11.39 seconds. Thus those children who were told that speed was a criterion modified their MFFT posttest latency accordingly. Error Scores The mean total error scores per test were analyzed in the same way as the mean MFFT latency to first response scores. A summary of the Competence x Criterion x Classification group means, standard deviations, and frequencies of the groups appears in Table 6. As is shown in Table 7, the 2 x 2 x 4 least squares analysis of variance revealed highly significant effects for Criterion (§(l,llZ) = 30.75, p3:.001) and for Classification (5(3,112) = 13.72, p <.001). Prediction 2 was that MFFT errors would not be affected by the child's- perception of the task success criterion. Children in the speed plus accuracy criterion condition, however, made more errors (X = 10.43) than did those in the accuracy condition (R = 5.86), The children in the speed plus accuracy criterion condition either focused their attention on the speed criterion alone, or they may have been unable to maintain a low 33 ‘\ 34-~ ""."""‘ Included irI t \ 2 x 3 ANOVA 324’ “ "“" Excluded 30’ from 2 x 3 ANOVA db \\ 28‘. 26v q. 24” \ Slow- ‘I \‘ accurate :2qu \‘ \ e “ x .4: 20" x \ (I \\ m \ 2 18+ \ O \ U " \ 3 16 ' ‘~ g h ‘ Slow- o --_. inaccurate z 14“ ‘-~~"‘-~~- 'I --“‘~ Fast- 12J’ accurate 4' \ 10" Fast- 1' inaccurate 8" 6‘t GI» 41' 2‘! Pretest Posttest Figure 1. Mean MFFT Latency Scores Per Item for Pre- and Posttest Sessions as a Function of Classification Group 34 341 32 -' 30 <- 28 d. 24 0 22 " Test- " ___ retest 20 Accuracy 18 1b 16 1’ Mean Seconds/Item 14 ~» 12 1b 10 ‘P Speed 8 .. Accuracy “ I l Pretest Posttest Figure 2. Mean MFFT Latency Scores Per Item for Pre- and Posttest Sessions as a Function of Criterion Group 35 Table 6. Mean Total Posttest Error Scores Per Test, Standard Deviations, and Frequencies of Competence, Success Criterion, and Pretest Classification Groups (2 x 2 x 4 ANOVA) Experimental Pretest Classification Success Criterion Condition SA FA 81 PT i 7.11 9.00 12.33 13.92 Speed 6 SD_ 4.48 7.76 3.51 4.68 Speed 8 Accuracy ACCuracy n 9 8 3 12 Competence x 5.00 6.13 10.00 7.55 x 10.43 Accuracy .52 2.98 3.91 4.24 4.82 SD_ 6.05 n 10 8 2 11 n 65 i 7.10 7.29 11.33 114.00 Speed & SD_ 7.58 1.70 2.52 5.20 Accuracy Accuracy n 10 7 3 13 Control 1 1.67 4.00 4.00 9.00 x 5.86 Accuracy SD_ 1.32 3.32 1.41 4.00 SD_ 4.22 n 9 9 2 12 n 63 Classification _ - X 5.26 6.50 9.90 11.25 SD_ 5.07 4.91 4.09 5.40 n 38 32 10 48 Note 1. Summary data are based on 128 experimental subjects. Note 2. Sixty was the maximum number of errors possible per subject. 36 Table 7. Summary of 2 x 2 x 4 Least Squares Analysis of Variance of; MFFT Posttest Total Errors Source §_S_ fl M_S_ _F_ Competence (A) 22.25 1 22.25. 1.02 Criterion (B) 669.21 1 669.21 30.75** Classification (C) 895.60 3 298.53 13.72** A x B 8.16 1 8.16 .38 A x C 56.30 3 18.77-- .86 B x C 36.27 3 12.09 .56 A x 8 x c 39.85 3 13.28 .61 Within 2437.07 112 21.76 ------ E_<.001. 37 60*5 16‘I 15. 144' Speed 8 11" Accuracy 10¢» 945 Mean Total Errors/Test Accuracy Test- 5.. retest I I - Pretest Posttest Figure 3. Mean Total Errors Per Test for Pre- and Posttest Sessions as a Function of Criterion Group 38 error rate because of anxiety or a speed-accuracy tradeoff (see Figure 3). Scheffe pairwise post hoc contrasts of Classification group means (see Table 8) revealed that the FI group differed significantly from SA and FA groups. The FI group made more errors (R= 11.25) than did the SA and FA groups. Also significant was the comparison in which "accurate" pretest classification groups were combined and compared to combined "inaccurate" pretest classification groups; the comparison of combined "slow" and combined "fast" groups, however, did not differ significantly. MFFT pretest accuracy groups remained significantly different after the introduction of experimental manipulations. Thus MFFT error scores appear to be stable, that is, they were not affected by the experimental manipulations in the present study. Table 8. Mean Posttest Error Score Differences for Scheffe Post Hoc Pairwise Comparisons of Classification Levels (2 x 2 x 4 ANOVA) Errors SA FA SI FI SA ------- 1.24 4.64 5.99* FA -------------- 3.40 4.75* 51 --------------------- 1.35 Fl ---------------------------- *p_<.05. 39 Since the main effect for Competence was not significant, Prediction 3 was not supported. Children told that they had performed well on the MFFT pretest did not make fewer errors than those children told nothing about their MFFT pretest performance.‘ The possibility that the competence message was ineffective because children did not believe the competence comments of the adult was explored. Of the 50 children who received the competence message, 41 said they believed the adult, six said they did not believe the adult, and three said that they did not remember the competence message. In view of the small number of children who did not believe the competence message, further analyses of these data were not warranted. The relationship between the child's evalua- tion of his MFFT competence in relation to that of his peers and MFFT posttest error scores was low (r = .11, ns). On the ordinal ranking, 57 percent of the children rated their MFFT performance as the same as that of their peers, 28 percent as better than that of their peers, 12 percent as not as good as that of their peers, and three percent refused to judge. Freeman's theta, a coefficient of differentiation between ordinal and nominal categories, was used to examine the relationship between the child's evaluation of his competence and assignment to experimental competence and criterion groups. Theta was low (0 = .09, ns), meaning that there was no relationship between the child's evaluation of his competence and the nominal competence-criterion group to which he was assigned. Neither the experimental manipulations that consisted of verbal comments about MFFT competence nor the child's evaluation of his MFFT competence in relation to that of his peers was related to MFFT scores . 40 The means, standard deviations, and frequencies for MFFT pretest and posttest error scores of the Classification and Criterion groups of the 2 x 3 least squares repeated measures analysis of variance are presented in Table 9. As Table 10 shows, significant effects were found for Classification (5(1,100) = 171.46, pf .001) and for Criterion (F(2,100) = 7.88, p_