Validity of Questionnaire and TAT Measures of Need for Achievement: Two Meta-Analyses

+ The Unconscious Mind

+ Psychology

Author

William D. Spangler

Psychological Bulletin. 1992. Vol. 112, No. 1,140-154

Validity of Questionnaire and TAT Measures of Need for Achievement: Two Meta-Analyses

William D. Spangler School of Management State University of New at Binghamton

Correspondence concerning this article should be addressed to William D. Spangler, School of Management, State University of New York, P.O. Box 6000, Binghamton, New York 13902-6000.

Proponents of the Thematic Apperception Test (TAT), most notably McClelland, have argued that the TAT and questionnaires are valid measures of different aspects of achievement motivation. Critics of the TAT have argued that questionnaires but not the TAT are valid measures of the need for achievement. Two meta-analyses of 105 randomly selected empirical research articles found that correlations between TAT measures of need for achievement and outcomes were on average positive; that these correlations were particularly large for outcomes such as career success measured in the presence of intrinsic, or task-related, achievement incentives; that questionnaire measures of need for achievement were also positively correlated with outcomes, particularly in the presence of external or social achievement incentives; and that on average TAT-based correlations were larger than questionnaire-based correlations. The theoretical implications of these findings are discussed.

Over the course of 4 decades, McClelland, Atkinson and their associates have studied the motivational bases of human behavior. Much of their work has focused on the sources and effects of achievement motivation. This work has ranged from laboratory studies of the effects of need for achievement on performance (Atkinson & Litwin, 1960), studies of performance and success of people such as entrepreneurs in vocational settings (McClelland & Winter, 1969), training efforts aimed to increase the need for achievement of individuals (McClelland, 1965), as well as studies linking the achievement motive to the economic growth and decline of civilizations (McClelland, 1961). During this period a number of theories of motivation have been developed (e.g., Atkinson, 1957; McClelland, 1985).

At the same time, McClelland, Atkinson, and their colleagues have devoted much research to the issue of measuring the need for achievement in individuals. The focus of this work has been the Thematic Apperception Test (TAT; Atkinson, 1982; McClelland, 1972, 1980, 1985; McClelland, Atkinson, Clark, & Lowell, 1958; McClelland, Clark, Roby, & Atkinson, 1958). TAT presents the subject with a set of pictures, general instructions to be creative, and a set of four questions to guide the respondent in writing stories. The respondent writes a short story interpreting each picture, and the stories are then coded for the presence of various types of achievement imagery.

The TAT method of measuring the achievement motive has inspired substantial criticism as well as defense. Critics have charged that TAT measures of the achievement motive demonstrate poor test-retest and internal consistency reliability (Entwisle, 1972; Fineman, 1977; Weinstein, 1969) and have low and inconsistent correlations with actual achievement-oriented be-

havior (Entwisle, 1972; Fineman, 1977; Klinger, 1966; Scott & Johnson, 1972; Weinstein, 1969). Critics have further argued that questionnaire measures demonstrate adequate reliability and greater predictive validity than TAT measures of the achievement motive (Fineman, 1977; Mischel, 1972; Scott & Johnson, 1972). Finally, critics have pointed out that questionnaire and TAT measures of achievement are virtually uncorrelated (Fineman, 1977; Weinstein, 1969), providing evidence of the poor convergent validity (Campbell & Fiske, 1959) of the TAT measure.

McClelland, Atkinson and their associates have responded to these criticisms (Atkinson, 1982; deCharms, Morrison, Reitman, & McClelland, 1955; McClelland, 1972,1980,1985). In particular, McClelland and his associates have argued that when the TAT is properly administered, derived motive scores have adequate test-retest reliability (McClelland, 1980; McClelland, Koestner, & Weinberger, 1989; Winter & Stewart, 1977). They have further asserted that the TAT predicts longterm “real-world” behavior better than do questionnaire measures, and McClelland has argued that TAT and questionnaire measures of the achievement motive are virtually uncorrelated because they are measures of distinct aspects of personality and therefore should not be correlated.

This dispute over the relative merits of TAT versus those of questionnaire measures of achievement motivation has been valuable in the sense that the published criticisms and defenses of TAT and questionnaire measures provide the basis for precise testable hypotheses. However, previous reviewers of the literature have faced a number of difficulties in testing hypotheses. First, thousands of empirical studies on achievement motivation are available, and it is virtually impossible to analyze more than a small fraction of them. It is not clear that the studies chosen for analysis from the mass of available research were randomly selected. Second, there have been no statistical tests of hypotheses. Third, empirical research on achievement motivation has used many statistics in addition to correlations coefficients such as chi-squared statistics and contingency tables, F and t tests, p values, and tables of means and standard deviations, but prior reviews have not made use of these statistics. Fourth, studies vary in sample size, and previous reviews have not weighted their statistics by sample size.

Over the last decade, a set of analytical and statistical techniques has been developed that allow researchers to address these four difficulties. These techniques collectively are referred to as meta-analytic techniques (Hedges & Olkin, 1985; Hunter, Schmidt, & Jackson, 1982; Rosenthal, 1984; Wolf, 1986). Using these newly developed procedures, it is possible to systematically analyze a large body of published research or a randomly selected subset of it, statistically test hypotheses, convert various measures of association to a common metric such as correlations (Rosenthal, 1984; Wolf, 1986), and weight individual statistics by their respective sample sizes.

The present investigation’s purpose is to test the competing claims of McClelland and colleagues against those of their critics by using newly developed meta-analytic procedures to analyze previously published empirical research.

Theoretical Background

To understand the nature of the controversy between advocates of TAT measurement and proponents of questionnaire measurement of motives, it is necessary to first discuss some theoretical concepts advanced by McClelland and his colleagues.

McClelland has argued that the dispositions measured by the TAT are implicit motives. Implicit motives are dispositions that have traditionally been labeled needs, for example, need for achievement (nAch), need for power (nPower), and need for affiliation (nAff). McClelland considers these TAT-measured dispositions to be motives because they “drive behavior (i.e., energize it), direct behavior (i.e., focus attention on relevant activity), and select behavior (i.e., produce better learning or performance)” (McClelland et al., 1989, p. 696). These motives are labeled implicit motives because the person writing stories to a set of pictures is not explicitly describing him- or herself. Rather, stories written reflect nonconscious motives of the author. McClelland et al. (1989) speculate that these implicit motives are built on affective experiences with natural incentives early in life before the development of language in the person. There is some evidence that implicit motives drive, direct, and select behavior through physiologically based reinforcement processes. For example, McClelland et al. (1989) discuss a study in which the presentation of a romantic film was associated with the increased release of dopamine for high nAff subjects.

McClelland has defined a second category of personality dispositions, which are currently called self-attributed motives, such as self-attributed achievement (sanAch), self-attributed power (sanPower), and self-attributed affiliation (sanAff). Previously, these self-attributed motives had been referred to as values, but this terminology evidently led to some confusion with values defined as “normative beliefs about desirable goals and modes of conduct” (McClelland et al., 1989, p. 690). These dispositions represent the value or worth to individuals of specific achievement-, affiliation-, or power-related activities.

According to McClelland and his associates, these self-attrib-

uted motives differ from implicit motives in at least four fundamental ways. First, they are uncorrelated with measures of implicit motives. Second, they correlate with different types of outcomes. Implicit motives tend to predict long-term spontaneous behavioral trends over time such as entrepreneurial success (nAch) or success in management (nPower), whereas self-attributed motives predict responses to immediate and specific situations and choice behavior. Third, self-attributed motives are relatively conscious perceptions of what is important to the individual and of what is valued by the individual’s culture. These self-attributed motives are part of the individual’s self-concept. Fourth, self-attributed and implicit motives have different developmental histories. Implicit motives develop early in life as a result of experiences with various incentives and do not require the presence of language for their development. Self-attributed motives develop somewhat later in life, require the presence of language, and come from the individual’s understanding of social incentives and demands made verbally by others in the environment. As a result of these distinct developmental histories, implicit motives are related to physiological processes such as release of norepinephrine and dopamine; self-attributed motives evidently are not related to such physiological processes.

From the perspective of McClelland (McClelland, 1980, 1985; McClelland et al., 1989), individual differences in implicit and self-attributed motives do not by themselves predict individual differences in behavior. Motives predict behavior only in the presence of appropriate incentives. If there are no achievement incentives in a given research or work situation, there is no reason to believe that achievement-oriented individuals will behave any differently than those low in achievement motive. That is, successful performance of some specified activity will not result in outcomes that are reinforcing to the performer. In achievement situations, on the other hand, there should be a positive correlation between strength of the achievement motive and achievement-related behavior because the person’s activity may lead to reinforcing outcomes. In other words, achievement behavior is an interactive effect of implicit and self-attributed motives for achievement and environmental achievement incentives.

Recently, McClelland et al. (1989) made a distinction between social incentives and activity incentives. Social incentives are characteristics of situations such as rewards, prompts, expectations, demands, and norms that come from outside the task itself. These incentives may be provided by a boss, by an experimenter, by co-workers, or by a group. Social achievement incentives include challenging goals set by an experimenter (McClelland, Atkinson, Clark, & Lowell, 1958), achievementoriented instructions in an experiment (French, 1955; McClelland, Clark, Roby, & Atkinson, 1958), achievement work norms, and pretreatment experimental manipulations. Activity incentives are characteristics of the task itself. The individual high in some implicit motive is reinforced by performing the task itself. Activity achievement incentives include moderate task risk (Atkinson, 1957; Atkinson & Feather, 1966; Atkinson & Litwin, 1960; Weinstein, 1969), task contingency (Raynor, 1969, 1970), achievement work content (McClelland, Atkinson, Clark, & Lowell, 1958), time pressure, and a high objective relationship between performance and some achievement-related outcome in the immediate situation.

The relevance to the present investigation of this distinction between social and activity incentives is the proposition made by McClelland et al. (1989) that specific classes of incentives interact with specific types of motives. Specifically, social incentives interact with self-attributed motives but not implicit motives, and activity incentives interact with implicit motives but not self-attributed motives. A number of studies have shown that social incentives but not activity incentives interact with self-attributed motives to predict behavior (deCharms et al., 1955; McClelland et al., 1989; Patten & White, 1977). Theoretically, if self-attributed motives arise as a result of exposure to demands, goals, incentives and values expressed by others, then it follows that these self-attributed motives should, later in life, respond to external or social incentives. Activity incentives but not social incentives will interact with implicit motives to produce behavior. The theoretical reason for this matching of implicit motives with activity incentives is the possibility that implicit motives are based on incentives involved in doing or experiencing certain things early in life. McClelland et al. (1989) have summarized a number of studies in which TAT-based implicit motives and activity incentives, but not social incentives, together predicted behavior.

To understand the controversy between proponents and opponents of TAT measurement, it is also necessary to examine McClelland’s use of the terms operant and respondent. Operant behavior is behavior that the subject generates spontaneously. At least it is not possible to specify or control the stimulus that elicits the behavior, and such behavior appears to be freely emitted. Respondent behavior is behavior that is controlled by characteristics of the subject’s environment. That is, it is behavior elicited by known stimuli in the environment. Behavior is more seriously and overtly constrained by the environment in respondent situations than in operant situations. Of course, behavior is not either entirely operant or entirely respondent. Some degree of environmental control is assumed even in supposedly operant situations, and even in respondent situations the subject has some degree of freedom in responding.

¹According to McClelland (1980), it is possible to arrange behavior along a continuum from behavior emitted under extreme environmental control to behavior that is relatively free from explicit environmental control. Relatively operant outcomes include income, job level attained in an organization, professional rank, publications, participation and leadership in community organizations, and social behavior occurring under natural conditions. These are examples of operant behavior because it is not possible to specify the exact stimuli that lead to the behavior in question. Respondent outcomes include school grades, intelligence and achievement test scores, results of personality inventories, and opinion surveys. According to McClelland (1980), these are examples of respondent behavior because the expression of the behavior is elicited and constrained by environmental stimuli. For example, a class exam may specify the stimulus items, response options, instructions, time allotted, and interactions with other students and the examiner. In an opinion survey, respondents may be limited to agreeing or disagreeing with pre-selected items. Behavior or performance typically measured in laboratory experiments falls somewhere between these extremes and may be labeled semioperant behavior (McClelland, 1972; McClelland et al., 1989).

With this theoretical background in mind, it is possible to derive the seven hypotheses that were tested in the present investigation.

Seven Hypotheses

Critics of TAT measurement have argued that questionnaires predict outcomes better than does the TAT. Fineman (1977) summarized correlations and other measures of association between questionnaire measures of need for achievement and criteria such as school grades, job success, and laboratory performance. Twenty-two of 30 measures were significant, and Fineman noted that the criterion validity of these measures was better than the criterion validity of the TAT measure nAch. Scott and Johnson (1972) directly compared the predictive validity of TAT measures of power, achievement, and affiliation with questionnaire measures and found correlations of questionnaire measures with criteria to be higher than correlations of TAT measures with outcomes. Likewise, Entwisle (1972) found that correlations between questionnaire measures and outcomes were generally higher than correlations between TAT measures and outcomes.

McClelland has argued that in general it is inappropriate to compare the predictability of TAT and questionnaire measures of the achievement motive regardless of the type of outcome predicted. The TAT generally is a better predictor of long-term operant behavioral trends, and questionnaires tend to be better predictors of choices and attitudes. However, McClelland (1972,1980) has agreed with critics of TAT measurement that under many circumstances questionnaire measures of the achievement motive will generate superior correlations. These circumstances include situations in which questionnaires and behavior are assessed within a short time of each other, situations in which respondents infer their need for achievement from their perceptions of their behavior, and occasions in which the questionnaire and the behavioral measures share items.

These considerations led to the first hypothesis tested in the present research:

Hypothesis 1. The motive-outcome correlation is the result of a type-of-motive-measure main effect. That is, the correlation between TAT measures of the achievement motive and outcomes will be smaller than the correlation between questionnaire measures of achievement and outcomes.

McClelland (1980; McClelland et al., 1989) has argued that the magnitude of the correlation between a measure of the achievement motive and an outcome depends both on the type of motive measure and the number of achievement incentives in the situation. Motives per se do not predict behavior, activity incentives interact with implicit motives, and the TAT measures implicit motives. Therefore, the correlation between a TAT measure of nAch and outcomes in a particular situation depends on the number of activity achievement incentives in the situation. Furthermore, questionnaires measure self-attributed motives and interact with social incentives rather than activity incentives. Therefore, the correlations between questionnaire measures of self-attributed achievement and outcomes should not vary substantially with variations in the number of activity incentives in the situation. From these arguments Hypothesis 2 was derived:

Hypothesis 2. The motive-outcome correlation is the result of a Type of Motive Measure X Number of Activity Incentives interaction. Specifically, the effect of number of activity incentives on the magnitude of the motive-outcome correlation will be greater for TAT measures of the achievement motive as compared with questionnaire measures of the achievement motive.

According to McClelland et al. (1989), questionnaires measure the self-attributed need for achievement, now labeled san-Ach. SanAch interacts with social incentives in the environment, and therefore, questionnaire-based motive-outcome correlations will increase with the number of social achievement incentives in the situation. Furthermore, according to the arguments of McClelland et al., the number of social incentives in the experimental or work situation should not substantially affect the predictability of TAT measures of achievement because the TAT measures nAch, which interacts primarily with activity incentives in the environment.

Hypothesis 3. The motive-outcome correlation is the result of a Type of Motive Measure X Number of Social Incentives interaction. The effect of number of social incentives on the magnitude of the motive-outcome correlation will be less for TAT measures of the achievement motive as compared with questionnaire measures of the achievement motive.

Critics of TAT measurement argue that in general questionnaires demonstrate superior predictability. According to McClelland (1980) and McClelland et al. (1989), the superiority of questionnaires versus the TAT depends on the type of outcome measured, specifically on the operant or respondent character of the outcomes.

The average correlation between TAT achievement and operant, or real-life, outcomes should be larger than the average correlation between TAT measures and respondent outcomes (McClelland, 1980,1985; McClelland et al., 1989). McClelland et al. (1989, p. 695) provide one explanation for this expected relationship. Social incentives found in the environment interact with the self-attributed need for achievement (sanAch); achievement incentives in the activity itself interact with the implicit motive for achievement (nAch). McClelland et al. (1989) suggest that social incentives are likely to be associated with respondent outcomes, whereas activity incentives are likely to be associated with operant outcomes. Therefore, the expected correlation between TAT achievement and operant outcomes will be positive because of the interaction between activity achievement incentives and need for achievement.

McClelland et al. (1989) imply, however, that correlations between TAT achievement and operant outcomes should be larger than correlations between TAT achievement and respondent outcomes for a second reason. In the case of TAT measurement and operant outcomes, environmental stimuli do not strongly control either TAT responses or operant outcome responses, so it is possible for variations in TAT implicit need for achievement to produce variations in behavioral responses. On the other hand, in the case of TAT measurement and respondent outcomes, environmental stimuli will elicit and constrain respondent behavior but not TAT responses, thereby reducing any association between TAT implicit need for achievement and respondent outcomes.

McClelland (1980, 1985) and McClelland et al. (1989) have contended that questionnaire measures may predict respondent behavior such as school grades or attitudes on pencil-andpaper inventories better than operant behavior such as career success. Three reasons account for this superiority of questionnaire measures in predicting respondent outcomes versus operant outcomes.

First, respondent outcomes are very likely associated with social incentives, and according to McClelland et al. (1989), social incentives in the environment interact with the self-attributed need for achievement measured by questionnaires to produce achievement-related respondent behavior. On the other hand, operant achievement behavior arises from the interaction of nAch interacting with activity incentives. In respondent situations with questionnaire measures, there are quite likely few activity incentives, and questionnaires measure self-attributed need for achievement, so operant behavior will not be well predicted by questionnaire measures of achievement.

Second, questionnaire measures of achievement motivation are respondent measures, that is, the responses are elicited and shaped by the instrument and situation. Likewise, respondent behavior is behavior elicited and shaped by stimuli. If the stimuli eliciting responses to the instrument are the same as those eliciting the respondent behavior, then there should be a strong correlation between the two sets of measures. In the extreme case, a questionnaire measure of achievement and achievement behavior may be two measures of the same thing (McClelland, 1972). For example, a questionnaire measure of achievement motivation relies on a subject’s conscious opinions about himor herself. This questionnaire is “validated” by friends’ or teachers’ observations of the respondent, but the observers merely report what they have heard the respondent tell of himself or herself. McClelland (1980) further suggested that observed correlations between questionnaire measures of achievement motivation and outcomes may be inflated by such contaminants as single-source response bias.

Third, according to McClelland (1980,1985), questionnaires measure the conscious value people place on achievement, a part of the self-concept, whereas the TAT measures the underlying and nonconscious motive for achievement, nAch. Therefore, questionnaires predict those aspects of behavior that are linked to conscious values and processes such as choice and susceptibility to expert opinion (deCharms et al., 1955), responses to formal instruments such as personality inventories, and performance on exams. However, according to McClelland (1980,1985), much of operant real-life behavior is determined by nonconscious motives that drive, direct, and select behavior.

From this reasoning, Hypothesis 4 was derived and tested in the present study.

Hypothesis 4. The motive-outcome correlation depends on a Type of Motive Measure X Outcome Operant Level interaction. The degree of increase in the motive-outcome correlation as a function of type of outcome (going from respondent to operant outcomes) will be greater for TAT measures of the

achievement motive as compared with questionnaire measures of the achievement motive.

McClelland and his associates have explored 2 two-way interactions between type of motive measure and number of achievement activity incentives (Hypothesis 2) and between type of motive measure and outcome operant level (Hypothesis 4). However, McClelland and his associates have not explored the possible three-way interaction of type of motive measure, number of achievement activity incentives in the environment, and outcome operant level. If there are no achievement incentives in the situations that generate either operant or respondent outcomes, then high need-for-achievement individuals would not be motivated to perform well in either situation and the average correlation for operant outcomes should not be much larger than the average correlation for respondent outcomes. However, in situations characterized by many achievement incentives, operant outcome measures should produce substantially higher correlations with TAT measures of need for achievement than respondent outcome measures. High-need-for-achievement persons performing respondent tasks may find themselves constrained by the nature of the task or situation or not reinforced by successful completion of the task, in which case the presence of incentives may have a small effect on behavior. Motive-outcome correlations that are based on questionnaire measures will not be substantially affected by number of activity incentives in the environment in either operant or respondent situations because activity incentives do not interact substantially with self-attributed motives measured by questionnaires.

Hypothesis 5. The motive-outcome correlation depends on a Type of Motive Measure X Number of Activity Incentives X Outcome Operant Level interaction. The greater effect of activity incentives on the motive-outcome correlation for TAT motive measures as compared with questionnaire motive measures will increase as outcomes become more operant.

Although McClelland and his colleagues have investigated the possible two-way interactions between type of motive measure and number of social incentives (Hypothesis 3) and between type of motive measure and outcome operant level (Hypothesis 4), they have not considered the possible three-way interaction of type of motive measure, number of achievement social incentives in the environment, and outcome operant level. Social incentives interact with the self-attributed need for achievement. Because this self-attributed need is measured by questionnaires, there should be a positive relationship between questionnaire-based motive-outcome correlations and number of social incentives in the environment. But operant outcomes may allow greater opportunity for the value of achievement as measured by questionnaires to be expressed as compared with respondent measures of outcomes, which restrict or constrain the expression of questionnaire-measured achievement motivation. Neither the nature of the outcome, operant versus respondent, nor the number of social incentives in the situation should have much effect on TAT-based motive-outcome correlations because the TAT primarily measures implicit motives, which interact with activity incentives rather than with social incentives in the environment.

Hypothesis 6. The motive-outcome correlation is a function of a Type of Motive Measure X Number of Social Incentives X Outcome Operant Level interaction. The effect of social incentives on the motive-outcome correlation for TAT motive measures as compared with questionnaire motive measures will become increasingly negative as outcomes become more operant.

A further aspect of the controversy between McClelland and his defenders and critics of TAT measurement centers on correlations between TAT and questionnaire measures of needs. Klinger (1966), Entwisle (1972), Fineman (1977) and others reported that correlations between questionnaire and TAT measures of achievement are low. Proponents of TAT measurement (e.g., Atkinson & Litwin, 1960; McClelland, 1980, 1985) have concurred. However, there are at least two interpretations of this pattern of results. One interpretation is that TAT and questionnaire measures are different measures of a single construct such as need for achievement, that they should be highly correlated, and that the failure of TAT measures to correlate highly with questionnaire measures indicates a lack of convergent validity on the part of the TAT measure. McClelland (1980,1985; McClelland et al., 1989) has argued that the two techniques measure different constructs. Questionnaires measure self-attributed motives; the TAT measures more physiological and nonconscious implicit motives. In the present study a positive but small correlation between questionnaire and TAT measures of achievement was expected, either because the two type of measures measure the same construct or because the two types of measures might be subject to common cues or response bias of measurement.

Hypothesis 7. The average correlation between TAT and questionnaire measures of achievement will be positive and significant.

Summary

To test the competing claims of McClelland and associates and those of critics of TAT-based measures of achievement motivation, 105 empirical studies using either TAT or questionnaire measures of achievement motivation or both were content analyzed to provide data for two separate meta-analyses. In the first meta-analysis, the dependent variable was the achievement motive-outcome correlation. Independent variables in the first meta-analysis included type of motive measure (questionnaire, TAT), the number of activity incentives in the situation, and the number of social achievement incentives in the situation. Outcomes were classified as respondent (e.g., attitudes, opinions, school outcomes, ability and achievement tests), semioperant (laboratory measures of performance), or operant (e.g., income earned, occupational success, naturally occurring social behavior) on the basis of the presumed degree of environmental control over measured behavior. The first six hypotheses discussed above were tested by regressing the dependent variable, namely the motive-outcome correlation, on the four independent variables and on interaction terms formed from these four independent variables. The seventh hypothesis was tested in a second meta-analysis that was based on correlations between TAT and questionnaire measures of the achievement motive.

Method

Selection of Articles

To test the seven hypotheses developed in the present study, a comprehensive meta-analysis of original empirical research was undertaken. The first step was to identify available studies. All original research articles cited in four major literature reviews (Entwisle, 1972; Klinger, 1966; McClelland, 1980; Weinstein, 1969) were listed. A computer search for additional articles was undertaken. These two steps produced a list of 286 potentially useful articles. Articles randomly selected from this list were coded until a total of 139 articles had been selected, from which 105 articles were coded. At this point I decided that further data collection would not materially affect the results of the three meta-analyses, because the available 105 articles provided 490 correlations for analysis. Of the 34 articles selected but not used, 11 had no dependent variables or achievement was the dependent variable (articles with achievement as a dependent variable were eliminated because the focus of the present study was the predictive validity of measures of achievement motivation), 7 were not relevant to the meta-analysis at hand (e.g., need for affiliation rather than need for achievement was related to performance), and 16 articles contained insufficient data to be used.

A list of the 105 articles used in the present study may be found in the Appendix.

Coding of Articles

The first step in coding the selected 105 articles for subsequent analysis was to develop a coding form. Next I prepared a coding manual, which explained each item on the form and gave examples of how material from the articles was to be coded. A graduate research assistant and I independently coded the research articles. However, I reviewed the articles coded by the research assistant, and we discussed and resolved discrepancies. All data were completely coded and input before any tests of the seven hypotheses developed in this research were undertaken.

Table 1 summarizes the coding of four articles used in the present three meta-analyses.

Unit of Observation

An article may report results from more than one independent study or may provide statistics for a number of independent groups within a given study. Furthermore, a given article may test relationships between TAT-based measures of achievement motivation and performance on the one hand and relationships between questionnaire measures of achievement and performance on the other. An article may report one correlation on the basis of an operant measure of performance and another correlation on the basis of a respondent measure of performance. One correlation was perhaps based on a situation with achievement incentives, and another correlation from the same investigation was based perhaps on a no-achievement-incentives situation.

In these circumstances, it is not legitimate to collapse distinct correlations into some average statistic representing all research findings reported in an article. For example, to average a correlation that is based on an operant measure of performance with one from the same study that was based on a respondent measure would make it impossible to test any hypothesis involving outcome operant level, specifically Hypotheses 4, 5, and 6.

In other cases however, a number of statistics may be found for a single group of people on the basis of a single situation with a given number of achievement incentives, in which all outcome measures are either operant or respondent, and in which only one type of achievement measure, for example, TAT, was used to measure performance. There are multiple statistics perhaps because several similar measures of performance were used or because the measure of achievement was individually correlated with items from a personality test. In these cases, it is not legitimate to analyze these statistics separately for two reasons. First, these statistics are not independent because they come from the same sample. Second, including several similar correlations from a single group of individuals artificially inflates the total sample size used in the analysis. That is, a group of subjects is counted more than once.

In the present investigation, the unit of observation was defined to be a single statistic that was based on one type of achievement measurement (TAT or questionnaire), the same outcome operant level (respondent, semioperant, or operant), and one set of achievement incentives for a single group of individuals. In cases in which several similar correlations were available on a single group of people in a study, the correlations were averaged to produce one correlation for subsequent analysis.

This definition of the unit of observation produced 490 correlations for subsequent analysis. Of these, 190 were correlations of TAT achievement motivation with outcomes that were based on a total of 12,961 subjects; 193 were correlations between questionnaire measures of achievement and outcomes that were based on a sample size of 15,328; and 36 were correlations between TAT and questionnaire measures of achievement motivation that were based on a sample size of 2,785. The remaining 71 correlations were either correlations between projective but non-TAT measures and outcomes or overall study correlations that duplicated subgroup correlations used in the three meta-analyses.

Treatment of Estimated Data

Of the 490 correlations available for analysis, 45 were based on one or more estimated statistics such as F < 1 or r not significant. In these cases, the source article did not report the actual value of the statistic. If these nonsignificant statistics had been eliminated from the present study, the calculated average correlations might have been substantially higher than their population values. Therefore, I decided to estimate statistics reported as not significant. F not significant was converted into an equivalent correlation of 0, F< 1 was set to F= .5, p <. 1 was set at .05, correlations not significant were set to 0, / not significant was set to .5 where there was some evidence of a positive but nonsignificant relationship, for example, group means. In cases in which a nonsignificant t was reported together with means that were approximately the same, t was set to 0.

Determination of Sign Direction

Each statistic was examined for sign direction according to available theories of achievement motivation. For example, Atkinson (1957) proposed that high-need-for-achievement individuals would exhibit a greater preference for moderate risk than individuals low on need for achievement. However, a specific researcher might test this hypothesis by correlating TAT nAch with an index of preference for extreme risk. A negative correlation or a negative / statistic would confirm Atkinson’s (1957) hypothesis, but to add this confirmatory negative correlation to other correlations would reduce the magnitude of the average correlation between TAT achievement and outcomes. So in these cases it was necessary to change the sign of the correlation or ( statistic so that a positive sign always indicated a finding consistent with theory and a negative sign indicated a finding inconsistent with achievement theory.

A further difficulty emerged in coding the data. Much of the relevant research used statistics other than correlations, for example, tables of means and standard deviations, F statistics, and chi-squared statistics. Because these statistics do not have a sign associated with

Table 1
Illustrative Coding of Articles

					Incentives
Source	Outcome	Group and size	Result	r	Activity	Social
		Measured with	Thematic Apperception Test (TAT)
Miyamoto (1981)	Res	135 students, Grades 1-7	Grades	.37	Tast contingency, achievement work content; objective relationship*	Achievement instructions
Miyamoto (1981)	Res	1 35 students, Grades 1-7	IQ	.26	Time pressure, task contingency, achievement work content	Achievement instructions
Miyamoto (198!)	Res	135 students, Grades 1-7	Rated activity	.49	Task contingency, achievement work content	—
Miyamoto (1981)	Res	135 students	Attitudes, anxiety	.41 . 1 5	Task contingency, achievement work content	—
Atkinson and Litwin(1960)	Res	45 college students	Final exam persistence	.28”	Time pressure, task contingency, achievement work content; objective relationship	Achievement instructions
Atkinson and Litwin(1960)	Res	44 college students	Final grade	.32*	Time pressure, task contingency, achievement work content; Objective	Achievement instructions
Atkinson and Litwin(1960)	Sem	45 college students	Ringtoss game	.25	Moderate risk, achievement work content; Objective relationship	Achievement instructions
Singh (1979)	Op	200 farmers	Farm output	.68	Task contingency, achievement work content; objective	— —
Singh (1979)	OP	100 entre preneurs	Industrial output	.48	relationship Task contingency, achievement work content; objective relationship	—
			Measured with questionnaire
Hickson and Driskill (1970)	Res	68 students	GPA	.42	Task contingency, achievement work content; objective relationship	Achievement instructions
Hickson and Driskill (1970)	Res	68 students	GPA	.36	Task contingency, achievement work content; objective relationship	Achievement instructions

					Incentives
Source	Outcome	Group and size	Result		Activity	Social
		Measured	with questionnaire (continued)
Hickson and Driskill (1970)	Res	68 students	Honors program	.34	Achievement work content; objective relationship	Achievement norms
Hickson and Driskill (1970)	Res	68 students	Sensation seeking	.14	—	—
Hickson and Driskill (1970)	Res	68 students	Sensation seeking	.11	—	—
Atkinson and Litwin(1960)	Res	44 college students	Final exam persistence	-.13”	Time pressure, task contingency, achievement work content; objective relationship	Achievement instructions
Atkinson and Litwin(I960)	Res	44 college students	Final grade	.18”	Time pressure, task contingency, achievement work content; objective relationship	Achievement instructions
Atkinson and Litwin(1960)	Sem	43 college students	Ringtoss game	-.26”	Moderate risk, achievement work content; objective relationship	Achievement instructions
Singh (1979)	Op	200 farmers	Farm output	-.01	Task contingency, achievement work content; objective relationship	—
Singh (1979)	Op.	200 farmers	Farm output	-.03	Task contingency, achievement work content; objective	—
Singh (1979)	Op	100 entre- preneurs	Industrial output	.00	relationship Task contingency, achievement work content; objective relationship	—
		Correlation	between TAT and questionnaire
Atkinson and Litwin(1960) Miyamoto	—	47 college students 135 students,		-.05 .41		—
(1981)	—	Grades 1-7	—		—

Table 1 (continued)

Note. All studies measured achievement. Res = respondent, Sem = semioperant, Op = operant, GPA = grade point average. Objective relation- ship = high probability that individual action leads to achievement-related outcomes.

* Correlation coefficient derived from a contingency table.

them or the sign is always positive, it was necessary to assign a sign, positive or negative, to each of these statistics to indicate whether the achievement-related hypothesis under investigation was supported.

Transformation of Raw Data

The raw data found on the coding forms were transformed to provide measures of effect size, type of motive measure, number of activity incentives, number of social incentives, and outcome operant level.

Effect size. The articles coded in the present investigation reported a number of statistics in addition to correlations. Specifically, contingency tables, chi-squared statistics, /-“tests,; tests, p values, and tables of means and standard deviations were used. I converted all of these statistics into correlations using formulas provided by Rosenthal (1984, pp. 24-26) and Wolf (1986, pp. 35-36). A given study might report a number of related statistics. For example, a study might report four correlations or F statistics relating TAT achievement to a number of similar outcome measures for a single group of participants. For the present investigation, these multiple measures were first converted into correlations where necessary. The multiple correlations constituting a single observation were then transformed using Fisher’s r-to-z transformation, a weighted average of the transformed correlations was calculated, and the average was transformed back into a correlation (Rosenthal, 1984, p. 27). In cases in which similar measures of association were averaged to form one final correlation, a simple arithmetic average of the component sample sizes was the sample size for the final average correlation.

Type of motive measure. I used a dummy variable (0 = questionnaire; 1 = TAT) to classify motive-outcome correlations.

Number of activity incentives. Each observation was coded for the presence or absence (absence = 0; presence = 1) of the following 5 activity achievement incentives: moderate task risk, task contingency, achievement work content, time pressure, and a high objective relationship between performance and some achievement-related outcome in the immediate situation. An overall index of activity incentives was created for each observation by adding up the 5 scores. In the present set of data, number of activity incentives ranged from 0 to 5 per observation.

Number of social incentives. Each observation was coded for the presence or absence (absence = 0; presence = 1) of the following social achievement incentives: challenging goals set by an experimenter, achievement-oriented instructions in an experiment, achievement work norms, and pretreatment experimental manipulations. These scores were added up to provide a total social achievement-incentive index for each observation that in principle could range from no social achievement incentives in the work or experimental situation (0) to 3 social incentives in the work or experimental situation, 5 activity incentives in the pretreatment situation, and 3 social incentives in the prelreatmenl situation (11). Pretreatment activity incentives were coded as social incentives because they were extrinsic to the task or situation that generated the motive-outcome correlations analyzed. The actual number of social achievement incentives recorded from the 105 articles used in the present research ranged from 0 to 4.

Outcome operant level. School outcomes, ability and achievement tests, and measures of attitudes, opinions, and personality were classified as respondent outcomes and coded 1. Measures of performance in laboratory settings were classified as semioperant measures and coded 2. Income, job level attained in an organization, professional rank, publications, participation and leadership in community organizations, and social behavior occurring under natural conditions were classified as operant outcomes and coded 3.

Interrater Reliability

Once the coding of the articles had been completed, I undertook an analysis of interrater reliability. I hired and trained three students

enrolled in a PhD psychology program. I randomly selected, on the basis of a table of random numbers, 20 articles from those which had been used in the basic analysis. The three assistants read 2 background articles on the work of McClelland (McClelland, 1985; McClelland et al., 1989), studied the coding instructions used in the original coding, discussed with me basic aspects of the work of McClelland, and then practice coded the first 10 of the randomly selected articles. Their codes were compared with the original codes for the same articles and discrepancies were discussed. Subsequently, the three coders independently coded the remaining 10 articles from the set. These 10 articles produced a total of 33 distinct observations. Although the three research assistants were familiar with the basic theoretical propositions of McClelland and associates from their training, they were not aware of the seven specific hypotheses tested in the present research.

Data from the receding of the second set of 10 articles were correlated with the original data for the same 10 articles to obtain 3 interrater reliability coefficients for motive-outcome correlation, type of motive measure, outcome operant level, number of activity incentives, and number of social incentives. Excluding 1 interrater reliability coefficient of .33 for social incentives, the coefficients ranged from .67 for number of social incentives to 1.00 for outcome operant level. The median interrater reliability coefficient for these 14 coefficients was .89. The low coefficient of .33 may be accounted for in large part by the failure of one coder to code correctly achievement-oriented group or organization norms in 6 of 33 cases.

Statistical Procedures

To test the first six hypotheses developed in this article, motive-outcome correlations were regressed on type of motive measure, number of activity incentives, number of social incentives, outcome operant level, and seven interaction terms calculated from these independent variables. In this regression, z transformations of the original correlations were used, and each observation was weighted by n - 3, that is, by its observed sample size - 3 (Hedges & Olkin, 1985, pp. 237-239). Each independent variable was “centered,” that is, the mean of the variable was subtracted from each observation. Interaction termsspecified by Hypotheses 2-6 were created by multiplying together specified centered variables. In addition to the five interaction terms specified by Hypotheses 2-6, terms representing Number of Activity Incentives X Outcome Operant Level interaction and the Number of Social Incentives X Outcome Operant Level interaction were included in the regression. These terms were included because valid tests of higher order interactions such as those predicted by Hypotheses 5 and 6 required the inclusion in the regression equation of all component lower order interaction terms (Cohen & Cohen, 1983, pp. 345-346). Centered variables rather than the original variables were used in this analysis to reduce multicollinearity among independent variables and interaction terms calculated from these independent variables (Cohen & Cohen, 1983, p. 325).

Hypothesis 7 was tested by first calculating the weighted average z scores zw of all 36 available z-transformed correlations between TAT and questionnaire measures of achievement motivation and then using the statistic ^(N— 3fc)!/2, where k = number of correlations, to test the null hypothesis that the population correlation was 0 against the onetailed alternative that it was greater than 0 (Hedges & Olkin, 1985, p. 231).

Results

Table 2 contains descriptive statistics for motive-outcome correlations arranged by outcome operant level. Results from the regression used to test Hypotheses 1-6 are summarized in

Table 2
Motive-Outcome Correlations by Type of Motive Measure
and Outcome Operant Level

	Outcome type
Statistic	Respondent	Semioperant	Operant
Questionnaire
M	.15	.15	.13
SD	.17	.24	.13
n	89	92	12
P	.0001	.0001	.006
TAT
M	.19	.22	.22
SD	.18	.30	.22
n	108	45	37
P	.0001	.0001	.0001

Note. Respondent outcomes = measures of attitudes, opinions, per- sonality, school outcomes such as grade point average, and achieve- ment and ability tests scores. Semioperant outcomes = measures of performance and social behavior in laboratory situations. Operant out- comes = income earned, occupational success, participation in and leadership of community organizations, sales success, job perfor- mance, social behavior occurring in natural settings. TAT = Thematic Apperception Test.

Table 3. All results are expressed in terms of the original correlations rather than z-transformed correlations.

Hypothesis 1 predicted that the average motive-outcome correlation would be larger for questionnaire than TAT measures of achievement. Contrary to expectation, TAT-based correlations were significantly larger than questionnaire-based motive-outcome correlations. Hypothesis 3 was not confirmed. Hypotheses 2, 4, 5, and 6 were confirmed. Hypothesis 2 predicted that the effect of activity incentives on the motiveoutcome correlation would be larger for TAT measures than for questionnaire measures of the achievement motive. According to Hypothesis 4, the motive-outcome correlation would depend on a Type of Motive Measure X Outcome Operant Level interaction. The degree of increase in the motive-outcome correlation as a function of type of outcome (going from respondent to operant outcomes) would be greater for TAT measures of the achievement motive as compared with questionnaire measures of the achievement motive. Hypothesis 5 predicted a Type of Motive Measure X Number of Activity Incentives X Outcome Operant Level interaction. The greater effect of activity incentives on the motive-outcome correlation for TAT motive measures as compared with questionnaire motive measures would increase as outcomes became more operant. According to Hypothesis 6, the motive-outcome correlation would be a function of a Type of Motive Measure X Number of Social Incentives X Outcome Operant Level interaction. The effect of social incentives on the motive-outcome correlation for TAT motive measures as compared with questionnaire motive measures would become increasingly negative as outcomes become more operant.

Hypothesis 7 predicted that the average correlation between TAT nAch and questionnaire measures of achievement motivation would be positive and significant. From the content analysis of 105 articles used in the present research, 36 correlations between TAT and questionnaire measures of achievement were found with an average correlation of .088. The z statistic testing the hypothesis of no significance against the alternative that the correlation was greater than 0 was 4.548, p< .001. Hypothesis 7 was therefore accepted.

Discussion

These findings have a number of implications for the relative merits of TAT and questionnaire measures of achievement motives, the predictability of achievement behavior, the distinction between implicit motives and self-attributed motives, the distinction between social incentives and activity incentives, the possible suppressing effects of incentives on behavior, and the utility of laboratory studies of personality.

Relative Merits of TAT and Questionnaire Measures of Achievement Motives

Opponents of the TAT have maintained that TAT measures of achievement motivation are unreliable and invalid and that therefore the average correlation between questionnaire achievement measures and outcomes should be larger than the average correlation between TAT measures of achievement and outcomes. From Table 2 it is evident that in the present study

Table 3

Tests of Hypotheses 1 Through 6 From the Regression of Motive-Outcome Correlations on Independent Variables and Interaction Terms

Regression term	Predicted sign	Coefficient
Hypothesis 1
Intercept		.165ftt
TAT		.072ft
Outcome operant level		.040}
Number of activity incentives		.032ftt
Number of social incentives		-.046ft
Hypothesis 2
TAT X No. Activity Incentives		.032*
Hypothesis 3
TAT X No. Social Incentives		.027
Hypothesis 4
TAT X Outcome Operant Level		.085*
No. Activity Incentives X		.025
Outcome Operant Level
No. Social Incentives X		.031
Outcome Operant Level
Hypothesis 5
TAT X No. Activity Incentives X		.102**
Outcome Operant Level
Hypothesis 6
TAT X No. Social Incentives X		-.118*
Outcome Operant Level

Note. Thematic Apperception Test (TAT) was coded 0 for question naire measures of achievement motivation and 1 for TAT measures of achievement motivation. Outcome operant level was coded 1 for re- spondent outcomes such as school outcomes, ability and achievement tests, and measures of attitudes, opinions, and personality, 2 forsemi- operant outcomes such as performance in laboratory settings, and 3 for operant outcomes such as job level, professional rank, participation and leadership in community organizations, and social behavior occur- ring in natural settings. Coefficients in this table are unstandardized

regression coefficients. *p < .05, one-tailed. **p < .001, one-tailed, f p < .05, two- tailed, ff p < .01, two-tailed, fff p < .001, two-tailed.

the average correlation between TAT achievement and outcomes was higher than the average correlation between questionnaire achievement and outcomes for respondent, semioperant, and operant outcomes. Furthermore, from Table 3 it is evident that motive-outcome correlations that were based on the TAT were significantly larger than correlations based on questionnaire measures by .072 correlation points.

Predictability of Achievement Behavior

Although five of seven hypotheses were statistically supported in the present investigation and an unexpected positive effect for TAT measures was found, a question remains: Do TATs and questionnaires predict behavior at a nontrivial level? That is, are the results practically significant as well as statistically significant? The results presented in Tables 2 and 3 do not appear to be impressive, even though they are highly significant statistically. In Table 2, the average correlation between achievement measures and outcomes ranges from. 13 to .22. That is, at most 5% of the variance in outcomes is predicted by either TAT or questionnaire measures of the achievement motive. Likewise, the coefficients in Table 3, although several are statistically significant, appear to be modest.

To understand the significance of the findings reported in Tables 2 and 3,1 calculated several expected motive-outcome correlations by inserting various values of independent variables and interaction terms into the regression equation summarized in Table 3 and by solving for the motive-outcome correlation. Certain combinations of type of motive measure, outcome operant levels, and levels of social and activity incentives produce low expected correlations. For example, the expected correlation between questionnaire achievement and respondent outcomes in the absence of activity and social achievement incentives is. 13, rather modest. Similarly, the expected correlation between TAT achievement and respondent outcomes in the absence of social and activity achievement incentives is. 13. On the other hand, the expected correlation between questionnaire achievement and real-world (i.e., operant) behavior in the presence of, for example, four social achievement incentives and no activity incentives is .35. The expected correlation between TAT achievement and operant outcomes in the presence of four activity incentives and no social incentives is .66.

In short, neither questionnaires nor the TAT predict achievement behavior well in respondent situations in the absence of appropriate incentives, but the TAT in the presence of activity incentives predicts operant behavior extraordinarily well, and questionnaire measures in the presence of social incentives strongly predict operant behavior. Of course, it was precisely under conditions of activity achievement incentives and operant outcomes that TAT measures of achievement were expected to predict behavior (Hypothesis 5), and it was precisely under conditions of operant outcomes and social incentives that questionnaire measures of achievement were expected to predict outcomes (Hypothesis 6).

Implicit Motives and Self-Attributed Motives

For some time, McClelland has contended that the TAT measures nonconscious needs for achievement and other motives, whereas questionnaires measure more conscious values of achievement, affiliation, and power. Opponents of TAT measurement have not generally accepted this distinction. They appear to be! ieve there is one set of motives with two alternative methods of measurement, TAT versus questionnaire. Questionnaires are reliable and valid, and TAT measures are neither reliable nor valid in their view.

Results of the present investigation support the distinction made by McClelland and his associates. TATs and questionnaires appear to be measuring different aspects of personality. First, TAT nAch and questionnaire measures of achievement were found to have a low average correlation of .088. Second, the argument that the TAT is an invalid measure of need for achievement does not appear to be correct, given the findings that TAT achievement strongly predicted operant behavior in the presence of activity achievement incentives. Third, one indication that two measures are measures of different underlying constructs is their differential interaction with environmental stimuli. In the present study, type of motive measure, number of activity incentives, and outcome operant level interacted to positively affect motive-outcome correlations (Hypothesis 5), but type of motive measure, social incentives, and outcome operant level interacted to negatively affect motive-outcome correlations (Hypothesis 6).

An Unexpected Finding: The Suppressing Effects of Incentives

Hypotheses 5 and 6 were based on the proposition advanced by McClelland et al. (1989) that activity incentives would interact with implicit motives as measured by the TAT and that social incentives would interact with self-attributed motives as measured by questionnaires. Both hypotheses were confirmed, but an inspection of the data suggested that activity incentives possibly interacted with self-attributed motives and that social incentives possibly interacted with implicit motives. In both cases, these Motive X Incentive interactions apparently acted to reduce the motive-outcome correlation, particularly for more operant outcomes.

To test this observation, I regressed motive-outcome correlations on activity incentives, social incentives, the Social Incentives X Outcome Operant Level interaction term, and the Activity Incentives X Outcome Operant Level interaction term twice, once for questionnaire-based motive-outcome correlations and once for TAT-based motive-outcome correlations. In the TATbased regression, the Activity Incentives X Outcome Operant Level interaction term significantly predicted the motive-outcome correlation (j3 = .082, / = 4.64, p < .001, one-tailed) as expected, but the Social Incentives X Outcome Operant Level interaction term was negatively related to the motive-outcome correlation (0 = -. 106, t = -3.24, p < .005, two-tailed). In the questionnaire-based regression, the Social Incentives X Outcome Operant Level interaction term was significant as expected (j8 = .093, / = 2.74, p < .005, one-tailed), but the Activity Incentives X Outcome Operant Level interaction term was negatively although not significantly related to the motive-outcome correlation (ft = -.027, t = -1.36, p = . 175).

A tentative explanation for this phenomenon may be advanced. It is possible that just as a high level of a motive and an appropriate incentive creates a pleasurable effect that reinforces future similar behavior, a high level of some motive combined with an inappropriate incentive may create an aversive effect that temporarily suppresses or extinguishes behavior.

Utility of Laboratory Studies of Personality

A characteristic of much laboratory research is the coercive nature of the research setting. Researchers want to control their experimental variables, measure reliably and validly their dependent variables, eliminate competing explanations of their findings, and provide enough structure to make their work replicable by other researchers. An unintended consequence of the scientific method may be to minimize the expression of individual differences and the interaction of individual differences and environmental characteristics.

The present study provides some evidence of this effect. As outcomes become more operant, motive-outcome correlations become larger for TAT-based correlations compared with questionnaire-based measures (Hypothesis 4). The impact of activity incentives on motive-outcome correlations measured by the TAT becomes larger as outcomes become more operant (Hypothesis 5), and the impact of social incentives on motive-outcome correlations that are based on questionnaire measures increases as outcomes become more operant (Hypothesis 6).

In summary, the relative merits of questionnaire versus TAT measures of achievement motivation have been debated in the literature for decades. The present study suggests that both TAT and questionnaire measures of motives have an important role in understanding and predicting human behavior.

References

Atkinson, J. W (1957). Motivational determinants of risk-taking behavior. Psychological Review, 64, 359-372.
Atkinson, J. W (1982). Motivational determinants of thematic apperception. In A. Stewart (Ed.), Motivation and society (pp. 3-40). San Francisco: Jossey-Bass.
Atkinson, J. W, & Feather, N. T. (1966). A theory of achievement motivation. New York: Wiley.
Atkinson, J. W, & Litwin, G. H. (1960). Achievement motive and test anxiety conceived as motive to approach success and motive to avoid failure. Journal of Abnormal and Social Psychology, 60, 52-63.
Campbell, D. T., & Fiske, D. W (1959). Convergent and discriminant validation by the multi-trait-multimethod matrix. Psychological Bulletin, 56, 81-105.
Cohen, C, & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences. Hillsdale, NJ: Erlbaum.
deCharms, R., Morrison, H. W, Reitman, W, & McClelland, D. C.

(1955). Behavioral correlates of directly and indirectly measured achievement motivation. In D. C. McClelland (Ed.), Studies in motivation (pp. 414-423). Englewood Cliffs, NJ: Prentice-Hall.

Entwisle, D. R. (1972). To dispel fantasies about fantasy-based measures of achievement motivation. Psychological Bulletin, 77, 377- 391.
Fineman, S. (1977). The achievement motive and its measurement: Where are we now? British Journal of Psychology, 68,1-22.
French, E. G. (1955). Some characteristics of achievement motivation. Journal of Experimental Psychology, 50, 232-236.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA: Academic Press.
Hunter, J. E., Schmidt, F. L., & Jackson, G. B. (1982). Meta-analysis: Cumulating research findings across studies. Beverly Hills, CA: Sage.
Klinger, E. (1966). Fantasy need achievement. Psychological Bulletin, 66,291-306.
McClelland, D. C. (1961). The achieving society. Princeton, NJ: Van Nostrand.
McClelland, D. C. (1965). Toward a theory of motive acquisition. American Psychologist, 20, 321-333.
McClelland, D. C. (1972). Opinions predict opinions: So what else is newt Journal of Consulting and Clinical Psychology, 38, 325-326.
McClelland, D. C. (1980). Motive dispositions: The merits of operant and respondent measures. In L. Wheeler (Ed.), Review of personality and social psychology (Vol. 1, pp. 10-41). Beverly Hills, CA: Sage.
McClelland, D. C. (1985). How motives, skills, and values determine what people do. American Psychologist, 40, 812-825.
McClelland, D. C, Atkinson, J. W, Clark, R. A., & Lowell, E. L. (1958). A scoring manual for the achievement motive. In J. W Atkinson (Ed.), Motives in fantasy, action, and society (pp. 179-204). Princeton, NJ: Van Nostrand.
McClelland, D. C., Clark, R. A., Roby, T. B., & Atkinson, J. W (1958). The effect of the need for achievement on thematic apperception. In J. W Atkinson (Ed.), Motives in fantasy, action, and society (pp. 64- 82). Princeton, NJ: Van Nostrand.
McClelland, D. C, Koestner, R., & Weinberger, J. (1989). How do selfattributed and implicit motives differ? Psychological Review, 96, 690-702.
McClelland, D. C, & Winter, D. (1969). Motivating economic achievement. New York: Free Press.
Mischel, W (1972). Direct versus indirect personality assessment: Evidence and implications. Journal of Consulting and Clinical Psychology, 38, 319-324.
Patten, R. L., & White, L. A. (1977). Independent effects of achievement motivation and overt attribution on achievement behavior. Motivation and Emotion, I, 39-59.
Raynor, J. O. (1969). Future orientation and motivation of immediate activity: An elaboration of the theory of achievement motivation. Psychological Review, 76, 606-610.
Raynor, J. O. (1970). Relationships between achievement-related motives, future orientation, and academic performance. Journal of Personality and Social Psychology, 15, 28-33.
Rosenthal, R. (1984). Meta-analylic procedures for social research. Beverly Hills, CA: Sage.
Scott, W A., & Johnson, R. C. (1972). Comparative validities of direct and indirect personality tests. Journal ofConsulting and’Clinical Psychology, 38, 301-318.
Weinstein, M. S. (1969). Achievement motivation and risk preference. Journal of Personality and Social Psychology, 13,153-172.
Winter, D. G., & Stewart, A. J. (1977). Power motive reliability as a function of retest instructions. Journal of Consulting and Clinical Psychology, 45, 436-440.
Wolf, F. M. (1986). Meta-analysis: Quantitative methods for research synthesis. Newbury Park, CA: Sage.

(Appendix follows on next page)

152 WILLIAM D. SPANGLER

Appendix

Articles Used in the Three Meta-Analyses

Adams, A. J., & Stone, T. H. (1977). Satisfaction of need for achievement in work and leisure time activities. Journal a/Vocational Behavior, 11, 174-181.
Andrews, J. D. W. (1967). The achievement motive and advancement in two typesof organizations. Journal of Personally and Social Psychology, 6, 163-168.
Atkinson, J. W, & Litwin, G. H. (1960). Achievement motive and test anxiety conceived as motive to approach success and motive to avoid failure. Journal of Abnormal and Social Psychology, 60, 52-63.
Atkinson, J. W, & Raphelson, A. C. (1956). Individual differences in motivation and behavior in particular situations. Journal of Personality, 24, 349-363.
Bailey, R. C., & Mettetal, G. W (1977). Sex differences in the congruency of perceived intelligence. Journal of Genetic Psychology, 131, 29-36.
Bendig, A. W (1958). Predictive and postdictive validity of need achievement scores. Journal of Educational Research, 52, 119-120.
Catalano, J. F, & Berger, D. F. (1970). The effect of need for achievement and normative information on rotary pursuit performance. Psychonomic Science, 18, 69-71.
Ceranski, D. S., Teevan, R., & Kalle, R. J. (1979). A comparison of three measures of the motive to avoid failure: Hostile press, test anxiety, and resultant achievement motivation. Motivation and Emotion, 3, 395-404.
Cooper, W H. (1983). An achievement motivation nomological network. Journal of Personality and Social Psychology, 44, 841 -861.
Covington, M V, & Omelich, C. L. (1979). Are causal attributions causal? A path analysis of the cognitive model of achievement motivation. Journal of Personality and Social Psychology, 37,1487-1504.
Crockett, H. J., Jr. (1962). The achievement motive and differential occupational mobility in the United States. American Sociological Review. 27, 191-204.
deCharms, R., & Dave, P. N. (1955). Hope of success, fear of failure, subjective probability, and risk-taking behavior. Journal of Personality and Social Psychology, 1, 558-568.
Durand, D., & Shea, D. (1974). Entrepreneurial activity as a function of achievement motivation and reinforcement control. Journal of Psychology, 88, 57-63.
Elder, G. H., & Maclnnis, D. J. (1983). Achievement imagery in women’s lives from adolescence to adulthood. Journal of Personality and Social Psychology, 45, 394-404.
Epps, E. G., Katz, I., Perry, A., & Runyon, E. (1971). Effect of race of comparison referent and motives on Negro competitive performance. Journal of Educational Psychology, 62, 201-208.
Erdwins, C. J., Tyer, Z. E., & Mellinger, J. C. (1980). Personality traits of mature women in student versus homemaker roles. Journal of Psychology, 105, 189-195.
Farley, F. H., & Mealiea, W L. (1973). Motivation and recall of completed and uncompleted achievement items. Journal of Educational Research, 66, 302-306.
Feather, N. T. (1965). The relationship of expectation of success to need achievement and test anxiety. Journal of Personality and Social Psychology, 1, 118-126.
Ferguson, E. D. (1971). Relationship between ego-involvement instructions and various motivation and performance measures. Psychological Reports, 29, 547-556.
Forward, J., Wells, K., Canter, R., & Waggoner, M. (1975). Teacher control strategies and choice of educational objectives among college students. Journal of Educational Psychology, 67, 757-763.
French. E. G. (1955). Some characteristics of achievement motivation. Journal of Experimental Psychology, 50, 232-236.
Gould, S. (1980). Need for achievement, career mobility, and the Mexican-American college graduate. Journal ofVocational Behavior, 16, 73-82.
Greene, D. L., & Winter, D. G. (1971). Motives, involvements, and leadership among Black college students. Journal of Personality, 39,319- 332.
Hall, L. H. (1972). Personality variables of achieving and non-achieving Mexican-American and other community college freshmen. Journal of Educational Research, 65, 224-228.
Hancock, J. G, & Teevan, R. C. (1964). Fear of failure and risk-taking behavior. Journal of Personality, 32, 200-209.
Harrell, A. M., & Stahl, M. J. (1981). A behavioral decision theory approach for measuring needs. Journal of Applied Psychology, 66, 242-247.
Heise, D. R. (1966). Sensitization of verbal response-dispositions by N affiliation and N achievement. Journal of Verbal Learning and Verbal Behavior, 5, 522-525.
Helmreich, R. L., Spence, J. X, Beane, W E., Lucker, G. W, & Matthews, K. A. (1980). Making it in academic psychology: Demographic and personality correlates of attainment. Journal of Personality and Social Psychology, 39, 896-908.
Helmreich, R. L., Swain, L. L., & Carsrud. (1986). The honeymoon effect in job performance: Temporal increases in the predictive power of achievement motivation. Journal of Applied Psychology, 71, 185-188.
Henschen, K. P., Edwards, S. W, & Mathinos, L. (1982). Achievement motivation and sex-role orientation of high school female track and field athletes versus nonathletes. Perceptual and Motor Skills, 55, 183-187.
Hickson, R. H., & Driskill, J. C. (1970). Needs for achievement: Differences between honors and non-honors students. Journal of Experimental Education, 38, 37-38.
Holmes, D. S., & Tyler, J. D. (1968). Direct versus projective measurement of achievement motivation. Journal of Consulting and Clinical Psychology, 32, 712-717.
Honor, S. H., & Vane, J. R. (1972). Comparison of Thematic Apperception Test and questionnaire methods to obtain achievement attitudes of high-school boys. Journal of Clinical Psychology, 28, 81 -83.
Hurley, J. R. (1957). Achievement imagery and motivational instructions as determinants of verbal learning. Journal of Personality, 25, 274-282.
Inkson, J. H. (1971). Achievement motivation and occupational choice. Australian Journal of Psychology, 23, 225-234.
Jardine, E., & Winefield, A. H. (1981). Achievement motivation, psychological reactance, and learned helplessness. Motivation and Emotion, 5, 99-113.
Johnson, D. M. (1973). Relationships between selected cognitive and noncognitive variables and practical nursing achievement. Nursing Research, 22, 148-153.
Johnston, R. A. (1955). The effects of achievement imagery on mazelearning performance. Journal of Personality, 23,145-152.
Johnston, R. A. (1957). A methodological analysis of several revised forms of the Iowa Picture Interpretation Test. Journal of Personality, 25, 283-293.
Karabenick, S. A. (1972). Valence of success and failure as a function of achievement motives and locus of control. Journal of Personality and Social Psychology, 21, 101-110.
Karabenick, S. A., & Youssef, Z. I. (1968). Performance as a function of achievement motive level and perceived difficulty. Journal of Personality and Social Psychology, JO, 4] 4-419.
Kelly, R. T, Harve, E., & Rodger, L. (1973). Interaction effects of

achievement need and situational press on performance. Journal of Social Psychology, 89,141-145.

Kukla, A. (1972). Attributional determinants of achievement-related behavior. Journal of Personality and Social Psychology, 21,166-174.
Lao, R. C. (1980). Differential factors affecting male and female academic performance in high school. Journal of Psychology, 104,119- 127.
Latta, R. M. (1978a). Hope of success and fear of failure components of Mehrabian’s scales of resultant achievement motivation. Journal of Research in Personality, 12,151.
Latta, R. M. (1978b). Interactive effects of initial achievement orientation and prior success feedback on the mastery of subsequent difficult and easy tasks. American Educational Research Journal, /5,17- 24.
Lifshitz, M. (1974). Achievement motivation and coping behavior of normal and problematic preadolescent kibbutz children. Journal of Personality Assessment, 38, 138-143.
Lindgren, H. C., Moritsch, B., & Thulin, E. K. (1986). Studies of three measures of achievement motivation. Psychological Reports, 59, 123-136.
Lowe, C. A., & Medway, F. J. (1976). Effects of achievement motive on divergent self-other attributions for success and failure. Perceptual and Motor Skills, 42,127-132.
Lowell, E. L. (1952). The effect of need for achievement on learning and speed of performance. Journal of Psychology, 33, 31 -40.
Lueptow, L. B. (1973). Need for achievement and continuation in college. Psychological Reports, 33, 455-458.
McAdams, D. P., & Vaillant, G. E. (1982). Intimacy motivation and psychosocial adjustment: A longitudinal study. Journal of Personality Assessment, 46, 586-593.
McClelland, D. C, & Liberman, A. M. (1949). The effect of need for achievement on recognition of need-related words. Journal of Personality, 75,236-251.
McFall, R. M., & Schenkein, D. (1970). Experimenter expectancy effects, need for achievement, and field dependence. Journal of Experimental Research in Personality, 4, 122-128.
McKeachie, W J. (1961). Motivation, teaching methods, and college learning. In M. R. Jones (Ed.), Nebraska symposium on motivation: 1961 (pp. 111-142). Lincoln: University of Nebraska Press.
McKeachie, W J., Isaacson, R. L, & Milholland, J. E. (1968). Student achievement motives, achievement cues, and academic achievement. Journal of Consulting and Clinical Psychology, 32, 26-29.
Meade, R. D. (1966). Achievement motivation, achievement, and psychological time. Journal of Personality and Social Psychology, 4, 577-580.
Melnick, M. (1974). Need achievement, need affiliation, and perception of laterally in other people. Perceptual and Motor Skills, 38, 467-470.
Mitchell,! V, Jr. (1961). An analysis of the factorial dimensions of the achievement motivation construct. Journal of Educational Psychology, 52,179-187.
Miyamoto, M. (1981). Instrumental activity in achievement motivation. Japanese Psychological Research, 23, 79-87.
Morgan, H. H. (1952). A psychometric comparison of achieving and nonachieving college students of high ability. Journal of Consulting Psychology, 16, 292-298.
Morgan, H. H. (1953). Measuring achievement motivation with “picture motivations.” Journal of Consulting Psychology, 17, 289-292.
Morris, J. L., & Fargher, K. (1974). Achievement drive and creativity as correlates of success in small business. Australian Journal of Psychology, 26,217-222.
Moulton, R. W (1965). Effects of success and failure on level of aspiration as related to achievement motives. Journal of Personality and Social Psychology, 1, 399-406.
Nygard, R. (1982). Achievement motives and individual differences in

situational specificity of behavior. Journal of Personality and Social Psychology, 43, 319-327.

O’Connell, A. N. (1980). Effects of manipulated status on performance, goal setting, achievement motivation, anxiety, and fear of success. Journal of Social Psychology, 112, 75-89.
Orso, D. P. (1969). Comparison of achievement and affiliation arousal on nAch. Journal of Protective Techniques and Personality Assessment, 33, 230-233.
Patrick, A. W, & Zuckerman, M. (1977). An application of the statetrait concept to the need for achievement. Journal of Research in Personality, 11, 459-465.
Patten, R. L., & White, L. A. (1977). Independent effects of achievement motivation and overt attribution on achievement behavior. Motivation and Emotion, 1, 39-59.
Ray, J. J. (1980). Effects of individual differences on productivity among farmers in India. Journal of Social Psychology, 112, 11-17.
Raynor, J. O. (1970). Relationship between achievement-related motives, future orientation, and academic performance. Journal of Personality and Social Psychology, 15, 28-33.
Raynor, J. O., & Smith, C. P. (1966). Achievement-related motives and risk-taking in games of skill and chance. Journal of Personality, 34, 176-198.
Reiter, H. H. (1964). Prediction of college success from measures of anxiety, ach ievement motivation, and scholastic aptitude. Psychological Reports, 15, 23-26.
Rodgers, C. W (1973). Relationship of projective to direct expression of selected needs for nonpsychotic Ss. Perceptual and Motor Skills, 36, 571-578.
Rosen, B. C. (1955). The achievement syndrome. American Sociological Review, 21, 203-211.
Sachs, P. R. (1982). Avoidance of diagnostic information in self-evaluation of ability. Personality and Social Psychology Bulletin, 82, 242- 246.
Sampson, E. E. (1963). Achievement in conflict. Journal of Personality, 31, 510-516.
Schroth, M. L. (1985). The effect of differing measuring methods on the relationship of motives. Journal of Psychology, 119, 213-218.
Schultz, C. B., & Pomerantz, M. (1976). Achievement motivation, locus of control, and academic achievement behavior. Journal ofPersonality, 44, 38-51.
Scodel, A., Ratoosh, P., & Minas, J. S. (1959). Some personality correlates of decision making under conditions of risk. Behavioral Science, 4, 19-28.
Scott, W A., & Johnson, R. C. (1972). Comparative validities of direct and indirect personality tests. Journal of Consulting and Clinical Psychology, 38, 301-318.
Sedney, M. A., & Turner, B. F. (1975). A test of causal sequences in two models for development of career-orientation in women. Journal of Vocational Behavior, 6, 281-291.
Sherwood, J. J. (1966). Self-report and projective measures of achievement and affiliation. Journal of Consulting Psychology, 30, 329-337.
Singh, N. P. (1969). rcAch among successful-unsuccessful and traditional progressive agricultural entrepreneurs of Delhi. Journal of Social Psychology, 79, 271-272.
Singh, N. P. (1970). n/Ach among agricultural and business entrepreneurs of Delhi. Journal of Social Psychology, 81, 145-149.
Singh, S. (1979). Relationships among projective and direct verbal measures of achievement motivation. Journal of Personality Assessment, 43, 45-49.
Skolnick, A. (1966). Motivational imagery and behavior over twenty years. Journal of Consulting Psychology, 30, 63-478.
Smith, C. P. (1963). Achievement-related motives and goal setting under different conditions. Journal of Personality, 31, 124-140.
Sorrentino, R. M., & Field, N. (1986). Emergent leadership over time: The functional value of positive motivation. Journal of Personality and Social Psychology, 50,1091-1099.
Steers, R. M. (1975). Task-goal attributes, n achievement, and supervisory performance. Organizational Behavior and Human Performance, 13, 392-403.
Tasaki, T., & Seki, F. (1975). Effects of affiliation motive, achievement motive, and pressure sources on conformity. Japanese Psychological Research, 17, 174-181.
Teevan, R. C, & Smith, B. D. (1975). Relationship of fear-of-failure and need achievement motivation to a confirming-interval measure of aspirational levels. Psychological Reports, 36, 967-976.
Teglasi, H. (1978). Sex-role orientation, achievement motivation, and causal attributions of college females. Sex Roles, 4, 381-397.
Trope, Y. (1975). Seeking information about one’s ability as a determinant of choice among tasks. Journal of Personality and Social Psychology, 32, m*-?>.
Weiner, B., & Kukla, A. (1970). An attributional analysis of achievement motivation. Journal of Personality and Social Psychology, 15, 1-20.
Weiner, B., & Sierad, J. (1975). Misattribution for failure and enhancement of achievement strivings. Journal of Personality and Social Psychology, 31,415-421.
Weinstein, M. S. (1969). Achievement motivation and risk preference. Journal of Personality and Social Psychology, 13,153-172.
Weitzenkorn, S. D. (1974). An adjusted measure of achievement motivation for males and females and effects of future orientation on level of performance. Journal of Research in Personality, 8, 361-377.
Williams, J. M., Hoepner, B. J., Moody, D. L., & Ogilvie, B. C. (1970). Personality traits of champion female fencers. Research Quarterly, 41, 446-453.
Winefield, A. H., & Jardine, E. (1982). Effects of differences in achievement motivation and amount of exposure on responses to uncontrollable rewards. Motivation and Emotion, 6, 245-257.
Winefield, A. H., & Norris, P. J. (1981). Effects of exposure to uncontrollable events as a function of achievement motivation and initial expectation of success. Motivation and Emotion, 5, 235-248.
Winter, D. G., & Wiecking, F. A. (1971). The new Puritans: Achievement and power motives of New Left radicals. Behavioral Science, 16, 523-530.
Wolk, S., & Du Cette, J. (1971). Locus of control and achievement motivation: Theoretical overlap and methodological divergence. Psychological Reports, 29, 755-758.
Yamauchi, H., & Doi, K. (1977). Factorial study of several motives. Psychological Reports, 41, 795-801.
Zander, A., & Forward, J. (1968). Position in group, achievement motivation, and group aspirations. Journal of Personality and Social Psychology, 8, 282-288.

Received October 25,1989 Revision received August 3,1991

Accepted August 12,1991

Footnotes

h1↩︎

Validity of Questionnaire and TAT Measures of Need for Achievement: Two Meta-Analyses

William D. Spangler School of Management State University of New at Binghamton

Theoretical Background

Seven Hypotheses

Summary

Method

Selection of Articles

Coding of Articles

Unit of Observation

Treatment of Estimated Data

Determination of Sign Direction

Table 1 (continued)

Transformation of Raw Data

Interrater Reliability

Statistical Procedures

Results

Discussion

Relative Merits of TAT and Questionnaire Measures of Achievement Motives

Table 3

Predictability of Achievement Behavior

Implicit Motives and Self-Attributed Motives

Social Achievement Incentives and Activity Achievement Incentives

An Unexpected Finding: The Suppressing Effects of Incentives

Utility of Laboratory Studies of Personality

References

152 WILLIAM D. SPANGLER

Appendix

Articles Used in the Three Meta-Analyses

Footnotes