Unconscious Motivation Implicit Motives In most of the studies previously described, automatic processes or preattentive stimuli have effects on conscious motivation. Now it is time to turn to the question originally raised by McClelland et al. (1989): whether motivational states, themselves, operate outside of the subject’s phenomenal awareness. Traditional self-reports of motivation, such as those collected by the Adjective Check List (ACL;) (Gough & Heilbrun, 1965), the Edwards Personal Preference Schedule (EPPS;) (Edwards, 1959), or the Personality Research Form (PRF;) (Jackson, 1967) are obviously unsuitable for assessing unconscious motives: subjects cannot report on something they don’t know. Accordingly, from an early stage in his research program McClelland preferred “indirect” motive measures such as Murray’s Thematic Apperception Test (TAT;) (Murray, 1943) to “direct” measures like questionnaires (deCharms, Morrison, Reitman, & McClelland, 1955). Later, McClelland embraced a distinction introduced by B.F. Skinner (Skinner, 1935), his Harvard colleague, between “operant” (performance) and “respondent” (selfreport) measures of motivation (D.C. McClelland, 1980), before settling on the explicitimplicit formulation. In either case, McClelland believed that performance measures of motivation were superior to self-report measures – a preference that he extended to intelligence tests as well (D. C. McClelland, 1973). Of, course, by their very nature selfreports cannot reflect unconscious motives or any other unconscious features of personality. To that end, McClelland introduced the “Picture-Story Exercise” (PSE) as the preferred method for assessing implicit motives (Figure 2). In the PSE, as in the TAT, Unconscious Motivation subjects are asked to generate a story in response to a picture selected to tap internal motives, such as the needs for achievement (nAch;) (D.C. McClelland et al., 1953), power (nPow;) (Winter, 1973), and affiliation or intimacy (nInt;) (McAdams, 1989). The stories are then coded with an empirically derived content-analysis scheme to yield an assessment of the individual’s standing on the motive in question (Pang, 2010; Schultheiss & Pang, 2007; Schultheiss & Schultheiss, 2014). But nowhere are subjects asked to introspect, or report, on their motives as such. Figure 2. A prodigy being called to the stage, or a boy wishing he were outside playing baseball? Actually, a photograph of the 20th-century violin virituoso Yehudi Menuhin, age 12. Another portrait from this series, redrawn by Christiana Morgan, served as the basis for Card 1 of the TAT (Jahnke & Morgan, 1997). Unconscious Motivation Dissociating Explicit and Implicit Motivation The idea that the PSE assesses unconscious motives was not entirely new. Murray himself had assumed that subjects might not be aware of some of their salient motives, but that they would identify with, and project their own internal needs onto, the “hero” or protagonist of their stories (Lindzey, 1952; Lindzey & Kalnins, 1958). He made this clear in his manual for the TAT: “[T]he content of a set of TAT stories represents second level, covert... personality, not first level, overt or public... personality.... [T]he TAT is one of the few methods available today for the disclosure of covert tendencies” (Murray, 1943) (p. 16). And Lindzey (Lindzey, 1952) noted that “The dispositions and conflicts that may be inferred from the story-teller’s creations are not always reflected directly in overt behavior or consciousness” (p. 3). Reviewing the early literature on the test, he concluded that “the assumed imperfect correlation” between “fantasied and overt behavior is warranted” (p. 20). But the evidence itself had more to do with the relation between TAT scores and overt behavior – i.e., the external validity of the TAT -than with the relation between conscious and unconscious motives. More recent studies have operationalized conscious motivation in terms of subjects’ self-reports on personality questionnaires. McClelland et al., for example, reported very low correlations betwen PSE and questionnaire measures of nAch, nPow, and nAff (unweighted, unsigned aggregate r = .11; ) (D.C. McClelland et al., 1989). Spangler (Spangler, 1992), reviewing 36 studies employing both TAT/PSE and questionnaire measures of nAch, reported an average correlation of .00. A more recent meta-analysis by Kollner and Schultheiss (Kollner & Schultheiss, 2014) oftained average correlations of .14 for nAch, .12 for nAff, and .04 for nPow. Although all these Unconscious Motivation aggregate correlations are statistically significant, they are low even by the standards of personality research (Mischel, 1968). As such, they provide prima facie evidence of a dissociation between explicit, conscious and implicit, unconscious motives. Are Implicit Motives Truly Unconscious? Of course, a cynic might reply that the PSE is simply a poor measure of motivation, which explains why it does not correlate with established self-report methods. And, indeed, the TAT and related measures have been criticized on psychometric grounds almost since their inception – particularly for their ostensibly low reliability (e.g., ) (Entwisle, 1972). On the other hand, as noted even as early as Lindzey (Lindzey, 1952), there is a great deal of evidence for the validity of TAT-like measures, in that they actually predict subjects’ performance on various motive-related tasks (for comprehensive coverage, see) (Schultheiss, 2008; Schultheiss & Brunstein, 2010). As McClelland himself often pointed out, if a test is valid, it must perforce be reliable (because, in psychometric terms, a test cannot correlate higher with an external criterion than it does with itself). McClelland also argued that test-retest reliability, at least, was problematic for measures of motivation, because a motive discharged at the time of an initial test might not have recharged at the time of the retest: once we’ve eaten, we’re not hungry again for a while. In any event, problems with inter-rater reliability, created by the lack of an explicit scoring manual for the TAT comparable to those developed for the Rorschach (e.g.,) (Beck, 1944), have been solved by the availability of clear manuals for scoring various motives on the PSE (e.g.,) (Schultheiss & Pang, 2007; C. P. Smith, 1992; Winter, 1991). Unconscious Motivation Still, a low correlation between explicit and implicit measures of motivation is only the first step in establishing a dissociation between explicit and implicit motives. There are many reasons why such correlations could be low, besides the fact that the variables in question reflect different levels of consciousness. To understand why this is the case, it will help to analyze explicit and implicit motivation in terms of the multitraitmultimethod matrix introduced by Campbell and Fiske (Campbell & Fiske, 1959), as illustrated in Table 2. Consider two traits – e.g., nAch and nPow; there are two explicit measures of each trait – e.g., a 0-10 self-rating scale and a self-report questionnaire; and there are two implicit measures of each trait – e.g., the PSE and a reaction-time measure such as the Implicit Association Test (IAT;) (Banaji & Greenwald, 2013; A.G. Greenwald et al., 1998).  In the table, the diagonal represents the reliability of each individual measure – i.e., its reliability. This value should be very high – ideally, a correlation approaching a perfect 1.0.  Measures that share both trait variance (e.g., both are measures of nAch) and method variance (e.g., both are explicit measures) should be highly correlated – perhaps, achieving correlations as high as .70 or so.  Measures that share neither trait variance nor method variance have nothing in common, and should be essentially uncorrelated, with correlations of approximately 0.  Measures of two different traits (e.g., self-reports of nAch and nPow) may be significantly correlated simply because they have some method variance in common – though such correlations should be relatively low. Unconscious Motivation  In principle, any two measures of the same trait should be significantly correlated. The exception is when the two measures differ on the explicit-implicit continuum: to the extent that implicit motivation can be dissociated from explicit motivation, the correlation between the two measures should be as low as possible approaching zero. Table 2 Explicit and Implicit Methods in the Multitrait-Multimethod Matrix Method Exp1 Exp2 Imp1 Imp2 Exp1 Exp2 Imp1 Method Trait nAch nPow Exp1 nAch 1.00 Exp2 nAch .80 1.00 Imp1 nAch .10 .10 1.00 Imp2 nAch .10 .10 .80 1.00 Exp1 nPow .30 .30 .00 .00 1.00 Exp2 nPow .30 .30 .00 .00 .80 1.00 Imp1 nPow .00 .00 .30 .30 .10 .10 1.00 Imp2 nPow .00 .00 .30 .30 .10 .10 .80 Imp2 1.00 The situation is somewhat analogous to the Implicit Assciation Test (IAT; Greenwald et al., 1998; see also Banaji & Greenwald, 2013), a reaction-time method which which is widely considered to be a measure of unconscious attitudes and prejudice. In fact, the correlation between IAT and self-report measures of attitude, such as an “attitude thermometer”, are relatively low: across the 56 attitude domains reviewed by Nosek, for example, the median correlation was .48 (Nosek, 2007). But again, this is not enough to establish the IAT as a measure of unconscious attitudes. Setting aside problems specific to interpretation of the IAT as a measure of attitudes (Arkes & Tetlock, 2004; Blanton et al., 2009; Kihlstrom, 2004b), it might be that implicit tests are simply unobtrusive measures which tap attitudes or motives of which the person is perfectly well aware -- but which, for whatever reason, he or she declines to Unconscious Motivation disclose to the experimenter (Webb, Campbell, Schwartz, & Sechrest, 1966). Supporting this hypothesis is the finding that, across 217 research reports surveyed in a recent meta-analysis, the external validity of the IAT was significantly higher in studies that found a relatively high implicit-explicit correlation (Kurdi et al., 2018). The explicit-implicit correlations involving the PSE are certainly lower than those typically observed with the IAT, which strengthens the case for dissociation, even if it does not clinch it. Thrash and his colleagues have proposed a general framework within which to explore the apparent incongruence between explicit and implicit motives (Thrash, Cassidy, Maruskin, & Elliot, 2010). For example, they sugest that if explicit and implicit motives have different antecedents, they will show little overlap; if they have the same antecedents, the overlap will be greater. The congruence between the two classes of motive measures can also be influenced by overlap in unidentified “random error” variables (e.g., a fire alarm that goes off during one of the tests), systematic “confounding error” variables (e.g., a third variable that is correlated with both the explicit and implicit measures), or, simply, observed indicators (e.g., response to the PSE, but not to a questionnaire, may be influenced by whether the subject likes to write stories). Although Thrash et al. begin their discussion with the assumption that explicit motives are consciously accessible while implicit motives are unconscious, consciousness is not necessarily an element in their framework. In fact, their framework is so general that it could be used to explore the relations between any two constructs, regardless of whether one is conscious and the other not. They do suggest that social desirability might be correlated with conscious motive expressions, but not unconscious Unconscious Motivation ones; but this could also differentiate between obtrusive and unobtrusive measures of conscious motives. Thrash et al. also point out that an aggregate correlation near zero can obscure the existence of subgroups for whom the variables in question are quite highly correlated. Accordingly, some studies have attempted to identify dispositional variables which might moderate the correlation between explicit and implicit measures of nAch (for full reviews, see Thrash et al., 2010; Thrash, Maruskin, & Martin, 2012). In a study by Thrash and Elliot (2002), for example, the correlation was -.07 for subjects low in self-determination and .40 for subjects high in self-determination. A later study found greater explicit-implicit congruence in individuals who are high in private body consciousness or preference for consistency, and in those who are low in selfmonitoring (Thrash, Elliot, & Schultheiss, 2007). Schattke et al. found higher explicitimplicit congruence for adults who, as children, had experienced higher levels of autonomy and relatedness (Schattke, Koestner, & Kehr, 2011). In establishing unconscious motivation, though, it is important that the variables moderating the explicit-implicit dissociation clearly relate to consciousness. Individuals high in private body consciousness might be more aware of their internal motivational states, leading to greater congruence between explicit and implicit motivation – but, perhaps, only if those motives are embodied in the way that hunger and thirst are. Selfmonitoring and a preference for consistency do not, intuitively, bear on consciousness, and the moderating effect of these variables is also consistent with an interpretation of PSE as an unobtrusive measure of conscious motivation. Unconscious Motivation It could also be the case that the implicit and explicit measures tap two quite different constructs, or at the very least nonoverlapping aspects of the same construct. For example, the PSE could assess aspects of nAch which are missed by questionnaires, or vice-versa. Schultheiss and his colleagues addressed this possibility by writing new questionnaire items for nAch which precisely parallel the coding categories of the PSE. In a study by Thrash and colleagues, the correlation between PSE and a standard questionnaire measure of nAch was .00; substituting the matchedcontent PSE-based questionnaire raised the correlation with PSE to .17 (Thrash et al., 2007). Correlations of this magnitude are not very high, of course – which strengthens the case that the low correlations observed between explicit and implicit measures of motivation are not an artifact of differences in construct definition. A similar point could be made with respect to differential validity. The PSE predicts actual behavior on motive-relevant tasks, establishing its validity as a motive measure, but McClelland et al. (1989) argued that the dissociation between explicit and implicit motivation was supported by the fact that the two sorts of measures predicted performance on very different kinds of tasks. In brief, implicit nAch predicts performance on “operant” tasks but not “respondent” tasks, especially under conditions of incentive, while the reverse is true for explicit motives tasks (Schultheiss & Brunstein, 2010; Spangler, 1992). On the surface, this double dissociation is very compelling – although the correlations in question are not strong even by the standards of personality research. On the other hand, for the most part the validation tasks employed in these studies have little to do with consciousness per se. Therefore, while the PSE and questionnaires may predict different kinds of motivated activity, that does not Unconscious Motivation necessarily mean that the PSE taps motives which are inaccessible to conscious awareness. Perhaps the greatest impediment to accepting the dissociation between explicit and implicit motives is that the tests that measure them are so different. The ACL, EPPS, and PRF are self-report questionnaires, while the PSE involves the coding of storytelling behavior. The lack of correlation between them, then, may be as much a matter of method variance as anything having to do with consciousness. In the study by Schultheiss et al. (2009), for example, the within-method correlations among PSE measures of nAch, nPow, and nAFF were higher (.19-.31) than the heteromethod correlations involving the explicit and implicit measures of the same motives. None of the correlations in this study were very high, but the fact that, in aggregate, the correlations among PSE measures of different motives were the highest obtained is cause to worry about the degree to which the entire pattern of results is contaminated by method variance. In the cognitive domain, the most convincing explicit-implicit dissociations are found in the domains of memory and perception, where investigators take great care to be sure that the cues presented to the subjects remain constant between the explicit and implicit tests, which vary only in their demands for conscious processing (Graf, Squire, & Mandler, 1984). In an experiment on implicit memory, for example, subjects might study a word such as ashcan. After a retention interval, they might be presented with a word stem such as ash___ and asked to complete the stem with a word from the studied list – a task requiring conscious recollection; or they might be asked simply to complete the stem with the first word that comes to mind – a task that does not require Unconscious Motivation episodic memory at all. In Jacoby’s (1991) PDP procedure, conscous recollection and unconscious priming are separated by asking subjects to complete the stem with a word that did not appear on the study list; if the studied word appears anyway, Jacoby infers that this is due to unconscious, automatic priming. In any case the cues remain constant, and the tasks vary in terms of whether they require consciousness or not. Cue-matching is not a problem unique to the PSE. The IAT has the same problem, as does implicit learning, and the problem has the same effect: it undercuts the conclusion that the motives assessed by the PSE are really unconscious in the sense that implicit percepts and memories are. Some progress in this direction was achieved by Schultheiss et al. (2009). These investigators asked subjects to imagine themselves to be one of the characters in each of the PSE stimuli, and then respond to a series of questions bearing directly on the PSE coding categories (Schultheiss, Yankova, Diritkov, & Schad, 2009). Scores for nAch, nAff, and nPow derived from this “PSE Questionnaire” (PSE-Q) showed little overlap with those derived from the standard PSE, with correlations ranging from .11 to .18. In a later study, employing depressed patients and adult controls instead of college students, the explicit-implicit correlations were higher but still generally low, ranging from -.25 to .24 (Neumann & Schultheiss, 2015). Then again, in the earlier study, PSE-Q scores also showed little overlap with explicit questionnaire measures of these motives, with correlations ranging from .08 to .23 (Schultheiss et al., 2009). On these grounds, it is hard to identify PSE-Q with explicit motivation – which leaves us with not two but three quite different motive assessments, none of which correlate very highly with each other. Maybe these instruments are measures of different constructs after all. Unconscious Motivation Alternative Assessments of Implicit Motivation Research on implicit motivation has now matured to the point that there are several alternative assessments of implicit motivation available for researchers to use. In addition to the standard PSE, there are several variants on the PSE. In the Operant Motive Test (OMT), for example, subjects are presented with a motive-evoking picture, as in the standard PSE; but instead of writing a story, they answer a series of four questions about one of the main characters (Runge et al., 2016). In the Multi-Motive Grid (MMG), subjects also view pictures, but then answer specific questions about what is going on in the minds of the characters (Sokolowski, Schmalt, Langens, & Puca, 2000). These variants retain the essential features of the PSE: subjects are presented with pictures, but are not asked about their own motives. A comparative study of nAch, nPow, and nAff found that all three implicit motivation measures showed very low correlations with three explicit measures, with a median r = .05; unfortunately, the intercorrelations among the three implicit measures were hardly better, median r = .09. By contrast, the three explicit measures showed consistently significant (if still low) intercorrelations, median r = .31 (Schüler, Brandstätter, Wegner, & Baumann, 2015). A verbal version of the MMG, in which brief vignettes substituted for the pictures, showed generally nonsignificant correlations with questionnaire measures (median r = .18); this figure is only slightly higher than that obtained with the standard pictorial version (median r = .11); unfortunately, the between-groups design of the study did not permit assessment of the relationship between the two forms of the MMG (Krumm, Schäpers, & Göbel, 2016). Unconscious Motivation As an alternative to TAT-like assessments of implicit motivation, Brunstein and Schmitt (2004) developed a version of the IAT to measure nAch: its scores were uncorrelated r = -.07) with scores on an explicit self-rating measure. Three subsequent studies confirmed essentially zero correlations between IAT scores and self-ratings; on the down side, however, they found only modest correlations between IAT and PSE, (median r = .27; Brunstein & Schmitt, 2010). One factor that might account for this lack of convergence between implicit motive measures is a difference in format: the PSE employs pictorial stimuli, whereas the stimuli in the IAT are verbal. Accordingly, Slabbinck and his colleagues introduced “pictorial-attitude” versions of the IAT (PA-IAT), in which subjects make judgements concerning motive-relevant situations (Slabbinck, De Houwer, & Van Kenhove, 2011). In a comparative study, Slabbinck et al. (2011) found that the PA-IAT measure of nPow correlated significantly (r = .31) with the PSE measure, but was essentially uncorrelated with questionnaire measures of dominance and aggression. Following up on a factor-analytic study by King (1995) Blisky and Schwartz (2008) employed the multitrait-multimethod matrix (Campbell & Fiske, 1959) to explore the relationships among TAT-like implicit and questionnaire-like explicit measures of nAch, nPow, and nAff. Multidimensional scaling applied to the data of five previously published studies (including King’s) revealed that explicit and implicit assessments of the same motive lined up together, providing evidence of significant shared trait variance between explicit and implicit methods despite near-zero correlations beween implicit and explicit measures. Like King, they argued that the various methods of motive assessment could be arrayed along a continuum from more vs. less explicit Unconscious Motivation (e.g., self-ratings vs. questionnaires), to less vs. more implicit (e.g., autobiographical memories vs. TAT/PSE). It is not yet known how the IAT fits into this scheme. Can Motivation Be Unconscious? Within the domain of cognition, there is now widespread agreement that episodic memories can be unconscious; so can percepts, and so can thoughts. All are revealed by priming effects observed in carefully controlled experiments. Given the evidence for the cognitive unconscious, it seems reasonable to ask whether the explicit-implicit distinction can be extended beyond cognition to emotion and motivation. It appears that motives can be activated automatically, and unconsciously, by relevant environmental cues, even when these cues themselves are processed outside conscious awareness. However, this claim needs to be supported by research which employs more rigorous criteria for automaticity; and which assesses the differential contributions of automatic and controlled processes to task performance. Similarly, initial results support the idea that implicit, motives, as assessed by techniques such as the PSE, are uncorrelated with explicit motives, as assessed by self-report questionnaires. Currently, however, evidence is lacking that the implicit motives assessed by the PSE are really inaccessible to conscious awareness, in the same sense that implicit percepts and memories are. There is no reason to think that motives and goals cannot be unconscious, and nevertheless influence experience, thought, and action; but further research is needed to make the case convincingly. In particular, it is important to show that implicit motive assessments are uncorrelated with explicit measures of the same motives, employing assessments which are as closely matched in their cue value as Unconscious Motivation possible, differing only in their demands for conscious awareness of the motive in question. At this point, however, the status of unconscious motivation remains: plausible on conceptual grounds, but not yet proven.