Indirect Assessment of Attitudes with Response-Time-Based Measures

Z. Soz. psychol. J. Degner 37 (3) et al.: 2006 Indirect by Verlag RT-Based Hans Huber, Assessment Hogrefe of Attitudes AG, Bern Zeitschrift für Sozialpsychologie, 37 (3), 2006, 131 139 Beitrag zum Themenheft Indirect Assessment of Attitudes with Response-Time-Based Measures Chances and Problems Juliane Degner 1, Dirk Wentura 1 and Klaus Rothermund 2 1 Universität des Saarlandes, 2 Friedrich-Schiller-Universität Jena Abstract: We review research on response-latency based ( implicit ) measures of attitudes by examining what hopes and intentions researchers have associated with their usage. We identified the hopes of (1) gaining better measures of interindividual differences in attitudes as compared to self-report measures (quality hope); (2) better predicting behavior, or predicting other behaviors, as compared to self-reports (incremental validity hope); (3) linking social-cognitive theories more adequately to empirical research (theory-link hope). We argue that the third hope should be the starting point for using these measures. Any attempt to improve these measures should include the search for a small-scale theory that adequately explains the basic effects found with such a measure. To date, small-scale theories for different measures are not equally well developed. Keywords: indirect attitude measures, affective priming, IAT, Affective Simon Task Indirekte Einstellungsmessung mit reaktionszeitbasierten Verfahren Chancen und Probleme Zusammenfassung: Dieser Beitrag bietet eine Übersicht über reaktionszeitbasierte («implizite») Verfahren der Einstellungsmessung, orientiert an der Frage, mit welchen Absichten und Hoffnungen diese Methoden entwickelt und eingesetzt wurden und werden. Dabei stehen drei Hoffnungen im Mittelpunkt: (1) die Hoffnung auf alternative bzw. bessere Maße zur Erfassung interindividueller Unterschiede in Einstellungen verglichen mit Selbstberichtmaßen; (2) die Hoffnung, auf bessere Verhaltensvorhersagen bzw. Vorhersagen anderer Verhaltensaspekte als mit Selbstberichtverfahren; (3) die Hoffnung auf eine direktere Anbindung sozial-kognitiver Theorien an empirische Forschung. Wir argumentieren, dass dieser letzte Aspekt die Basis der Anwendung dieser Verfahren sein muss. Jeder Versuch der Weiterentwicklung und Verbesserung dieser Methoden sollte die Suche nach Theorien zur Erklärung der zugrunde liegenden Prozesse beinhalten. Gegenwärtig sind solche Prozessmodelle nicht für alle Verfahren gleichermaßen gut entwickelt. Schlüsselwörter: indirekte Einstellungsmessung, affektives Priming, IAT, Affektive Simon Aufgabe In recent years, there has been an increase in the development and use of response-latency based attitude measures. These measures have in common that they claim to assess the strength of associations between attitude objects and evaluations in memory, thereby providing an estimate of attitudes without having to directly ask the participant for a verbal report. There is a wide range of such measures used in attitude research (see Fazio & Olson, 2003, for an overview); the most prominent of which are the Affective Priming task (AP; Fazio, Sanbonmatsu, Powell & Kardes, 1986; Fazio, Jackson, Dunton & Williams, 1995), the Affective Simon Task (AST; De Houwer & Eelen, 1998), the Extrinsic Affective Simon Task (EAST; De Houwer, 2003a), and the Implicit Association Test (IAT, Greenwald, McGhee & Schwartz, 1998). DOI 10.1024/0044-3514.37.3.131 We want to discuss the chances and problems associated with these measures by asking which expectations are linked to them, and if and how these prospects are met by the discussed paradigms. The current prevalence and popularity of these measures, especially in social cognition research on attitudes, may have resulted from the long and enduring quest for attitude measures that are less prone to distortion than are self-report measures (the quality hope). Associated with this expectation was the further hope that these measures would do a better job in predicting attituderelated behaviors, which could not or only poorly be predicted by self-reports (the incremental validity hope). Furthermore, and most notably, these measures offer a more direct access to the cognitive structures and processes that are postulated in social-cognitive theories (the theory-link

132 J. Degner et al.: Indirect RT-Based Assessment of Attitudes hope). However, researchers often use these paradigms without caring too much about the explanation of underlying processes. We will argue that the formulation of some kind of small-scale theories that adequately explain the basic effects found with these measures must be the groundwork for using and improving them. First of all, however, we will provide a description of the three basic paradigms that were suggested to measure automatic evaluations. The Affective Priming Task. The Affective Priming Task was introduced by Fazio and colleagues (1986; for reviews see Klauer, 1998; Klauer & Musch, 2003; Wentura & Rothermund, 2003). The participant s task is to categorize positive and negative target items with regard to their valence (e.g., press one key for positive and another key for negative words). Preceding each target, a prime stimulus is presented briefly (e.g., 200 ms). Typically, participants respond faster and more accurately if prime and target are congruent in valence than if they are incongruent in valence (Fazio et al., 1986). This technique works even if the prime is presented very briefly (e.g., 40 ms) and is immediately replaced by a mask, so that participants cannot identify the prime event (e.g., Draine & Greenwald, 1998). 1 The Affective Simon Task. The Affective Simon Task (AST; De Houwer & Eelen, 1998) and its variants (e.g., Beckers, De Houwer & Eelen, 2002; De Houwer, 2003a; Voss, Rothermund & Wentura, 2003) follow the basic principles of the spatial Simon Task, a paradigm well known in cognitive psychology (e.g., Simon, 1990). In a typical affective Simon study, participants have to select a positive or a negative response on the basis of a non-affective stimulus feature, while ignoring stimulus valence. For example, participants have to respond positive if the stimulus word is a noun and respond negative if the stimulus is an adjective (De Houwer & Eelen, 1998). Typically, responses are faster and more accurate if the valence of the response is congruent to the (to-be-ignored) valence of the stimuli (e.g., De Houwer & Eelen, 1998; De Houwer, Crombez, Baeyens & Hermans, 2001; Tipples, 2001). For a more practicable assessment of attitudes, De Houwer (2003a) introduced a variant, termed the Extrinsic Affective Simon Task (EAST). In a typical EAST, two classes of stimuli (e.g., white words and colored words) have to be classified according to their color or their meaning. In a first practice block, participants see only white words with a clear positive or negative meaning, which they have to categorize on the basis of their valence by pressing either a positive or negative key. In a further phase, attitude-related stimuli are presented in either a greenish or bluish color. Participants are instructed to ignore the meaning of the stimuli and to respond according to their color. The responses are assigned to the same keys as in the evaluation task, for example, blue to the positive key and green to the negative key. In the test phase white and colored words are presented alternately, and participants are instructed to classify white words according to their valence and colored words according to their color, using the same keys. Thereby, each attitude-relevant stimulus demands equally often positive and negative responses respectively in one experimental block. If the colored stimuli are associated with valence, responses are faster and more accurate if the correct response is associated with the same valence as the valence of the stimulus (e.g., De Houwer, 2003a; Stahl & Degner, in press). The Implicit Association Test. The Implicit Association Test (IAT; Greenwald et al., 1998) is at present the most popular and widely used measure (e.g., Fazio & Olson, 2003; Hofmann, Gawronski, Geschwender, Le & Schmitt, 2005). In a typical IAT study, stimuli are presented one by one on the computer screen, e.g., positive and negative words (attribute stimuli), and names of insects and flowers (target stimuli). In a dual categorization task, participants have to classify attributes with regard to their valence (positive vs. negative), and targets with regard to their category membership (insect vs. flower) by pressing one of two response keys. Two categories are always assigned to the same response key; for example, positive and flower and negative and insect share the same keys. In a second block, the assignment of one category dichotomy to the keys is switched (i.e., positive and insect vs. negative and flower ). Participants typically react faster and more accurately if categories that match in valence share a key (congruent block; e.g., flowers and positive words vs. insects and negative words) than if categories that mismatch according to their valence share a key (incongruent block; Greenwald et al., 1998). The IAT-score is computed by subtracting mean response latencies of the congruent block from those of the incongruent block (for an alternative scoring algorithm, see Greenwald, Nosek & Banaji, 2003). This score is interpreted as representing the relative association strength between target and attribute categories. The IAT is not restricted to evaluative associations (see, e.g., Asendorpf, Banse & Mücke, 2002, or Egloff & Schmuckle, 2002, for applications in personality research). In this paper, however, we focus exclusively on its use in attitude research. Recently, variations or first-grade relatives of the IAT have been developed, for example, the Go/No-Go Association Task (GNAT; Nosek & Banaji, 2001) or the Single Target IAT (ST-IAT; Wigboldus, Holland & Van Knippenberg, 2005). These try to measure associations to single concepts without requiring complementary or contrasting categories as in the original IAT. 1 The basic design of orthogonally varying prime and target valence was also employed in other studies, in which the evaluation task was replaced by a naming task (i.e., the target word has to be pronounced as quickly as possible; De Houwer & Randell, 2004; Spruyt, Hermans, Pandelaere, De Houwer & Eelen, 2004), by a lexical decision task (i.e., the target has to be categorized as a word or a non-word as fast as possible; Hermans, Smeesters & De Houwer, 2002), or by a semantic categorization task (i.e., targets have to be classified as, e.g., persons or objects; De Houwer, Hermans, Rothermund & Wentura, 2002). However, these manipulations result in a completely different paradigm (see, e.g., Wentura & Rothermund, 2003).

J. Degner et al.: Indirect RT-Based Assessment of Attitudes 133 The rationales behind adapting response-latency paradigms to attitude research are manifold. The current great interest in these measures can partly be explained by the prospect to gain alternative approaches to the assessment of attitudes as compared to questionnaires or other direct measures. Using indirect RT-based measures is seen as one potential avenue to assess attitudes while decreasing the risks of interfering influences well known to social psychologists, such as social desirability concerns, intentional faking, (self-) deception, and self-presentational strategies. Because of the indirect nature of these measures, researchers hope that participants are not aware of which attitude is measured, that they do not understand the functioning of the measure, and that therefore they are neither motivated nor able to strategically manipulate or fabricate effects. Can these hopes be appropriately fulfilled by using indirect response-latency based attitude measures? In the following section we will first review empirical evidence on the unobtrusiveness of indirect response-latency based attitude measures. Furthermore, we will give a short overview of the discussion on what is actually assessed with these tools. The Quality Hope Susceptibility to Strategic Control The most intransparent measure from the participants perspective is an Affective Priming task with a masked ( subliminal ) presentation of primes (De Houwer, 2006). If participants are not aware of the priming event, it is likely that they are also unaware of the measurement s purpose. Thus, the unobtrusiveness of this instrument cannot be questioned (Klauer, Mierke & Musch, 2003). This, however, cannot be assumed for any of the other presented measures. It could be argued that participants might guess the purpose of any study if the attitude-related stimuli are presented in such a way that they are clearly visible. This questions the unobtrusiveness of these paradigms but does not necessarily permit conscious control of the measurements outcome. For example, several studies have shown that strong Affective Priming effects are obtained with short time-lags (e.g., 300 ms) between onset of prime and onset of target (stimulus onset asynchrony, SOA), but disappear with longer SOAs (e.g., Hermans, De Houwer & Eelen, 2001; Fazio et al., 1986; Klauer, Roßnagel & Musch, 1997). This provides evidence that the effect is caused by automatic processes, because time is too short for intentional control. However, Klauer et al. (1997) found the short-soa effect to be dependent on the relatedness proportion (i.e., the ratio of congruent to incongruent prime-target pairs), a finding typically seen as an indicator of strategy use. Given that there are no studies we know of that have systematically tested influences of manipulation intentions on the outcome of priming tasks, it is difficult to conclude if or to which extent the Affective Priming task is prone to strategic control. Likewise, there are no published studies that have tested the susceptibility of the AST or the EAST to strategic control. Given the procedural similarity of IAT and EAST (at least from the participant s perspective), one might consider them equally (in-)transparent and equally (un-)controllable. Yet, this claim remains to be empirically tested. Although the purpose and functioning of an IAT is somewhat easy to figure out (Monteith, Voils & Ashburn- Nardo, 2001) inexperienced participants do not seem to intuitively come up with the correct strategies to successfully control its outcome (e.g., Asendorpf et al., 2002; Banse, Seise & Zerbes, 2001). They are nevertheless able to follow instructions to reduce IAT-effects by slowing down responses in congruent trials (Kim, 2003; Steffens, 2004). More importantly, Fiedler and Bluemke (2005) found that, after prior experience with only a single IAT, participants were able to think of and implement faking strategies that were not detectable. Studies that explore socially sensitive issues (e.g., racial prejudice) often use cover stories to assure that participants do not gain insight into the purpose of the study (e.g., Fazio et al., 1995). A plausible cover story can decrease the likelihood of participants developing the motivation or intention to manipulate the outcome of the measurement. However, more systematic research on the controllability of the measurements outcome is needed before firm conclusions can be drawn. Nevertheless, it seems appropriate to assume that indirect attitude measures are less voluntarily controllable than questionnaire measures (at least for inexperienced participants). Yet, that does not imply that they are not susceptible to influences of social desirability at all. As Blair (2002) states, automatic activation processes are influenced by contextual factors of the experimental situation, such as the activation of social and self-related motives. Even if participants might not intentionally manipulate the measure, context features can still influence its outcome in important ways (see Lowery, Hardin & Sinclair, 2001, for an example). Two Different Types of Attitudes? It is a subject of considerable controversy whether these methods assess attitudes that are conceptually different from those assessed with self-report measures, or whether the different kinds of measures tap the same concepts in different ways. In this controversy, the terms implicit, unconscious, or automatic in contrast to explicit, conscious, and controlled very often come into play, establishing a seemingly clear-cut conceptual distinction between two different types of attitudes. Greenwald and Banaji (1995), for example, stated that so called implicit and explicit attitudes differ in their introspective accessibility, often redefined as consciousness. Wilson, Lindsay, and Schooler (2000) define implicit and

134 J. Degner et al.: Indirect RT-Based Assessment of Attitudes explicit attitudes as two different and independent evaluations of the same object: One that is old, well-learned, and therefore habitual and automatic, and the other that is more recently constructed and represents deliberate beliefs (for a similar perspective see Devine, 1989). Other researchers take a different perspective and see indirect response-latency based attitude measures as approaching the same structures in memory as self-report measures. From this perspective, so called implicit and explicit attitudes can be best understood as different effects of one and the same construct on different measures. Thus any attitude that is stored in memory can (but does not have to) be activated automatically and / or outside one s awareness or control, which does not obligatorily imply that the attitude is unconscious, unidentifiable, or unknown to the person (Dambrun & Guimond, 2004; Fazio, 1990; Fazio & Towles-Schwen, 1999). Both perspectives were empirically supported in correlational studies. On the one hand, meaningful correlations between different RT-based measures and questionnaires have been found repeatedly (e.g., Fazio et al., 1995; Mc- Connel & Leibold, 2001; Wentura & Degner, 2005), suggesting that both tap into the same construct. On the other hand, low or zero correlations between direct and indirect attitude measures have been reported in other studies (e.g., Gawronski, Geschke & Banse, 2003; Karpinski & Hilton, 2001), which has been interpreted as evidence that the constructs assessed by these measures diverge (e.g., Greenwald et al., 1998). In their meta-analysis on the IAT, Hofmann and colleagues (2005) report a small but significantly positive mean population correlation of.19 (.24 after correcting for measurement error) between self-reports and IAT effects (including 126 studies). This correlation increased as a function of increasing spontaneity of self-report and conceptual correspondence between measures. The latter finding supports the dependency perspective: Direct and indirect measures of attitudes might tap the same mental representations. To date there is no agreement whether there are different types of attitudes, and if so, whether they draw on the same or dual mental representations in memory, and whether they are dissociable or consistent in content. Nevertheless, there is one important overlap of the two presented perspectives. Researchers do agree that attitudes can differ in their mode of operation: They can operate automatically (differing in the degree of efficacy, effort, consciousness, and/or controllability; De Houwer, 2006), or be included in a controlled, deliberative decision, considering all available information on the issue in question. Under specific circumstances, a person might indeed be prompted to deliberately express attitudes that differ from automatically activated evaluations, but that does not necessarily imply their conceptual dissociation. Different models like the Motivation and Opportunity as Determinants (MODE)-Model (Fazio, 1990; Fazio & Towles-Schwen, 1999) or the Reflective- Impulsive Model (RIM; Strack & Deutsch, 2004) try to clarify how and under which conditions attitudes influence social behavior via these different modes of operation. Despite the present ambiguity about what RT-based measures actually assess, their application offers valuable complements to classical attitude research, in that they might provide a different perspective on the attitude-behavior link. The Incremental Validity Hope Prediction of Behavior Explaining, predicting (and changing) people s behavior has always been a major focus of social psychological theory and research. Thereby, attitudes have at all times held a dominant role in any behavior model (e.g., Ajzen, 2001). However, behavioral predictions often fail when based on self-reported attitudes. The decision to use indirect RTbased measures such as the IAT, the EAST, or the Affective Priming task is closely tied to the hope of improving behavior predictions. Can the accuracy of behavior predictions be increased? Can different types of behavior be predicted by including attitude measures that tap the automatic aspects of attitudes? The answer to these questions is by and large affirmative. It is one main potential advantage of indirect RT-based measures that they enable us to register the automatic impact of attitudes on behavior (De Houwer, 2006; Fazio & Olson, 2003). In several studies, it has been shown that response-time based attitude measures were related to automatic or spontaneous behavior (or behavior that is not seen as an expression of an attitude by the actor; Wilson et al., 2000), whereas questionnaire measures were better predictors for more deliberative behavioral decisions. We will give some examples of the three classes of measures, most of them dealing with the assessment of negative prejudice and prediction of race-related interaction behaviors. Fazio and colleagues (1995) were the first to demonstrate that the Affective Priming measure predicted subtle behavior more accurately than self-report measures of racial prejudice. The friendliness and interest of participants toward a black experimenter was correlated with an Affective Priming effect, but not with self-reported prejudice. However, explicit judgments of race-related items were correlated to the questionnaire but not to the priming data. Similar findings were obtained with a subliminal version of the Affective Priming task assessing racial attitudes (Degner, Wentura, Gniewosz & Noack, 2006; Dovidio et al., 1997) and with a race-iat (McConnel & Leibold, 2001). In an unpublished review, Poehlman, Uhlmann, Greenwald, and Banaji (2005) showed that IAT measures were similarly effective at predicting responses that are difficult versus easy to control. In contrast, explicit measures were strongly affected by the controllability of the criterion measure: They were good predictors of highly controllable

J. Degner et al.: Indirect RT-Based Assessment of Attitudes 135 behaviors, but weaker predictors of spontaneous behaviors. Furthermore, they reported that the predictive validity of the IAT was not affected by the social sensitivity of the domain, whereas predictability with explicit measures decreased with increasing influence of social desirability. Especially in the stereotyping and prejudice domain it was shown that the IATs were significantly better predictors than were explicit measures. Psychometric Qualities Of course, in order to accept these measures as alternatives to questionnaires in predicting behavior, several criteria have to be met regarding their psychometric properties, such as internal consistency and re-test reliability. However, reliabilities of Affective Priming tasks as well as Affective Simon Tasks are generally not satisfying. Cronbach s αs and split-half reliabilities are typically lower than.60 in supraliminal as well as subliminal priming tasks (e.g., Bosson, Swann & Pennebaker, 2000; De Houwer, 2003b; Wentura & Degner, 2005, Voss et al., 2003; but see Cunningham, Preacher & Banaji, 2001 and Perugini, O Gorman & Conner, 2005, for αs ranging from.52 to.74). Reported test-retest reliabilities are often low, too (e.g., r s ranging from.08 to.28; Bosson, et al., 2000). Nevertheless, when correcting for measurement error, Cunningham and colleagues (2001) found stability estimates of.83 for a supraliminal priming task. Low internal consistency is an unresolved matter in both the EAST and in the affective Priming paradigm. In contrast, reliabilities of IAT measures are satisfying in terms of internal consistency with Cronbach s αs typically ranging between.80 to.90, and test-re-test correlations ranging from.50 to.70 (e.g., Banse et al., 2001; Bosson et al., 2000; Cunningham et al., 2001; Gawronski et al., 2003). Importantly, however, high internal consistencies (a =.95) and substantial retest correlations (r tt =.66) were also observed with artificial variants of the IAT that were unrelated to attitudes (geometry IAT, Mierke & Klauer, 2003, Exp. 1A). These findings should be taken as a caveat not to interpret reliability coefficients in the IAT as a measure of systematic interindividual variance related to attitudes. Instead, internal consistencies in the IAT might also reflect interindividual differences with regard to cognitive control abilities. In general, reliability estimates of response-latency measures are generally lower than is typically the case for selfreport measures (Cunningham et al., 2001; Fazio & Olson, 2003), a fact that somewhat questions the applicability of these paradigms as measures of interindividual differences. It should be noted, however, that meaningful correlations to questionnaires and behavioral measures are often found despite low indices of internal consistency. It might be worthwhile to think about more sophisticated ways to calculate estimates of reliability, which might be better suited for RT-based measures. The Theory-Link Hope At this point, we take a step back and discuss the different measurement procedures from a more theoretical view. How do indirect response-latency based attitude measures function compared to self-report measures? We have to be aware of the fact that indirect, RT-based attitude measures focus on a completely different theoretical level than selfreports; we might call it a functional or sub-personal level. Without requiring the knowledge, intention, good will, or effort of the participants to disclose personal preferences or dislikes, these measures enable the drawing of conclusions from observed behavior (response latencies and error rates) on cognitive structures and processes. However, these conclusions rely on small-scale theories that link the basic effect found with the measures (e.g., the RT-difference between congruent and incongruent trials in an Affective Priming task) to the large-scale theoretical level (e.g., Fazio s conceptualization of strong attitudes, i.e., an association of attitude object and evaluation in long-term memory). It is the most important advantage of indirect response-latency based measures to directly link the concept of attitudes to social-cognitive theories. In order to do so, however, any measure must be associated with a valid small-scale theory that explains the underlying processes through which the construct attitude produces and/or determines the outcome of the measurement procedure (De Houwer, 2006). Considering the underlying processes cannot only provide information about the focus and functioning of the different measures and their similarity to other measures (Fazio & Olson, 2003), but can also be crucial to the further development of large-scale theories on the attitude in question. Unfortunately, very often these measures are used and outcomes are interpreted without an examination of the underlying processes. In the following section we will review the current small-scale theories that have been developed to explain the underlying processes of the measures discussed in this article. The Affective Priming paradigm. To date, the processes that drive Affective Priming effects are considered as fairly well understood (De Houwer, 2003b; Fazio & Olson, 2003; Klauer, 1998; Klauer & Musch, 2003). Interestingly, although the Affective Priming paradigm was initially developed as a tool for social-cognitive research on attitudes, it has been rigorously utilized for years in other domains. The most dominant question in this context was: What kind of process mediates between the activated prime valence and the target response? Should we assume a process of spreading activation between the mental representation of the prime, the representation of its valence, and all other concepts that share valence with the prime? We do not want to recapitulate this discussion at this point (for reviews, see Fazio, 2001; Klauer & Musch, 2003; Wentura & Rothermund, 2003), but it seems fair to conclude that Affective Priming (in the evaluation task) can be explained by a rath-

136 J. Degner et al.: Indirect RT-Based Assessment of Attitudes er simple process (e.g., De Houwer & Hermans, 1994; Klauer et al., 1997; Wentura, 1999). The prime triggers (via the activation of its valence) either the correct response (if the target is congruent) or the wrong one (in the case of an incongruent target). In incongruent trials, additional evidence of the target valence is needed to arrive at the correct response as compared to congruent trials (e.g., Fazio & Olson, 2003); consequently responses take longer than in congruent trials. Thus Affective Priming effects in an evaluation task can only be explained if one assumes that the affective valence of the primes is processed, even though this is not necessary for the task at hand (Hermans, Spruyt, De Houwer & Eelen, 2003). For the large-scale theory of the mental representation of attitudes and Fazio s concept of strong attitudes (Fazio, et al., 1986), the small-scale theory of response priming is all that is needed. Affective Simon Task. A similar argument applies to the affective Simon task. Although valence of stimuli is a taskirrelevant stimulus feature, it systematically influences speed and accuracy of valence-related responses. Thus, the valence of stimuli must be processed first. The underlying rationale is the same as for the Affective Priming paradigm: The automatic activation of stimulus valence automatically triggers a matching response. This facilitates responding if the valence of stimulus and response are congruent. In the case of incongruence, the misleading activation of the stimulus valence has to be overcome to give a correct response (De Houwer & Eelen, 1998). Because effects also emerge if participants are explicitly instructed to ignore stimulus valence (e.g., De Houwer & Eelen, 1998, Study 2), it can be inferred that the processing of valence occurs automatically, that is, unintentionally and uncontrollably (see Bargh, 1992). It is assumed that the affective connotation of stimuli is automatically retrieved from memory, which in turn leads to an activation of the corresponding evaluative response tendencies. This explanation is directly deduced from theories on the spatial Simon effect (see Simon, 1990). For the Extrinsic Affective Simon Task (EAST), one additional assumption has to be made: It is assumed that the a priori neutral responses (i.e., a left or right key press) acquire the connotation of positivity or negativity during the task. This is attained by reactions on the evaluation trials that are interspersed among the Simon trials (i.e., the trials with attitude related stimuli). However, recently, alternative approaches have been developed that question the response-incompatibility explanation of the EAST and thus its relation to the classical Simon paradigm (Degner & Wentura, 2006; Voss & Klauer, in press). The Implicit Association Test. Greenwald and colleagues (1998) have proposed that the IAT is a measure of associations between concepts: Responding in the IAT should be facilitated if associated concepts are assigned to the same rather than to different responses. This postulate, however, has not been substantiated by a more detailed small-scale theory of the mediating processes that should translate associations into IAT effects. An influential account of these mediating mechanisms has been proposed by De Houwer (2001). According to him, the IAT is a response compatibility measure, much like the AST or EAST. Previously neutral key presses acquire short-term associations with either positive or negative valence during the compatible block of an IAT, because in these blocks, each response is linked to two categories sharing the same valence (e.g., assigning flower and positive to the right key implies that pressing the right key becomes associated with positive valence). In the compatible block, executing the required response (pressing the right key) is facilitated because the valence of the correct response is similar to the valence of the automatically activated evaluative response tendency. In an incompatible block, on the other hand, response keys do not acquire a clear valence because categories of opposite valence are assigned to the same key. This reduces facilitatory effects of response compatibility and might even lead to response interference. However, the association account of the IAT has recently been challenged by an alternative model. Rothermund and Wentura (2001, 2004) have proposed that salience-asymmetries are crucial for the emergence of IAT effects. In many IATs, the two categories that make up the target or the attribute dimension differ with regard to their salience. Previous research has shown that in a typical situation, novel, unfamiliar, or negative categories tend to be more salient as compared to categories that are familiar or positive (e.g., Johnston & Hawley, 1994; Pratto & John, 1991; Wentura, Rothermund & Bak, 2000). The members of the salient categories (e.g., insect and negative) thus create the impression of standing out against the background of the non-salient categories (e.g., flower and positive). According to the salience asymmetry account, IAT effects result from a consistent (vs. inconsistent) assignment of salient and non-salient categories to responses. Thus, assigning the salient categories to the same response in the compatible block facilitates responding because salience can be used to distinguish between responses. Rothermund and Wentura (2001, 2004; for related evidence, see Brendl, Markman & Messner, 2001; Mitchell, 2004) supported the salience asymmetry account by a series of experiments in which salience and associations were experimentally dissociated. The debate regarding the underlying processes of the IAT has direct implications for its use as a measure of attitudes. If IAT effects can be influenced by factors that are only loosely related to evaluations (e.g., familiarity, novelty, attentional focus), and if evaluations exert their influence on the IAT only indirectly by being confounded with salience, then the IAT cannot unquestionably be taken as a valid indicator of attitudes. In support of this conclusion, Rothermund and Wentura (2004) have shown that correlations between IAT effects and self-report measures can in fact be spurious: Controlling for interindividual differences in salience asymmetries reduced or eliminated the correlation between self-report measures and the IAT. The implications of these findings do not only concern the use of the IAT as a diagnostic tool for assessing attitudes. They also

J. Degner et al.: Indirect RT-Based Assessment of Attitudes 137 draw our attention to the role of salience asymmetries as a fundamental dimension that differentiates between social categories, and which might also play an important role in the development and maintenance of attitudes. It becomes clear, that although the tasks reported in this article share superficial similarities in their procedures, they are fundamentally different on a structural level (De Houwer, 2003b). They tap into different cognitive processes, which might be the cause of the often missing intercorrelations between different measures (e.g., Bosson et al., 2000; Olson & Fazio; 2003). The development and exploration of small-scale theories that explain the basic processes underlying the paradigms is still in progress. Researchers should pay attention to the basic research on these paradigms when using them. It cannot be overemphasized that the improvement of social cognitive theory-building hinges on the development and usage of experimental paradigms that are associated with a validated small-scale theory. Of course, in scientific progress any theory, thus also any small-scale theory, can be questioned. What follows is paradigm-oriented research that is business-as-usual in cognitive psychology. From the outside, it might appear as ivory tower research that does not help to solve the real scientific problems. In our view, this evaluation is a mistake. Final Considerations In this article we have given an overview on the most prominent RT-based attitude measures, focusing on the potential benefits these measures offer as compared to classical selfreport based attitude research. We investigated three main benefits of these measures. First of all, they assess attitudes in a way that is less prone to influences of social desirability, faking, deception, and self-presentational strategies than self reports. Decidedly the most unobtrusive measures in this respect are masked (subliminal) versions of the Affective Priming paradigm. This surplus makes these measures especially valuable for the assessment of socially sensitive issues, such as intergroup attitudes. However, the hope to assess the real core attitude (if one believes that such a stable concept exists) can not be fulfilled, because these tools do not measure anything independently of context effects (Blair, 2002). Second, these measures offer another access to predicting behavior, and enable the prediction of different behaviors, than do self-report measures. In this regard, all three classes of measures have been shown to do a good job especially in predicting spontaneous behavior not accessible by self-reports. The implementation of these tools into research expands theorizing on attitude-behavior-relations (e.g., the MODE-model by Fazio et al., the RIM by Strack & Deutsch, 2004). Third and most importantly, research with and on these measures allows a different perspective on the cognitive representation of attitudes and their modes of operation within our cognitive systems. In this regard, indirect RT-based attitude measures are much more than just valuable tools in complementing and possibly correcting self-reports. Their usage paves the way to an understanding of the cognition of attitudes. However, whereas cognitive psychology has a long tradition in putting forward large-scale theories (e.g., what is the structure of memory and what are the dominant processes?) as well as small-scale theories (e.g., what is the adequate explanation of semantic priming effects?), and most importantly of exploring the delicate relationship between those types of theories, social cognition seems to have somewhat neglected small-scale theories and has largely overlooked the fact that a given paradigm is not logically bound to only one small-scale theory. We suggest that the development and exploration of small-scale theories must be the starting point for using RT-based measures. Any attempts to using and improving these measures have to begin with a search for a theory that adequately explains their basic effects. Only if we get a clear picture of how these measures work, can we build up theories on what these measures actually assess. Acknowledgments The preparation of this article was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG) to Dirk Wentura (We 2284/3). References Ajzen, I. (2001). Nature and operation of attitudes. Annual Review of Psychology, 52, 27 58. Asendorpf, J.B., Banse, R., & Mücke, D. (2002). Double dissociation between implicit and explicit personality self-concept: The case of shy behavior. Journal of Personality and Social Psychology, 83, 380 393. Banse, R. (2001). Affective priming with liked and disliked persons: Prime visibility determines congruency and incongruency effects. Cognition and Emotion, 15, 501 521. Banse, R., Seise, J., & Zerbes, N. (2001). Implicit attitudes toward homosexuality: Reliability, validity, and controllability of the IAT. Zeitschrift für Experimentelle Psychologie, 48, 145 160. Bargh, J.A. (1992). The ecology of automaticity. Toward establishing the conditions needed to produce automatic processing effects. American Journal of Psychology, 105, 181 199. Beckers, T., De Houwer, J., & Eelen, P. (2002). Automatic integration of non-perceptual action effect features: The case of the associative affective Simon effect. Psychological Research, 66, 166 174. Blair, I. (2002). The malleability of automatic stereotypes and prejudice. Personality and Social Psychology Review, 6, 242 262. Bosson, J.K., Swann, W.B. Jr., & Pennebaker, J.W. (2000). Stalk-

138 J. Degner et al.: Indirect RT-Based Assessment of Attitudes ing the perfect measure of implicit self-esteem: The blind men and the elephant revisited? Journal of Personality and Social Psychology, 79, 631 643. Brendl, M., Markman, A., & Messner, C. (2001). How do indirect measures of evaluation work? Evaluating the inference of prejudice in the Implicit Association Test. Journal of Personality and Social Psychology, 81, 760 773. Cunningham, W.A., Preacher, K.J., & Banaji, M.R. (2001). Implicit attitude measures: Consistency, stability, and convergent validity. Psychological Science, 12, 163 170. Dambrun, M., & Guimond, S. (2004). Implicit and explicit measures of prejudice and stereotyping: Do they assess the same underlying knowledge structure? European Journal of Social Psychology, 34, 663 677. Degner, J., & Wentura, D. (2006). The Extrinsic Affective Simon Task as an instrument for indirect assessment of prejudice. Manuscript submitted for publication. Degner, J., Wentura, D., Gniewosz, B., & Noack, P. (2006). Can the masked Affective Priming task be used as an unobtrusive measure of prejudice in adolescents? Manuscript submitted for publication. De Houwer, J. (2001). A structural and process analysis of the Implicit Association Test. Journal of Experimental Social Psychology, 37, 443 451. De Houwer, J. (2003a). The extrinsic affective Simon task. Experimental Psychology, 50, 77 85. De Houwer, J. (2003b). A structural analysis of indirect measures of attitudes. In J. Musch & K.C. Klauer (Eds.), The psychology of evaluation: Affective processes in cognition and emotion (pp. 219 244). Mahwah, NJ: Erlbaum. De Houwer, J. (2006). What are implicit measures and why are we using them. In R. W. Wiers & A. W. Stacy (Eds.), The handbook of implicit cognition and addiction (pp. 11 28). Thousand Oaks, CA: Sage. De Houwer, J., Crombez, G., Baeyens, F., & Hermans, D. (2001). On the generality of the affective Simon effect. Cognition and Emotion, 15, 189 207. De Houwer, J., & Eelen, P. (1998). An affective variant of the Simon paradigm. Cognition and Emotion, 12, 45 61. De Houwer, J., & Hermans, D. (1994). Differences in the affective processing of words and pictures. Cognition and Emotion, 8, 1 20. De Houwer, J., Hermans, D., Rothermund, K., & Wentura, D. (2002). Affective priming of semantic categorization responses. Cognition and Emotion, 16, 643 666. De Houwer, J., & Randell, T. (2004). Robust affective priming effects in a conditional pronunciation task: Evidence for the semantic representation of evaluative information. Cognition and Emotion, 18, 251 264. Devine, P.G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 56, 5 18. Dovidio, J.F., Kawakami, K., & Johnson, C. (1997). On the nature of prejudice: Automatic and controlled processes. Journal of Experimental Social Psychology, 33, 510 540. Draine, S.C., & Greenwald, A.G. (1998). Replicable unconscious semantic priming. Journal of Experimental Psychology: General, 127, 286 303. Egloff, B., & Schmukle, S.C. (2002). Predictive validity of an implicit association test for assessing anxiety. Journal of Personality and Social Psychology, 83, 1441 1455. Fazio, R.H. (1999). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. Advances in Experimental Social Psychology, 23, 75 109. Fazio, R.H. (2001). On the automatic activation of associated evaluations: An overview. Cognition and Emotion, 15, 115 141. Fazio, R.H., Jackson, J.R., Dunton, B.C., & Williams, C.J. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69, 1013 1027. Fazio, R.H., & Olson, M.A. (2003). Implicit measures in social cognition research: Their meaning and uses. Annual Review of Psychology, 54, 297 327. Fazio, R.H., Sanbonmatsu, D.M., Powell, M.C., & Kardes, F.R. (1986). On the automatic activation of attitudes. Journal of Personality and Social Psychology, 50, 229 238. Fazio, R.H., & Towles-Schwen, T. (1999). The MODE model of attitude-behavior processes. In S. Chaiken & Y. Trope (Eds.). Dual-process theories in social psychology (pp. 97 116). New York: Guilford. Fiedler, K., & Bluemke, M. (2005). Faking the IAT: Aided and unaided response control on the Implicit Association Test. Basic and Applied Social Psychology, 27, 307 316. Gawronski, B., Geschke, D., & Banse, R. (2003). Implicit bias in impression formation: Associations influence the construal of individuating information. European Journal of Social Psychology, 33, 573 590. Greenwald, A.G., & Banaji, M.R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102, 4 27. Greenwald, A.G., McGhee, D.E., & Schwartz, J.L.K. (1998). Measuring individual differences in implicit cognition: The Implicit Association Test. Journal of Personality and Social Psychology, 74, 1464 1480. Greenwald, A.G., Nosek, B.A., & Banaji, M.R. (2003). Understanding and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197 216. Hermans, D., De Houwer, J., & Eelen, P. (2001). A time course analysis of the affective priming effect. Cognition and Emotion, 15, 143 166. Hermans, D., Smeesters, D., & De Houwer, J. (2002). Affective priming for associatively unrelated primes and targets. Psychologica Belgica, 42, 191 212. Hermans, D., Spruyt, A., De Houwer, J., & Eelen, P. (2003). Affective priming with subliminally presented pictures. Canadian Journal of Experimental Psychology, 57, 97 114. Hofmann, W., Gawronski, B., Gschwendner, T., Le, H., & Schmitt, M. (2005). A meta-analysis on the correlation between the Implicit Association Test and explicit self-report measures. Personality and Social Psychology Bulletin, 31, 1369 1385. Johnston, W.A., & Hawley, K.J. (1994). Perceptual inhibition of expected inputs: The key that opens closed minds. Psychonomic Bulletin and Review, 1, 56 72. Karpinski, A., & Hilton, J.L. (2001). Attitudes and the Implicit Association Test. Journal of Personality and Social Psychology, 81, 774 788. Kim, D.-Y. (2003). Voluntary controllability of the Implicit Association Test (IAT). Social Psychology Quarterly, 66, 83 96. Klauer, K.C. (1998). Affective priming. In W. Stroebe & M. Hew-

J. Degner et al.: Indirect RT-Based Assessment of Attitudes 139 stone (Eds.), European review of social psychology (Vol. 8, pp. 67 103). New York: Wiley. Klauer, K.C., Mierke, J., & Musch, J. (2003). The positivity proportion effect: A list context effect in masked affective priming. Memory and Cognition, 31, 953 967. Klauer, K.C., & Musch, J. (2003). Affective priming: Findings and theories. In J. Musch & K.C. Klauer (Eds.), Psychology of evaluation: Affective processes in cognition and emotion (pp. 7 49). Mahwah, NJ: Erlbaum. Klauer, K.C., Roßnagel, C., & Musch, J. (1997). List-context effects in evaluative priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 246 255. Lowery, B.S., Hardin, C.D., & Sinclair, S. (2001). Social influence effects on automatic racial prejudice. Journal of Personality and Social Psychology, 81, 842 855. McConnell, A.R., & Leibold, J.M. (2001). Relations among the Implicit Association Test, discriminatory behavior, and explicit measures of racial attitudes. Journal of Experimental Social Psychology, 37, 435 442. Mierke, J., & Klauer, K.C. (2003), Method-specific variance in the Implicit Association Test. Journal of Personality and Social Psychology, 85, 1180 1192. Mitchell, C.J. (2004). Mere acceptance produces apparent attitude in the Implicit Association Test. Journal of Experimental Social Psychology, 40, 366 373. Monteith, M.J., & Voils, C.I., & Ashburn-Nardo, L. (2001). Taking a look underground: Detecting, interpreting, and reacting to implicit racial bias. Social Cognition, 19, 395 416. Nosek, B.A., & Banaji, M.R. (2001). The Go/No-Go Association Task. Social Cognition, 19, 625 669. Olson, M.A., & Fazio, R.H. (2003). Relations between implicit measures of prejudice: What are we measuring? Psychological Science, 14, 636 640. Perugini, M., O Gorman, R., & Conner, M.T. (2005). Moderators of the predictive validity of implicit and explicit attitudes: Predicting sweets consumption. Unpublished manuscript. Poehlman, T.A., Uhlmann, E., Greenwald, A.G., & Banaji, M.R. (2005). Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. Unpublished manuscript. Pratto, F., & John, O.P. (1991). Automatic vigilance: The attention-grabbing power of negative social information. Journal of Personality and Social Psychology, 61, 380 391. Rothermund, K., & Wentura, D. (2001). Figure-ground asymmetries in the Implicit Association Test (IAT). Zeitschrift für Experimentelle Psychologie, 48, 94 106. Rothermund, K., & Wentura, D. (2004). Underlying processes in the Implicit Association Test (IAT): Dissociating salience from associations. Journal of Experimental Psychology: General, 133, 139 165. Simon, J.R. (1990). The effects of an irrelevant directional cue on human information processing. In R.W. Proctor & T.G. Reeve (Eds.), Stimulus response compatibility (pp. 31 86). New Yoschrk: North-Holland. Spruyt, A., Hermans, D., Pandelaere, M., De Houwer, J., & Eelen, P. (2004). On the replicability of the affective priming effect in the Pronunciation Task. Experimental Psychology, 51, 109 115. Stahl, C., & Degner, J. (in press). Assessing automatic activation of valence: A multinomial model of EAST performance. Experimental Psychology. Steffens, M.C. (2004). Is the Implicit Association Test immune to faking? Experimental Psychology, 51, 165 179. Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8, 220 247. Strack, F., & Mussweiler, T. (1997). Explaining the enigmatic anchoring effect: Mechanisms of selective accessibility. Journal of Personality and Social Psychology, 73, 437 446. Tipples, J. (2001). A conceptual replication and extension of the Affective Simon Effect. Cognition and Emotion, 15, 705 711. Voss, A., Rothermund, K., & Wentura, D. (2003). Estimating the valence of single stimuli: A new variant of the affective Simon task. Experimental Psychology, 50, 86 96. Voss, A., & Klauer, K. C. (in press). Cognitive Processes in the Extrinsic Affective Simon Task: A Task-Shift Account of EAST Effects. Experimental Psychology. Wentura, D. (1999). Activation and inhibition of affective information: Evidence for negative priming in the evaluation task. Cognition and Emotion, 13, 65 91. Wentura, D., & Degner, J. (2005). Masked affective priming by pictures: Evidence for a dissociation of hostility and depreciation. Unpublished Manuscript. Wentura, D., & Rothermund, K. (2003). The meddling-in of affective information: A general model of automatic evaluation effects. In J. Musch & K.C. Klauer (Eds.), Psychology of evaluation: Affective processes in cognition and emotion (pp. 51 86). Mahwah, NJ: Erlbaum. Wentura, D., Rothermund, K., & Bak, P. (2000). Automatic vigilance: The attention-grabbing power of approach- and avoidance-related social information. Journal of Personality and Social Psychology, 78, 1024 1037. Wigboldus, D.H.J., Holland, R.W., & Van Knippenberg, A. (2005). Single target implicit associations. Unpublished manuscript. Wilson, T.D., Lindsey, S., & Schooler, T.Y. (2000). A model of dual attitudes. Psychological Review, 107, 101 126. Juliane Degner Department of Psychology Saarland University Building A2.4 P.O. Box 15 11 50 D-66041 Saarbrücken Germany E-Mail: j.degner@mx.uni-saarland.de