3. Response compatibility tasks 1 THE IMPLICIT ASSOCIATION TEST (IAT)
3.3. EXTRINSIC AFFECTIVE SIMON TASK (EAST)
De Houwer (2003) describes the EAST as a “modified version” of the IAT. Like the IAT, it is a response-compatibility task, where participants’ relative response times are contrasted to assess the strength of associations between different concepts. It is therefore implicit in the sense of being indirect and time-constrained, but is not non-conscious, unlike subliminal priming paradigms. The EAST is based on the Affective Simon Task (AST) (De Houwer & Eelen, 1998). The Simon paradigm (Simon & Rudell, 1967) refers to quicker response times when the
59
stimulus and response correspond, compared to when they do not (e.g., De Jong, Liang, & Lauber, 1994). In the AST, participants are asked to make a response (usually verbal) to the stimuli they see: e.g., “good” for names of flowers and “bad” for names of insects. The response associations are then reversed, and relative response times are calculated for each block. In contrast to the AST, participants completing the EAST will additionally make their responses to task-irrelevant features of the stimuli, such as their colour. For example, participants are instructed to classify adjectives (presented in a white font) on the basis of their valence: e.g., press Q for good words and P for bad words. Nouns, on the other hand, are classified on the basis of their colour: e.g., press Q for blue words and P for green words. The consistent categorisation of the white valence words conditions positive or negative associations with each response key; the strength of positive or negative associations with the nouns can be calculated from the response times made when the noun is presented in blue (Q: “good” key response) or green (P: “bad” key response).
The term “extrinsic” in the EAST refers to the fact that the response keys have no intrinsic valence: P or Q are not “good” or “bad”. However, they acquire those associations as a consequence of performing the task. De Houwer (2003) argues that the task-switching
“strategy” thought to be employed by people completing the IAT is problematic, because they are not directly engaged in making the desired associations. By contrast, the EAST allows for response-time comparison within the same task: target words can be presented in either colour during the same trial block, therefore changing their key assignment. The way in which the EAST is structured also does not require target stimulus pairs: i.e., it can be used to assess implicit associations between flowers and “positive” or “negative”. This is in contrast to the IAT, which can only assess associations with flowers relative to insects.
Single-category implicit measures can be useful for certain types of research. The EAST has been commonly used in studies focused on associations in semantic networks, rather than valenced associations (i.e., implicit attitudes). For instance, research has been done in spider phobia, to assess associations between ambiguous spider/human stimuli and threat (Ellwart,
60
Becker, & Rinck, 2005). Vancleef, Gilissen and de Jong (2007) have assessed people’s
associations between illness and anxiety, and health protective behaviour. De Jong, Wiers, van de Braak and Huijding (2007) have also argued that the EAST paradigm has helped disambiguate research in implicit attitudes towards alcohol and alcohol abuse. Assessing attitudes towards alcoholic and non-alcoholic beverages in separate blocks suggests that alcohol attitudes in problem drinkers are ambivalent and not, as previously thought, negative. However, from a conceptual standpoint, the IAT is likely to be a better methodological choice for research which is specifically concerned with contrasting two attitude objects.
Psychometric properties.
The EAST has been criticised for its poor predictive validity and internal consistency. De Houwer and De Bruycker (2007) found that while IAT- assessed attitudes towards beer and sprouts significantly predicted self-reported consumption behaviour, EAST scores did not. In a subsequent study of attitudes towards different sexual orientations, split-half reliabilities of the IAT were found at .83; for the EAST they were .02. Teige, Schnabel, Banse and Asendorpf (2004) took similar issues with the measure’s reliability. Participants’ self-associations with personality traits (shyness and angriness) were assessed using the IAT and the EAST. While the mean internal consistency of the IAT was found at α=.76; for the EAST it was α<.24. The IATs were also moderately correlated with the relevant self- reported personality measures only, indicating good discriminant validity. By contrast, the EASTs were not correlated with any of the relevant self-report measures.It must be noted that a lack of correspondence between implicit attitudes, as assessed by the EAST, and self-reported traits or behaviour is not necessarily problematic. Self report requires introspection and accuracy, and the motivation to respond honestly when this may be at odds with social desirability. By contrast, implicit attitudes are thought to be more closely associated with behaviour which is less conscious and more spontaneous (e.g., Dovidio, Kawakami, Johnson, Johnson, & Howard, 1997). Additionally, some research does lend credibility to the EAST as predictive measure: Vermeulen, Corneille and Luminet (2007) found
61
that faster response times on positively-valenced words were negatively associated with expressions of negative affect (e.g., depression, anxiety, alexithymia). While this is not strictly speaking the intended application of the EAST, it does support a potential for its application in research.
The same can be said for convergent validity: for instance, De Houwer and De Bruycker (2007) have found a very low correlation between the IAT and the EAST (r= –.10). This may be problematic for some multi-method experimental designs but is not altogether surprising, as the two paradigms are designed to assess different types of cognitive processes: the IAT for relative associations, the EAST for individual ones.
Conclusion.
The EAST falls under the broader category of response-compatibility tasks. Unlike the IAT, it can be adapted to assess associations with single stimulus categories and does not require task-switching on the part of the participant, which can be useful in some experimental paradigms. However, its psychometric properties, both in terms of predictive validity and internal consistency, appear to fall short of the IAT. The EAST may therefore be useful in research where a single-category attitude assessment is required, but the IAT is otherwise preferable due to its superior psychometric attributes.4. Conclusions
Various approaches to implicit attitude measurement have been developed, even before the concept of dual-process cognition was articulated in the literature. Measures of
physiological or emotional response have been used to indirectly assess valenced reactions to target stimuli. However, it cannot be said with certainty whether these indirect measures specifically measure automatic associations between concepts, or tap into physiological responses to other stimuli. Affective priming procedures are effective and are consistent with some theoretical approaches, such as the IR model. Non-conscious priming procedures have
62
been proposed as the least biased form of assessment, but their ecological validity is debateable.
Response-compatibility tasks, including the IAT, IRAP and the EAST, are most common in recent research. The IAT is the most commonly used and, perhaps as a consequence, the most frequently criticised. The severity of its limitations are debatable: many can be controlled for with careful material development and the extent of the role which others play is unclear. For the moment, the psychometric properties, versatility, and wide use of the IAT suggest it to be the most obvious choice for research in which a dual-category paradigm is theoretically acceptable. The psychometric properties of the IRAP, a relatively new measure, are similar to the IAT. However, the model is based in language comprehension theory, which suggests that it may not always be appropriate in research more concerned with affective or intuitive
associations. Finally, the EAST may be an acceptable single-category alternative to the IAT, however research into its psychometric properties suggest that it may be less robust.
The measures discussed here all have relative merits and limitations. They vary in the definition of “implicitness” which they employ, from indirect assessment to automatic
association. While psychometric properties can be directly contrasted as an index of quality, the theoretical background of each measure suggests that some may be more appropriate to different research questions than others. The IAT has been used for studies in the current thesis. Although its psychometric properties and widespread use (helpful for the contrasting of results and awareness of limitations) have played a role in that decision, individual reasons for method and stimulus selection are discussed in the Method section of each individual study.
63