On the predictive validity of indirect attitude measures: Prediction of consumer choice behavior on the basis of affective priming in the picture–picture naming task☆
Introduction
Throughout the history of psychology, it has been widely assumed that attitudes can serve as a powerful energizer of behavior (e.g., Allport, 1935). It is not surprising then that behavioral scientists have long sought for reliable attitude assessment techniques in order to understand and predict behavior. Recently, a number of attitude measurement techniques have been developed that assess an individual’s attitudes ‘indirectly’, that is, without having to ask for a direct, verbal report (Fazio & Olson, 2003). Well-known examples are the Implicit Association Test (IAT, Greenwald, McGhee, & Schwartz, 1998), the Extrinsic Affective Simon Task (EAST, De Houwer, 2003), the Go/No-Go Association Task (GNAT, Nosek & Banaji, 2001), and the affective priming paradigm (e.g., Fazio, Jackson, Dunton, & Williams, 1995), but many other attitude measurement techniques are also available (see Fazio & Olson, 2003, for a review). As a rationale for the use of these ‘indirect’ attitude measures, it is typically argued that they (a) are less likely to be affected by social desirability and intentional deception as compared to direct verbal reports (e.g., Fazio & Olson, 2003; but see Steffens, 2004) and (b) might be able to register traces of past experience that are introspectively unidentified (e.g., Asendorpf, Banse, & Mücke, 2002; see also Banaji and Greenwald, 1994, Greenwald, 1990, Greenwald and Banaji, 1995, Nisbett and Wilson, 1977, Olson and Fazio, 2003, Wilson et al., 2000).
The aim of the present research was to examine the usefulness of the picture–picture naming task, a specific version of the affective priming paradigm, as an indirect attitude measurement procedure. The motivation for this enterprise was twofold. First of all, we reasoned that the reliability of affective priming scores may be dependent upon the nature of the response task that is used. Indeed, recent studies have revealed that (a) different underlying processes can drive the affective priming effect, and (b) that the extent to which each process contributes to the observed priming effects is conditional upon specific task demands. Consider, for instance, the standard evaluative categorization task in which participants are asked to judge the affective connotation of positive and negative target stimuli (e.g., the word ‘LOVELY’) that are preceded by affectively polarized prime stimuli (e.g., the word ‘CANCER’) (e.g., Fazio, Sanbonmatsu, Powell, & Kardes, 1986). Evidence suggests that affective priming effects that are obtained with this task are largely due to the fact that the primes can trigger response tendencies that either facilitate or interfere with target responding (e.g., De Houwer et al., 2002, Fazio, 2001, Klauer, 1998, Klauer et al., 1997, Klinger et al., 2000, Musch et al., 2004, Rothermund and Wentura, 1998, Wentura, 1999, Wentura, 2000). However, such a response interference mechanism cannot contribute to the affective priming effect in the so-called naming task (e.g, pronouncing the word ‘LOVELY’) because the correct response in this task depends on the identity of the targets rather than on the valence of the targets. Nevertheless, affective priming of naming responses can be found (e.g., Bargh et al., 1996, Hermans et al., 1994, Spruyt et al., 2002, Spruyt et al., 2004, Spruyt et al., in press), suggesting that other processes also contribute to the affective priming effect. More specifically, it has been argued that affective priming of naming responses is best explained on the basis of processes that operate at an encoding level. According to such an account, affectively polarized prime stimuli automatically pre-activate the memory representations of affectively related target stimuli, thus making it easier to encode target stimuli with the same valence than targets with a different valence (e.g., Bargh et al., 1996, Chen and Bargh, 1999, De Houwer et al., 2001, De Houwer and Randell, 2004, Duckworth et al., 2002, Ferguson et al., 2005, Spruyt et al., 2002, Spruyt et al., 2004, Spruyt et al., in press; see also Bargh, 1997).
Crucially, to the extent that affective priming in the evaluative categorization task is based on processes that operate at a response selection stage, individual difference measures that are obtained with this task may be affected by factors that are unrelated to an individual’s attitude towards the prime objects. Consider, for example, the findings of Wentura (1999). He demonstrated that the time needed to evaluate a target on trial n increases when the valence of that target matches with the valence of an incongruent prime on trial n − 1. Such an effect can be explained if it is assumed that a response conflict on trial n − 1 results in a suppression of the response alternative that is trigged by the prime. If the information that was irrelevant on trial n − 1 is then relevant on trial n, this inhibition needs to be overcome and responses will be delayed. Clearly, such an effect can impair the usefulness of the evaluative categorization task as an indirect attitude measurement procedure. In contrast, given that it is unlikely that affective priming in the naming task is based on processes that operate at a response selection stage (e.g., De Houwer et al., 2001, De Houwer and Randell, 2004, Spruyt et al., 2002, Spruyt et al., 2004, Spruyt et al., in press, Spruyt et al., 2004), individual difference measures that are obtained with this task may provide a more unbiased estimate of an individual’s attitude towards the prime objects than the standard evaluative categorization task.
Second, we reasoned that the picture–picture naming task might be relatively insensitive to so-called “extra-personal” associations–associations that, although available in memory, are irrelevant to one’s evaluation of a particular attitude object (see Olson & Fazio, 2004, p. 663). Recent evidence shows that the predictive validity of indirect attitude measures can be crippled when they are influenced by this kind of information (Olson & Fazio, 2004). As an example, consider Experiment 2 of Karpinski and Hilton (2001). In that study, it was examined whether it would be possible to predict the choice between an apple and a candy bar on the basis of the IAT. Despite the socially uncontroversial nature of the relevant attitude objects, Karpinski and Hilton (2001) failed to find any relation between IAT scores reflecting relative preferences for apples versus candy bars and participants’ subsequent choice behavior. However, the IAT did reveal a marked preference for apples over candy bars. Both Karpinski and Hilton, 2001, Olson and Fazio, 2004 suggested that this pattern of results may have been due to the fact that society portrays apples quite positively relative to candy bars, and that these pro-apple extra-personal associations were reflected in their IAT scores. In a replication study, Olson and Fazio (2004) obtained supporting evidence for this reasoning. They found that IAT scores revealed little preference for either apples or candy bars when using a modified version of the IAT that reduced the impact of extra-personal associations. Moreover, they also observed significant correlations between IAT scores that were obtained with their modified IAT and a behavioral intention measure. These findings strongly suggest that the standard IAT can be contaminated with extra-personal associations and that it is important to control for these extra-personal associations in case one wants to predict behavior that is primarily driven by personal associations.
Recently, Olson and Fazio (2004) suggested that the affective priming paradigm may be less affected by extra-personal associations. These authors demonstrated that the IAT assesses associations to categories whereas the affective priming paradigm assesses evaluations of exemplars (see also De Houwer, 2001, Mitchell et al., 2003). To the extent that the IAT’s susceptibility to extra-personal associations is due to the fact that it assesses associations to categories rather than evaluations of exemplars, it could indeed be argued that a measurement procedure that assesses evaluations at an exemplar level (like the affective priming paradigm) should be less affected by extra-personal associations. Crucially, this reasoning implies that the naming task may even be less sensitive to extra-personal associations than the evaluative categorization task, because it does not rely on an explicit (normative) classification of the target stimuli as positive or negative.
In sum, the naming task may be particularly suited to be used as an indirect attitude measurement procedure because priming scores obtained with this task are probably less affected by (a) attitude-irrelevant processes operating at a response selection stage, and/or (b) extra-personal associations.
Before presenting our experimental work, some elaboration on the nature of the priming tasks used in the present studies is in order. Over the past decade, several researchers reported that they were unable to obtain reliable affective priming of naming responses (e.g., De Houwer et al., 1998, Hermans, 1996, Experiment 8; Klauer and Musch, 2001, Spruyt et al., 2004). Recent studies suggest, however, that reliable affective priming effects can nevertheless be obtained with the naming task when procedures are used that increase the extent to which naming is semantically mediated (De Houwer et al., 2001, De Houwer and Randell, 2004, Spruyt et al., in press). For example, De Houwer and Randell (2004) recently showed that significant affective priming effects can be found in a word naming task when participants are asked to name only those target words that belong to a specific semantic category. Because such an effect was not found when word naming was made conditional upon the detection of a (non-semantic) perceptual feature of the target words, De Houwer and Randell (2004) concluded that the affective relationship between a prime and a target will influence the naming of the target word only if and to the extent that semantic information can feed into the orthography-to-phonology translation process (see also De Houwer et al., 2001). As an alternative method to increase the extent to which naming is semantically mediated, one could also use pictures instead of words as primes and targets. According to the model of Glaser and Glaser, 1989, Glaser, 1992, pictures have privileged access to a semantic system that contains all semantic knowledge whereas words first need to access a non-semantic lexical system before they can activate semantic stimulus information. Given that affective information is stored within the semantic system (e.g., Bower, 1991, De Houwer and Hermans, 1994, De Houwer and Randell, 2004, Fiske and Pavelchak, 1986), it could thus be predicted that it is more likely to obtain reliable affective priming of naming responses when pictures, instead of words, are used as primes and targets. In line with this reasoning, it is indeed found that the affective priming effect replicates rather easily in a picture–picture naming task (e.g., Spruyt et al., 2002, Spruyt et al., 2004, Spruyt et al., in press), whereas studies that failed to produce reliable affective priming of naming responses all used words as primes and targets. In the present study, we therefore used the picture–picture naming task. For convenience, we will simply refer to the picture–picture naming task as the “naming task”.
Section snippets
Experiment 1
The aim of the present study was to examine the predictive validity of the naming task. More specifically, we were interested in how well individual difference scores that were obtained with the naming task would predict participants’ behavior when they were given the choice between an apple and a candy bar (see Karpinski & Hilton, 2001). In addition, for the sake of comparability with previous research, two other indirect attitude measures were administered: (a) an affective priming task with
Experiment 2
The results of Experiment 1 suggest that the naming task can be successfully used as an indirect and behaviorally predictive attitude measurement tool. However, in Experiment 1, the experimenter was always present when participants made their choice between the candy bar and the apple. Thus, one could argue that the presence of the experimenter might have influenced the choice behavior. This problem was addressed in Experiment 2.
General discussion
In two experiments, we observed that individual difference scores that were obtained with the naming task predicted participants’ choice behavior relatively well. Moreover, attitude scores that were obtained with the naming task showed predictive validity over and above direct attitude measures (Experiment 1). Accordingly, it can be concluded that the naming task may be a useful addition to the arsenal of indirect attitude measures that is currently available. In contrast, however, no relation
References (54)
- et al.
The automatic evaluation effect: unconditional automatic attitude activation with a pronunciation task
Journal of Experimental Social Psychology
(1996) A structural and process analysis of the Implicit Association Test
Journal of Experimental Social Psychology
(2001)- et al.
Affective priming of pronunciation responses: effects of target degradation
Journal of Experimental Social Psychology
(2001) - et al.
After-affects: how automatic evaluations influence the interpretation of subsequent, unrelated stimuli
Journal of Experimental Social Psychology
(2005) Picture naming
Cognition
(1992)- et al.
Non-associative semantic priming: Episodic affective priming of naming responses
Acta Psychologica
(2004) - et al.
Double dissociation between implicit and explicit personality self-concept: the case of shy behavior
Journal of Personality and Social Psychology
(2002) Attitudes
- et al.
Implicit stereotyping and prejudice
The automaticity of everyday life
Mood congruity of social judgments
Consequences of automatic evaluation: immediate behavioral predispositions to approach and avoid the stimulus
Personality and Social Psychology Bulletin
The extrinsic affective Simon task
Experimental Psychology
Differences in the affective processing of words and pictures
Cognition and Emotion
Affective and identity priming with episodically associated stimuli
Cognition and Emotion
Affective priming of semantic categorization responses
Cognition and Emotion
Robust affective priming effects in a conditional pronunciation task: evidence for the semantic representation of evaluative information
Cognition and Emotion
The automatic evaluation of novel stimuli
Psychological Science
On the automatic activation of associated evaluations: an overview
Cognition and Emotion
Variability in automatic activation as an unobtrusive measure of racial attitudes: a bona fide pipeline?
Journal of Personality and Social Psychology
On the automatic activation of attitudes
Journal of Personality and Social Psychology
Implicit attitude measures in social cognition: their meaning and use
Annual Review of Psychology
Category-based versus piecemeal-based affective responses: Developments in schema-triggered affect
Who is watching Big Brother? TV consumption predicted by masked affective priming
European Journal of Social Psychology
Context effects in Stroop-like word and picture processing
Journal of Experimental Psychology: General
Action phases and mind-sets
Deliberative and implemental mind-sets: cognitive tuning toward congruous thoughts and information
Journal of Personality and Social Psychology
Cited by (53)
Predicting meat consumption from concurrent, automatic appraisals: Introducing nuance to product appraisals
2022, AppetiteCitation Excerpt :Yet, even consumers that express awareness of the ethical, environmental, and health-related caveats of meat consumption often do not adhere to meat-free diets. Past research shows that automatic appraisals play a role in food behaviours (e.g., Prestwich et al., 2011; Spruyt et al., 2007). However, evidence concerning the exact role of automatic appraisals in meat-eating behaviours remains inconclusive.
Effect of context on fine fragrance-elicited emotions: Comparison of three experimental methodologies
2022, Food Quality and PreferenceCitation Excerpt :Individual means for reaction time and standard deviations were calculated for each subject and for each condition. Responses that were more than 2.5 standard deviations (as suggested by Spruyt et al, 2007) from the mean were rejected (3.1%). In the end, data from 30 participants were kept for the implicit and explicit questions (a total of 3,373 data items for the implicit test).
Distinguishing the affective and cognitive bases of implicit attitudes to improve prediction of food choices
2016, AppetiteCitation Excerpt :Thus, the evaluative IAT does not predict choice. This replicates past findings (e.g., Karpinski & Hilton, 2001; Olson & Fazio, 2004; Spruyt et al., 2007). Using logistic regressions, we found that the affective IAT significantly influences choice between an apple and chocolate (b = 1.36, χ2(1) = 8.28, p < .01) and this holds (marginally) under low cognitive load (b = 1.08, χ2(1) = 2.94, p = .08) and significantly under high cognitive load during choice (b = 1.72, χ2(1) = 5.70, p < .05), the interaction between the affective IAT and cognitive load being not significant (b = .63, χ2(1) = .39, p = .53).
Implicit Association Test
2015, International Encyclopedia of the Social & Behavioral Sciences: Second EditionImplicit affective associations predict snack choice for those with low, but not high levels of eating disorder symptomatology
2014, AppetiteCitation Excerpt :Within the realm of dietary behavior, the current findings are consistent with prior evidence of an association between implicit, but not explicit attitudes about various healthy and unhealthy foods and future consumption of them (Hollands et al., 2011; Perugini, 2005; Sheeran et al., 2013). Given prior inconsistencies in the strength of the relation between implicit measures and spontaneous eating behavior, the current study also sought to identify one individual-level factor, ED symptomatology, that may moderate the strength of the relation (Ayres et al., 2012; Karpinski & Hilton, 2001; Spruyt et al., 2007). To the extent that ED symptomatology involves exceptionally elevated efforts to control one's eating behavior and thoughts about food, one possible explanation for the current findings is that this elevated self-control reduces the strength of the effect of implicit affective associations on behavior (Friese & Hofmann, 2009).
- ☆
Adriaan Spruyt, postdoctoral researcher of the Research Fund K.U. Leuven (Belgium); Dirk Hermans, Department of Psychology, University of Leuven (Belgium); Jan De Houwer, Department of Psychology, Ghent University (Belgium); Joachim Vandekerckhove, Department of Psychology, University of Leuven (Belgium); Paul Eelen, Department of Psychology, University of Leuven (Belgium).