Propositional Accounts of Implicit Evaluation: Taking Stock and Looking Ahead

Abstract

Associative accounts suggest that implicit (indirectly measured) evaluations are sensitive primarily to co-occurrence information (e.g., pairings of gorges with positive experiences) and are represented associatively (e.g., Gorge–Nice). By contrast, recent propositional accounts have argued that implicit evaluations are also responsive to relational information (e.g., gorges causing vs. preventing ennui) and are represented propositionally (e.g., “I find gorges fascinating”). In a review of 30 empirical papers involving exposure to contradictory co-occurrence information and relational information, we found overwhelming evidence for the latter dominating the updating of implicit evaluations, supporting the propositional perspective. However, unlike explicit evaluations, implicit evaluations seem recalcitrant in the face of relational information that requires retrospective revaluation of already encoded co-occurrence information. These findings may be jointly explained by a “common currency” hypothesis under which implicit evaluations emerge from compressed summary representations, which are sensitive to relational information but are not fully propositional.

Implicit social cognition: A brief (and gentle) introduction

“…To summarize, although the propositional approach has reinvigorated research on the mechanisms of implicit attitude acquisition and change over the past decade, it seems clear that without auxiliary assumptions it cannot account for the totality of existing data (Kurdi & Dunham, 2020). Hybrid theories, such as the associative-propositional model (Gawronski & Bodenhausen, 2006), offer more flexibility but are also more difficult to falsify.…”

Sensitivity of implicit evaluations to accurate and erroneous propositional inferences

“…Specifically, contrary to the authors' preferred inferential interpretation, the second half of the learning phase in which the numbered gray squares were presented together with the neutral conditioned stimuli may have resulted in one-shot associative learning. Mutatis mutandis, similar arguments apply to a host of other studies that have used relational qualifiers going beyond the negation operator to investigate the sensitivity of implicit evaluations to relational information (for reviews, see De Houwer et al, 2020;Kurdi & Dunham, 2020).…”

“…Under this definition, implicit evaluations differ from their explicit counterparts in features of the measurement context, with the former relying on indirect indices of underlying knowledge (such as response latencies) and the latter on direct indices of underlying knowledge (such as different forms of selfreport). Whether implicit and explicit evaluations also differ from each other more deeply, especially in the types of learning and information to which they are sensitive, has been a matter of debate both in psychology (e.g., De Houwer, 2014;DeCoster, Banner, Smith, & Semin, 2006;Gawronski & Bodenhausen, 2006;Hughes, Barnes-Holmes, & De Houwer, 2011;Kruglanski & Gigerenzer, 2011;Kurdi & Dunham, 2020;Mitchell, De Houwer, & Lovibond, 2009;Smith & DeCoster, 2000;Strack & Deutsch, 2004) and philosophy (e.g., Gendler, 2008;Levy, 2014;Madva, 2016;Mandelbaum, 2016). Specifically, among many other modalities of learning, both explicit and implicit evaluations have been shown to respond to verbal statements about social targets (e.g., Cone & Ferguson, 2015;Kurdi & Banaji, 2017;2019;Peters & Gawronski, 2011;Rydell, McConnell, Strain, Claypool, & Hugenberg, 2006).…”

“…In the five experiments reported below we investigate patterns of updating in explicit and implicit evaluations in cases where the evaluative implications of co-occurrence information and of logical structure diverge. Existing theoretical perspectives are aligned in their prediction that, in such cases, explicit evaluations should be sensitive to logical structure above and beyond mere co-occurrence information (De Houwer, 2014;DeCoster et al, 2006;Gendler, 2008;Hughes et al, 2011;Kruglanski & Gigerenzer, 2011;Kurdi & Dunham, 2020;Levy, 2014;Madva, 2016;Mandelbaum, 2016;Mitchell et al, 2009;Smith & DeCoster, 2000;Strack & Deutsch, 2004). Specifically, in the example above, if it turns out that the speaker is mistaken, explicit evaluations of Laapians should remain unchanged assuming that the observer reasons in line with the rules of propositional logic under which A → B and ¬A do not jointly license any accurate inference.…”

What can the implicit social cognition literature teach us about implicit social cognition?

“…Not all memory research seeks to cure dementia; not all phonological awareness research tries to eradicate dyslexia; and not all auditory perception research contributes to the development of hearing aids. Similarly, much implicit social cognition research explores basic aspects of thought and behavior, including learning and representation (Kurdi & Dunham, 2020), social cognitive development (Dunham, Baron, & Banaji, 2008), and cultural change (Charlesworth & Banaji, 2019), without making any claim of immediate applicability to real-world problems. Thus, whether implicit social cognition research can explain real-world inequality should not be treated as its sole measure of success.…”