Improving the AMP for Sociologists: Study 2

Much of sociological theory and research relies on assumptions about how the human mind works. However, sociologists do not always test these assumptions empirically, instead treating them as metatheoretical concerns (DiMaggio 1997). Traditional sociological methods, such as forced choice surveys or qualitative interviews, are often unable to accurately tease out the influence of deliberative versus automatic thought (Vaisey 2009; Pugh 2013; Vila-Henninger 2015). While a handful of sociologists have begun to adopt measures of implicit cognition from psychology via the Implicit Association Test (IAT) (Greenwald et al. 1998; Shepherd 2011), this measure carries potentially untenable assumptions about the nature of automatic cognition (Miles et al. 2019) and may not accurately reflect implicit processes (Schimmack 2021). The current research builds on an alternate measure of implicit cognition for sociologists: the Affect Misattribution Procedure (AMP). This experimental method pairs primes and target symbols over multiple trials, and uses averaged responses to indicate priming effects net of the ‘noise’ generated by idiosyncratic target stimuli (Payne et al. 2005). We provide a crucial improvement to this method – namely, by developing a set of neutral language targets (NLTs), instead of the dominant Chinese pictographs currently used as target symbols. Although Chinese pictographs are assumed to be “affectively bland, novel, and ambiguous” nature (Murphy et al 1995:592), this assumption may no longer be tenable in an era of globalization (O’Neill 2024). Some AMP studies exclude not only Chinese-speaking or Chinese-identifying individuals (e.g., Payne et al. 2005) but other East Asian respondents as well (e.g., Ross et al. 2020; Fourakis and Cone 2020), under the assumption that neighbouring nations may recognize or interpret the Chinese pictographs. To compare the neutrality of Neutral Language Targets (NLTs) with Chinese pictographs traditionally used in the Affective Misattribution Procedure (AMP), three key features—recognizability, likeability, and interpretability—were evaluated in Study 1. This study found key differences in the recognizability, likeability, and meaning of Chinese pictographs versus neutral language targets (NLTs) across linguistic groups. However, it did not employ an AMP design, meaning it did not directly assess how these differences influence implicit affective responses. While Study 1 provided crucial insights into the explicit evaluations of CPs and NLTs, it remains unclear whether these differences translate into measurable biases in automatic cognition, particularly within the Affective Misattribution Procedure (AMP). To address this uncertainty, Study 2 embeds both CPs and NLTs into a full AMP paradigm to test how target type and language background jointly shape implicit evaluations. It does so by distinguishing two forms of measurement error: (1) systematic error, in which AMP responses vary consistently by target type, prime valence, or language group, and (2) random error, in which responses show excess residual variance due to ambiguous or semantically active targets. Mixed-effects logistic models are used to test the systematic error hypotheses (H1), examining whether CPs amplify or dampen prime-induced affective responses differently across language groups. Structural equation models test the random-error hypothesis (H2), comparing residual variances between CP and NLT targets. We predict that AMP responses will track prime valence across all participants, but that for Chinese and Japanese speakers, semantic knowledge of CPs will produce systematic Prime × Target interactions (H1). Furthermore, we expect that CPs will generate greater residual variance than NLTs—especially among Chinese speakers—reflecting increased random measurement error (H2). By introducing NLTs and explicitly modeling both systematic and random sources of variation, this study extends the AMP for sociological research, offering a methodologically rigorous framework for isolating implicit affective processes across cultural and linguistic contexts.