I aim to elaborate on Quilty-Dunn et al.'s illustrative argumentation in support of the language-of-thought (LoT) hypothesis through the social psychological lens presented in the last section of their article. Although the authors’ address of conflict problems (target article, sect. 6.1) and implicit attitude research (target article, sect. 6.2) is compelling in its experimental context, it is not far-reaching enough, that is, it is only weakly informative about the actual socially embedded reasoning of individuals. What is missing is a conceptual test of the LoT hypothesis in real-life social situations of logical reasoning – for example, such situations prevalent in research on competitive or collaborative games and strategic thinking (e.g., Colman, Reference Colman2003; Grüning & Krueger, Reference Grüning and Krueger2021, Reference Grüning and Krueger2022; Hedden & Zhang, Reference Hedden and Zhang2002). Realistic social situations of reasoning are different to the cases addressed by Quilty-Dunn et al. in several aspects. In this commentary, I outline two aspects in more detail: (1) social and self-referential projection, and (2) deliberate thinking through experimental artificiality.

First, social situations of logical reasoning are highly complicated by experiential social learning and self-referential projection. For illustration let us turn to an example: Quilty-Dunn et al. iterate an experiment by Kurdi and Dunham (Reference Kurdi and Dunham2021) in which participants were presented with, among other statements, the following simple logical statement: “If you see a green circle, you can conclude that Ibbonif is malicious” (target article, sect. 6.2, para. 3). Adapting this straightforward statement to a context of social inference – for instance: “If you see a smirk on the face, you can conclude that Peter is malicious,” – can quickly ascend individuals into a rabbit hole of applying their (1) own social learning and (2) induction from introspection about their self. Both confounds substantially with learning and testing phases as presented in the original study. For one, the absence of a smile is not as unambiguously informative as the absence of the green circle in Kurdi and Dunham's (Reference Kurdi and Dunham2021) experiment. Social cues, like facial expressions, are predominantly multicausal and, hence, ambiguous in their information about the real state of the world, more so the less contextual information is provided (e.g., “Peter is smiling after something has happened.” vs. “[…] after something terrible has happened.”). An individual can never “conclude” with full certainty what a social signal informs one about. Second, when evaluating real social situations, individuals cannot step away from using themselves as referential source of information to evaluate the situation. In the asocial situation of coloured circles and Ibbonifs, self-referential inference is not applicable, unless in the unlikely event that an individual draws connections between these concepts and their self and personal experiences. However, in social situations, that is, situations including other people in interaction, self-referentiality is a very prominent strategy for social reasoning (e.g., Krueger, Reference Krueger2008, Reference Krueger2013; Krueger & Grüning, Reference Krueger, Grüning, Forgas, Crano and Fiedler2021; Krueger, Grüning, & Heck, Reference Krueger, Grüning and Heck2023). Both of the here outlined complications occur when we move from quasi-social to in-fact social statements. They are intended to illustrate that simple cases where associative (i.e., social learning) and propositional logic are easily distinguishable, and self-referential projection is no confound are difficult to find in actual social reasoning.

Second, the high artificiality of the experimental context and task in both of the authors’ research examples should be taken into account when interpreting their results as evidence for a specific reasoning hypothesis. The experimental context itself is expected to increase participants’ cognitive alertness and motivation for accuracy (e.g., Orne, Reference Orne1962; Zizzo, Reference Zizzo2010). The artificiality of most experimental reasoning tests, including the authors’ examples, is further likely to encourage participants’ deliberate instead of intuitive thinking regarding reasoning statements (see, as process explanation; Evans, Reference Evans2008; Kahneman, Reference Kahneman2011; recently, De Neys, Reference De Neys2022) as stimulus materials. In this respect, a strong interpretation of the discussed experiments might commit the same fallacy as early interpretations of human bias (e.g., Kahneman, Slovic, Slovic, & Tversky, Reference Kahneman, Slovic, Slovic and Tversky1982) that were later challenged to contain experimental artefacts (e.g., Gigerenzer, Reference Gigerenzer1996; Hertwig, Leuker, Pachur, Spiliopoulos, & Pleskac, Reference Hertwig, Leuker, Pachur, Spiliopoulos and Pleskac2022; but also see, Vranas, Reference Vranas2000). I hasten to note that this is largely an inherent problem of the experimental context created by conversational norms and the idiosyncrasy of the experimental design (e.g., Schwarz, Reference Schwarz and Zanna1994, Reference Schwarz1999), not a shortcoming by the authors. Experimental exploration is, by all means, meaningful. However, at the same time, it is just a first step to investigate a psychological phenomenon, even more so when considering social cognition phenomena like social reasoning. The experimental artificiality can be fled by also using observational and field study designs, exchanging some internal for ecological and external validity. Before the experiments, that Quilty-Dunn et al. call upon to argue for LoT in the social psychological space, have been extended to more ecologically valid contexts, generalizable claims of any sort, including the LoT hypothesis, should be modest.

Concluding, I welcome Quilty-Dunn et al.'s attempt for an exhaustive integration of the LoT hypothesis in psychological theory and empirics. Relevantly, with my commentary I do not attempt to rebut or support the LoT hypothesis. I seek to make the authors and readers aware of the fact that for a robust, that is, a persuasive, test of the LoT hypothesis in the social context, researchers cannot exclusively revert to simple experimental imitations of social reasoning. Instead, existing findings from realistic social inference-making scenarios have to be considered by the authors and observational and field experimental approaches need to be focused on in the future. Cross-cultural exploration, as an advanced extension of social psychology, would provide an additional opportunity to test the generalizability of the LoT hypothesis.