There are towns in which language-of-thought (LoT) is the best game. But do we live in one? I go through three properties that characterize the LoT hypothesis: Discrete constituents, role-filler independence, and logical operators, and argue that in each case predictions from the LoT hypothesis are a poor fit to actual human cognition. As a hypothesis of what human cognition ought to be like, LoT departs from empirical reality.

The target article signal boosts important ongoing work across the cognitive sciences. However, its theoretical claims, generative value, and purported contributions are – where not simply restatements of arguments extensively explored elsewhere – imprecise, noncommittal, and underdeveloped to a degree that makes them difficult to evaluate. The article's apparent force results from engaging with straw rather than steel opponents.

We propose that episodic thought (i.e., episodic memory and imagination) is a domain where the language-of-thought hypothesis (LoTH) could be fruitfully applied. On the one hand, LoTH could explain the structure of what is encoded into and retrieved from long-term memory. On the other, LoTH can help make sense of how episodic contents come to play such a large variety of different cognitive roles after they have been retrieved.

There are two ways to understand any proposed properties of language-of-thoughts (LoTs): As diagnostic or constitutive. We argue that this choice is critical. If candidate properties are diagnostic, their homeostatic clustering requires explanation via an underlying homeostatic mechanism. If constitutive, there is no clustering, only the properties themselves. Whether deep neural networks (DNNs) are alternatives to LoTs or potential implementations turn on this choice.

Cognitive science has evolved since early disputes between radical empiricism and radical nativism. The authors are reacting to the revival of radical empiricism spurred by recent successes in deep neural network (NN) models. We agree that language-like mental representations (language-of-thoughts [LoTs]) are part of the best game in town, but they cannot be understood independent of the other players.

In this commentary, I contend that a representative sample of the arguments in the target article miss the mark. In particular, the interface problem provides no warrant for positing similarities between representational formats, and the evidence from neurocognitive, animal, and behavioral studies is inconclusive at best. Finally, I raise doubts about whether the authors' central hypothesis is falsifiable.

This commentary seeks to supplement the case Quilty-Dunn et al. make for the psychological reality of languages-of-thought (LoTs) in two ways. First, it focuses on the reduced physical demands which LoT architectures often make compared to alternative architectures. Second, it embeds LoT research within a broader framework that can be leveraged to understand the evolution of LoTs.

Quilty-Dunn et al. maintain that language-of-thought hypothesis (LoTH) is the best game in town. We counter that LoTH is merely one source of models – always wrong, sometimes useful. Their reasons for liking LoTH are compatible with the view that LoTH provides a sometimes pragmatically useful level of abstraction over processes and mechanisms that fail to fully live up to LoT requirements.

The target article focuses on evidence from nonlinguistic faculties to defend the claim that cognition generally traffics in language-of-thought (LoT)-type representations. This focus creates needed space to discuss the mounting accumulation of nonclassical evidence for LoT, but it also misses relevant work in linguistics that directly offers a perspective on specific hypotheses about candidate LoT representations.

Quilty-Dunn et al. argue that deep convolutional neural networks (DCNNs) optimized for image classification exemplify structural disanalogies to human vision. A different kind of artificial vision – found in reinforcement-learning agents navigating artificial three-dimensional environments – can be expected to be more human-like. Recent work suggests that language-like representations substantially improves these agents’ performance, lending some indirect support to the language-of-thought hypothesis (LoTH).

Object individuation provides a test case for the claim that infants already have a prelinguistic language-of-thought (LOT). By 12 months, infants represent several sortal-kinds: Object, agent, animate, and perhaps artifact. Infants have also encountered many words for object kinds, animals, people, and artifacts, therefore it remains a viable hypothesis that language learning may play a causal role in the acquisition of sortal-kinds, contra Quilty-Dunn et al.

The target article attempted to draw connections between broad swaths of evidence by noticing a common thread: Abstract, symbolic, compositional codes, that is, language-of-thoughts (LoTs). Commentators raised concerns about the evidence and offered fascinating extensions to areas we overlooked. Here we respond and highlight the many specific empirical questions to be answered in the next decade and beyond.

Why do many societies moralize apparently harmless pleasures, such as lust, gluttony, alcohol, drugs, and even music and dance? Why do they erect temperance, asceticism, sobriety, modesty, and piety as cardinal moral virtues? According to existing theories, this puritanical morality cannot be reduced to concerns for harm and fairness: It must emerge from cognitive systems that did not evolve for cooperation (e.g., disgust-based “purity” concerns). Here, we argue that, despite appearances, puritanical morality is no exception to the cooperative function of moral cognition. It emerges in response to a key feature of cooperation, namely that cooperation is (ultimately) a long-term strategy, requiring (proximately) the self-control of appetites for immediate gratification. Puritanical moralizations condemn behaviors which, although inherently harmless, are perceived as indirectly facilitating uncooperative behaviors, by impairing the self-control required to refrain from cheating. Drinking, drugs, immodest clothing, and unruly music and dance are condemned as stimulating short-term impulses, thus facilitating uncooperative behaviors (e.g., violence, adultery, free-riding). Overindulgence in harmless bodily pleasures (e.g., masturbation, gluttony) is perceived as making people slave to their urges, thus altering abilities to resist future antisocial temptations. Daily self-discipline, ascetic temperance, and pious ritual observance are perceived as cultivating the self-control required to honor prosocial obligations. We review psychological, historical, and ethnographic evidence supporting this account. We use this theory to explain the fall of puritanism in western, educated, industrialized, rich, and democratic (WEIRD) societies, and discuss the cultural evolution of puritanical norms. Explaining puritanical norms does not require adding mechanisms unrelated to cooperation in our models of the moral mind.

The suggestion that there is a need to moralize bodily pleasures for uncooperative self-control failures doesn't fit with the historical record. I counter that the development of puritanical values was an instrument of coercion and control, rather than an adaptation for cooperation. Confusing cooperation with coercion and moral principles with conventional norms leads to misconceptions about societal arrangements.

In this commentary we challenge Fitouchi et al.'s puritanical morality account by presenting evidence showing (1) that pursuing harmless pleasures can actually support self-regulation, and (2) that sharing pleasurable experiences can foster cooperation. We conclude that puritanical morality is not as adaptive as presented, and may even suppress the potential benefits pleasure can have for the individual and society.

The moral disciplining theory proposes that people moralize excessive innocent behavior to discipline others to behave in ways that facilitate cooperation. However, such disciplining might not always be effective. To solve this puzzle of ineffective moralization we should think of puritanism in terms of moral advertisement aimed at reputation management rather than the manipulation of others.

Puritanism may evolve into a technological variant based on norms of delegation of actions and perceptions to artificial intelligence. Instead of training self-control, people may be expected to cede their agency to self-controlled machines. The cost–benefit balance of this machine puritanism may be less aversive to wealthy individualistic democracies than the old puritanism they have abandoned.

In our research on lay prototypes of immorality, we found that Chinese consider immoral behaviors to be more about showing coarse character, rather than being violent and harmful (called criminal behaviors). The target article provides a satisfying rationale for why this Chinese immorality concept, which has many similarities to the puritanical morality described here, is connected to the morality of cooperation.

We extend the target authors' moral disciplining theory (MDT) by discussing signaling, proscriptive and prescriptive morality, and the dynamics by which signaling may operate in tandem with proscriptive and prescriptive forms of moral disciplining. We also suggest that MDT can help explain challenges to economic and social progress by revealing fundamental tensions between puritanical intuitions and liberal ideals.

Self-control provides one cooperative explanation for “purity.” Other types of cooperation provide additional explanations. For example, individuals compete for status by displaying high-value social and sexual traits, which are moralised because they reduce the mutual costs of conflict. As this theory predicts, sexually unattractive traits are perceived as morally bad, aside from self-control. Moral psychology will advance more quickly by drawing on all theories of cooperation.