Eric Mandelbaum

How do children form beliefs from testimony? Previous work with adults suggests a "Spinozan" account of belief formation on which propositions are accepted as true by default, and rejecting them is effortful. But it is unclear whether this asymmetry between accepting and rejection propositions arises because people learn that others generally say true things or because of the cognitive architecture supporting belief formation. To distinguish between these accounts, we presented 3- to 8-year-olds with statements about a novel animal that were confirmed or denied by a knowledgeable expert. Children with poorer executive function misremembered denied statements more often than confirmed ones. A second experiment confirmed that these results did not reflect a "yes bias", and that children could select the expert's testimony over the novice's when these were pitted against each other. Together, these results suggest that rejecting statements is challenging because of the cognitive architecture supporting belief formation.

Humans express generalizations using both generics (e.g., “Dogs bark”) and explicit quantifiers (e.g., “All dogs bark”). Quantified statements are systematically misremembered as generics more than vice versa, a phenomenon known as the Generic Recall Bias. Yet, the memory processes underlying this bias remain unclear. We investigate whether this phenomenon arises during memory encoding, retention, or retrieval. Here, we taught adults (N = 1,189) generic or quantified statements about a novel social category, and assessed recall immediately and one week later. Replicating and extending prior findings, we found that “all”, “most”, and “some” statements were each significantly misremembered as generics. Critically, longitudinal analyses revealed that misrecall patterns were stable across time and consistent within items, supporting an encoding-based account: quantified information is often encoded without its specific proportional information from the outset. Overall, these results clarify the memory mechanisms underlying human generalization.

Logical reasoning is one of humanity's most powerful abilities. A widespread assumption across psychology, linguistics, and philosophy holds that reasoning operates over concepts that refer to objects and properties in the world, yet this has rarely been tested empirically. We introduce a novel paradigm that exploits lexical ambiguity to differentiate candidate representations for human inference: word-forms, reference-fixing concepts, or more abstract "underspecified representations" that constrain meaning without fully determining it. Across three experiments (N=158), participants evaluated deductive arguments equivocating over polysemous or homonymous terms. Although these arguments are logically equivalent under standard analyses, participants reliably accepted non-truth-preserving polysemous arguments while rejecting homonymous ones. This asymmetry—the Uncommon Sense Effect—points to underspecified representations as the default basis of logical inference. This challenges a foundational assumption about meaning, and opens new empirical avenues for studying the link between what we think and why we think it.

(Commentary on Bai, D., Hafri, A., Izard, V., Firestone, C., & Strickland, B. (2025). “Core Perception”: Re-imagining Precocious Reasoning as Sophisticated Perceiving. Behavioral and Brain Sciences, 1–75.) A central motivation of the core knowledge framework has been its promise to illuminate the foundations of cognition. Reclassifying core knowledge as perceptual alters what it can plausibly explain about conceptual development. We consider three broad possibilities for how the authors’ proposal could interact with cognitive development, showing that each raises novel questions that should be addressed.

This chapter attempts to establish three theses. First, in response to recent work by Frankish, it argues that DDV is false-i.e., that people lack the power to form beliefs at will directly. Second, drawing on recent studies in social psychology, it contends that people have a propensity to form beliefs for non-epistemic reasons. Third, it argues that although DDV is false, once people become aware of their propensities to form beliefs for non-epistemic reasons, they have obligations to avoid triggering them-which are similar in kind to the sorts of obligations people would have if DDV were true.

Philosophers, psychologists and linguists routinely use truth-value judgments as a source of evidence for the meaning of specific expressions. This method presupposes that truth-value judgments track whether what was said corresponds with facts. However, a growing body of evidence suggests that ordinary people’s behavior on the truth-value judgment task is sensitive to a range of factors beyond correspondence with facts. In this paper, we investigate how judgments of truth are influenced by considerations of truthfulness. In a series of three experiments, we present results indicating that the truth-truthfulness entanglement is at the heart of the debate over the semantics of color predicates. We use these findings to challenge contextualism about color predicates. While our substantive contribution is quite targeted, we show that the methodological lessons are likely to generalize broadly to other areas of research in cognitive science.

Emerging single-process propositional perspectives in psychology and philosophy have introduced the key idea that, much like their explicit (deliberately retrieved) counterparts, implicit (automatically retrieved) attitudes should be sensitive to logical operations such as negation. In the present project, we subject this idea to a particularly stringent test by probing not only whether the formation of implicit attitudes is sensitive to negation but also whether such sensitivity additionally reflects the distinction between easy-to-negate bipolar adjectives (those with clear antonyms, e.g., strong) and difficult-to-negate unipolar adjectives (those without clear antonyms, e.g., unique). Across four experiments (three preregistered; total N = 6,707), we found (a) no significant difference in the formation of implicit attitudes in response to affirmed bipolar versus affirmed unipolar adjectives (β = 0.07); (b) modulation of the formation of implicit attitudes by whether adjectives were affirmed or negated (β = 0.67); and, critically, (c) stronger modulation of the formation of implicit attitudes by the negation of bipolar relative to unipolar adjectives (β = 0.23). These results generalized across three sets of adjectives, two implicit attitude measures (the Implicit Association Test and the Affect Misattribution Procedure), and two novel targets (social and nonsocial). Together, these data provide strong evidence to suggest that the formation of implicit attitudes is sensitive to logical operations, including not only whether an adjective is affirmed or negated but even whether adjectives are relatively easy or difficult to negate. We discuss theoretical implications for single-process and dual-process accounts of attitude formation and higher-order human cognition.

The language of thought hypothesis is a thesis about the structure of mental representations. It is an example of the computational–representational theory of mind, according to which much of cognition consists in formal computations over mental representations. What distinguishes the language of thought hypothesis from other such theories is the idea that mental representations share core features with formal languages. The language of thought hypothesis states that thinking is the transformation of mental representations in a language of thought, in which most of what we call “thinking” is in fact inferences performed over sentences in the language. The language of thought is not a natural language, such as Spanish or Haitian Creole. It is a formal language, comparable to the machine language of a computer or formal languages of logic. The language of thought is often claimed to be innate, universal, and pivotal in learning, and it is implicated in processes such as natural language acquisition. However, it is also compatible with learning and reasoning in nonlinguistic creatures as well as in mental processes seemingly unconnected to language such as perception and motor control.

The empirical study of belief is emerging at a rapid clip, uniting work from all corners of cognitive science. Reliance on belief in understanding and predicting behavior is widespread. Examples can be found, inter alia, in the placebo, attribution theory, theory of mind, and comparative psychological literatures. Research on belief also provides evidence for robust generalizations, including about how we fix, store, and change our beliefs. Evidence supports the existence of a Spinozan system of belief fixation: one that is automatic and independent of belief rejection. Independent research supports the existence of a system of fragmented belief storage: one that relies on large numbers of causally isolated, context-sensitive stores of belief in memory. Finally, empirical and observational data support at least two systems of belief change. One system adheres, mostly, to epistemological norms of updating; the other, the psychological immune system, functions to guard our most centrally held beliefs from potential inconsistency with newly formed beliefs. Refining our under- standing of these systems can shed light on pressing real-world issues, such as how fake news, propaganda, and brainwashing exploit our psychology of belief, and how best to construct our modern informational world.

How do we acquire beliefs? According to the Spinozan model, merely having a thought entails believing it. Only through a further, effortful process can one reject automatically-accepted beliefs. This paper supersedes and extends previous expositions of the Spinozan model in four ways. First, we introduce the mechanisms of the model, by characterizing the interactions of acceptance, rejection, and endorsement. Second, we present recent developmental and theoretical work that supports the Spinozan model. Third, we draw out some sociopolitical implications of having a Spinozan mind, particularly for fake news and misinformation. Finally, we extend the model from its traditional domain, belief acquisition, to belief updating and action. The result provides the foundation for a psychofunctional model of belief proper.

Lexical ambiguity has classically been categorized into two kinds. Homonyms are single word forms that map to multiple, unrelated meanings (e.g., “bat” meaning baseball equipment or a flying mammal). Polysemes are single word forms that map to multiple, related senses (e.g., “breakfast” meaning a plate of food or an event). Yet there is a longstanding debate as to whether polysemy and homonymy reflect distinct cognitive representations. Some (e.g., Fodor & Lepore, 2002; Klein & Murphy, 2001) posit that they do not — merely describing differing patterns of usage — while others (e.g., Falkum & Vicente, 2015; Pietroski, 2018) argue that polysemes, but not homonyms, involve an underspecified representation that is neutral with respect to the form’s multiple senses. While some extant experimental evidence supports the latter view (Klepousniotou, Titone, & Romero, 2008; Srinivasan, Berner, & Rabagliati, 2019), there has not yet been clear evidence of the representation of lexical ambiguity affecting domain-general reasoning. Using a novel inference paradigm, we compare participants’ dispositions to endorse deductive, Aristotelian arguments with equivocating polysemes versus comparable arguments with equivocating homonyms. We find that participants endorse the former substantially more than the latter, a phenomenon that we dub the "Uncommon Sense Effect". Our results provide direct evidence that polysemes and homonyms have underlyingly distinct mental representations — in particular that polysemes uniquely invoke an underspecified representation that allows for rule-based inferences across distinct senses.

We aim to do two things in this article. On the positive end, our goal is to explain how some seemingly incompatible aspects of belief live together, by presenting distinct mechanistic explanations of each of them: in particular we want to show how belief can be discerning, credulous, rational, and irrational. After clarifying our positive view, we take aim at some competitor views in the second half of the paper, particularly offering critiques of epistemic vigilance and social marketplace accounts of belief. We conclude by inviting proponents to provide an architecture for belief that can be tested against existing models

Four studies explore the role of perceptual fluency in attenuating bullshit receptivity, or the tendency for individuals to rate otherwise meaningless statements as “profound”. Across four studies, we presented participants with a sample of pseudoprofound bullshit statements in either a fluent or disfluent font and found that overall, disfluency attenuated bullshit receptivity while also finding little evidence that this effect was moderated by cognitive thinking style. In all studies, we measured participants’ cognitive reflection, need for cognition, faith in intuition, and superstitious beliefs. Superstition strongly predicted bullshit receptivity regardless of fluency. Inconclusive results were found for the remaining scales. Potential links for the role of perceptual disfluency in promoting analytic thinking are discussed.

“What is the structure of thought?” is as central a question as any in cognitive science. A classic answer to this question has appealed to a Language of Thought (LoT). We point to emerging research from disparate branches of the field that supports the LoT hypothesis, but also uncovers diversity in LoTs across cognitive systems, stages of development, and species. Our letter formulates open research questions for cognitive science concerning the varieties of rules and representations that underwrite various LoT-based systems and how these variations can help researchers taxonomize cognitive systems

Mental representations remain the central posits of psychology after many decades of scrutiny. However, there is no consensus about the representational format(s) of biological cognition. This paper provides a survey of evidence from computational cognitive psychology, perceptual psychology, developmental psychology, comparative psychology, and social psychology, and concludes that one type of format that routinely crops up is the language-of-thought (LoT). We outline six core properties of LoTs: (i) discrete constituents; (ii) role-filler independence; (iii) predicate–argument structure; (iv) logical operators; (v) inferential promiscuity; and (vi) abstract content. These properties cluster together throughout cognitive science. Bayesian computational modeling, compositional features of object perception, complex infant and animal reasoning, and automatic, intuitive cognition in adults all implicate LoT-like structures. Instead of regarding LoT as a relic of the previous century, researchers in cognitive science and philosophy-of-mind must take seriously the explanatory breadth of LoT-based architectures. We grant that the mind may harbor many formats and architectures, including iconic and associative structures as well as deep-neural-network-like architectures. However, as computational/representational approaches to the mind continue to advance, classical compositional symbolic structures – that is, LoTs – only prove more flexible and well-supported over time.

An analysis of a valenced corpus of English words revealed that words that rhyme with slurs are rated more poorly than their synonyms. What at first might seem like a bizarre coincidence turns out to be a robust feature of slurs, one arising from their phonetic structure. We report novel data on phonaesthetic preferences, showing that a particular class of phonemes are both particularly disliked, and overrepresented in slurs. We argue that phonaesthetic associations have been an overlooked source of some of the more peculiar, arational aspects of slurs. We conclude by drawing broader morals about the evolution of the lexicon.

Whole Brain Emulation has been championed as the most promising, well-defined route to achieving both human-level artificial intelligence and superintelligence. It has even been touted as a viable route to achieving immortality through brain uploading. WBE is not a fringe theory: the doctrine of Computationalism in philosophy of mind lends credence to the in-principle feasibility of the idea, and the standing of the Human Connectome Project makes it appear to be feasible in practice. Computationalism is a popular, independently plausible theory, and Connectomics a well-funded empirical research program, so optimism about WBE is understandable. However, this optimism may be misplaced. This article argues that WBE is, at best, no more compelling than any of the other far-flung routes to achieving superintelligence. Similarly skeptical conclusions are found regarding immortality. The essay concludes with some positive considerations in favor of the Biological Theory of consciousness, as well as morals about the limits of Computationalism in both artificial intelligence and the philosophy of mind more generally.

Belief storage is often modeled as having the structure of a single, unified web. This model of belief storage is attractive and widely assumed because it appears to provide an explanation of the flexibility of cognition and the complicated dynamics of belief revision. However, when one scrutinizes human cognition, one finds strong evidence against a unified web of belief and for a fragmented model of belief storage. Using the best available evidence from cognitive science, we develop this fragmented model into a nascent theory of the cognitive architecture of belief storage.

Ensemble perception—the encoding of objects by their group properties—is known to be resistant to outlier noise. However, this resistance is somewhat paradoxical: how can the visual system determine which stimuli are outliers without already having derived statistical properties of the ensemble? A simple solution would be that ensemble perception is not a simple, one-step process; instead, outliers are detected through iterative computations that identify items with high deviance from the mean and reduce their weight in the representation over time. Here we tested this hypothesis. In Experiment 1, we found evidence that outliers are discounted from mean orientation judgments, extending previous results from ensemble face perception. In Experiment 2, we tested the timing of outlier rejection by having participants perform speeded judgments of sets with or without outliers. We observed significant increases in reaction time (RT) when outliers were present, but a decrease compared to no-outlier sets of matched range suggesting that range alone did not drive RTs. In Experiment 3 we tested the timing by which outlier noise reduces over time. We presented sets for variable exposure durations and found that noise decreases linearly over time. Altogether these results suggest that ensemble representations are optimized through iterative computations aimed at reducing noise. The finding that ensemble perception is an iterative process provides a useful framework for understanding contextual effects on ensemble perception.

The core of Zimmerman’s picture posits an inverse correlation between an action’s automaticity and belief’s role in the action’s execution. This proposal faces serious problems. First, high-attention, high-control actions don’t seem to heighten awareness of one’s beliefs. Second, low-attention, low-control actions are caused by the same states at play when executing high-attention, high-control actions, in which case there is no ontological difference in the states involved in these behaviors. Third, on Zimmerman’s view it is unclear what it is for a state to be involved in behaviors at all, as the basic realist response—that beliefs cause behavior—is unavailable to a Zimmerman-style pragmatist. Lastly, if Zimmerman's view were right and low-level behaviors weren't caused by beliefs, then we should turn our attention to those states instead, as most of our behavior isn’t executed under conditions of high control and attention.