About

Roman Feiman received his PhD in Psychology from Harvard University in 2015. He completed his postdoctoral work at Harvard and UC San Diego before coming to Brown University. His work draws on a variety of approaches and methods from cognitive developmental psychology, language acquisition, psycholinguistics, and formal semantics. Roman directs the Brown Language and Thought Lab, focusing on research related to development and higher-level cognition.

Research topics

• Computer Science
• Psychology
• Sociology
• Mathematics
• Cognitive psychology
• Cognitive science
• Neuroscience
• Statistics
• Philosophy
• Developmental psychology
• Epistemology

Selected publications

• Separating Cognitive Development from Language Development using International Adoption

  OSF Preprints (OSF Preprints) · 2026-02-27

  other1st authorCorresponding
  Publisher

• ManyNumbers 3: A Multi‐Lab Study of Demographic Correlates of Early Number Knowledge

  Developmental Science · 2026-04-20

  article

  Large scale studies have documented socioeconomic (SES) and racial/ethnic disparities in children's standardized math achievement at kindergarten entry. These early math skills predict future mathematics achievement and career success. However, limited research has been conducted using large sample sizes to understand how SES and race/ethnicity are related to children's numerical skills at even younger ages. The current study aims to investigate sociodemographic variability in three fundamental areas of early numeracy: nonverbal numerosity discrimination, rote counting, and cardinal number word knowledge. In addition, we will examine if the relations between numerical skills might be explained by their shared correlations to sociodemographic factors and if differences in numerical skills between sociodemographic groups can be explained by variability in working memory. Finally, we also investigate whether childcare attendance moderates early sociodemographic differences in numerical abilities. To achieve these goals, data from children aged 2; 6-6; 0 will be gathered from ∼ 45 US sites, drawn from a larger multi-lab international project (ManyNumbers project). The findings of this research will enhance our understanding of early emerging variability in numerical skills and provide insights into developing responsive and inclusive educational practices that support diverse learning needs in the early years. SUMMARY: Early mathematical skills are crucial for long-term academic and career achievement. SES and race/ethnicity-related disparities in math achievement emerge as early as preschool. Most studies use standardized math assessments that combine different numerical skills to assess achievement gaps, leaving uncertain which specific skills vary with demographic variables. We explore disparities in developmentally significant numerical skills and their relation to demographic variables. We also report relations between WM, childcare attendance and numerical skills. Data from approximately N = 1080 children aged 2;6-6;0 will be collected from ∼ 45 US labs, including demographic information and numeracy measures.

  PublisherDOI

• Separating Cognitive Development from Language Development using International Adoption

  OSF Preprints (OSF Preprints) · 2026-02-27

  other
  Publisher

• Separating Cognitive Development from Language Development in the Acquisition of Negation using International Adoption

  2026-05-03

  articleSenior author

  Children learning words face at least one challenge, maybe two. First, children must gather enough information to identify which concept maps to which word (a linguistic bottleneck). Second, for this mapping to be possible, children may need to acquire the concept itself if it is not already available (a conceptual bottleneck). We investigate whether children face both challenges or only the first when learning the English words for negation, a concept that is abstract and non-referential, yet universal across human thought and languages. We tested whether a linguistic bottleneck is sufficient to explain how children learn to express negation by comparing negation production in typical English-learners and older internationally-adopted preschoolers, who are more conceptually mature but face a similar language-learning task. If infants normally face a conceptual bottleneck, older adoptees, who do not, should learn negation faster. If the only bottleneck is linguistic, the two groups should learn negation at the same rate relative to their acquisition of English. In Study 1, parent reports on the CDI showed that older international adoptees and language-matched infant first-language learners begin producing no and not at the same point in vocabulary growth. Study 2 examined children’s production of logical denial uses of negation words in longitudinal corpora of parent-child interactions. Adopted preschoolers’ and language-matched infants’ production of denials increased at the same rate relative to the mean length of children’s utterances. Matching patterns of negation acquisition in internationally-adopted preschoolers and first-language-learning infants support a solely linguistic, not conceptual, bottleneck on learning negation words.

  PublisherDOI

• Separating Cognitive Development from Language Development in the Acquisition of Negation using International Adoption

  PsyArXiv (OSF Preprints) · 2026-05-03

  preprint1st authorCorresponding

  Children learning words face at least one challenge, maybe two. First, children must gather enough information to identify which concept maps to which word (a linguistic bottleneck). Second, for this mapping to be possible, children may need to acquire the concept itself if it is not already available (a conceptual bottleneck). We investigate whether children face both challenges or only the first when learning the English words for negation, a concept that is abstract and non-referential, yet universal across human thought and languages. We tested whether a linguistic bottleneck is sufficient to explain how children learn to express negation by comparing negation production in typical English-learners and older internationally-adopted preschoolers, who are more conceptually mature but face a similar language-learning task. If infants normally face a conceptual bottleneck, older adoptees, who do not, should learn negation faster. If the only bottleneck is linguistic, the two groups should learn negation at the same rate relative to their acquisition of English. In Study 1, parent reports on the CDI showed that older international adoptees and language-matched infant first-language learners begin producing no and not at the same point in vocabulary growth. Study 2 examined children’s production of logical denial uses of negation words in longitudinal corpora of parent-child interactions. Adopted preschoolers’ and language-matched infants’ production of denials increased at the same rate relative to the mean length of children’s utterances. Matching patterns of negation acquisition in internationally-adopted preschoolers and first-language-learning infants support a solely linguistic, not conceptual, bottleneck on learning negation words.

  Publisher

• Title Pending

  Glossa a journal of general linguistics · 2025-01-01

  articleOpen accessSenior author

  This is an accepted article with a DOI pre-assigned that is not yet published.Children can use distributional information about where words occur to figure out their meanings. But what happens when two very different words not only have most of their distribution in common, but also compose to form indistinguishable sentential meanings in those common cases? As a negative polarity item (NPI), any is selectively licensed by certain linguistic environments, the most common of which is negation. This is the context in which children hear any in around 80% of their input. However, under negation, the meaning of any looks just like a negative quantifier in concord with the higher negation (a negative concord item; NCI). While studies of children’s production indicate that they hardly ever produce any without a licensing negation, suggesting competence with its distribution, we hypothesize that some children may have misanalysed its meaning. To investigate what children think any means, we tested 106 monolingual English-speaking children between 2 and 6 years of age, as well as 20 adults, in two picture-choice comprehension tasks. These tasks assessed their interpretation of any without a preceding negation, both in a licensed (free choice) and an unlicensed context. While most children interpreted any the same way adults did, we also found a group of children who systematically responded to any as if it meant no, consistent with a negative concord (mis)analysis. In addition to illustrating how much children rely on distributional information to learn such abstract words, this finding bears on several debates. It raises the question of whether it is possible to represent the licensing conditions of NPIs prior to knowing their meanings. And it suggests that children may be biased to assume that their language uses negative concord constructions even when it does not.

  PublisherDOI

• Language comprehension reveals natural logical ability

  OSF Preprints (OSF Preprints) · 2025-01-04

  other

  Does logic only dictate how people ought to reason or can it also describe how they actually do? Research documenting pervasive reasoning errors has been taken to show that people are not naturally logical. However, previous studies did not examine the kinds of logic inherent in the formal structure of language. We hypothesized that, if ordinary language comprehension requires computing certain logical inferences, understanding a narrative would be disrupted just when those inferences are violated. Three self-paced reading experiments tested the signatures of these computations. A fourth tested GPT-2 and 3 as models of language based only on probabilistic prediction, without built-in logical structure. We find that humans, but not GPTs, spontaneously made fast, correct inferences, suggesting that natural logic constitutes part of human thought.

  Publisher

• The development of negation in language and thought

  2025-03-17 · 4 citations

  book-chapterSenior author
  PublisherDOI

• LLMs model how humans induce logically structured rules

  ArXiv.org · 2025-07-05

  articleOpen accessSenior author

  A central goal of cognitive science is to provide a computationally explicit account of both the structure of the mind and its development: what are the primitive representational building blocks of cognition, what are the rules via which those primitives combine, and where do these primitives and rules come from in the first place? A long-standing debate concerns the adequacy of artificial neural networks as computational models that can answer these questions, in particular in domains related to abstract cognitive function, such as language and logic. This paper argues that recent advances in neural networks -- specifically, the advent of large language models (LLMs) -- represent an important shift in this debate. We test a variety of LLMs on an existing experimental paradigm used for studying the induction of rules formulated over logical concepts. Across four experiments, we find converging empirical evidence that LLMs provide at least as good a fit to human behavior as models that implement a Bayesian probablistic language of thought (pLoT), which have been the best computational models of human behavior on the same task. Moreover, we show that the LLMs make qualitatively different predictions about the nature of the rules that are inferred and deployed in order to complete the task, indicating that the LLM is unlikely to be a mere implementation of the pLoT solution. Based on these results, we argue that LLMs may instantiate a novel theoretical account of the primitive representations and computations necessary to explain human logical concepts, with which future work in cognitive science should engage.

  PublisherOA PDF

• Why Do Children Think Words Are Mutually Exclusive?

  Psychological Science · 2024-11-21 · 4 citations

  article

  How do children learn what a word means when its uses are consistent with many possible meanings? One influential idea is that children rely on an inductive bias that ensures that novel words get assigned distinct meanings from known words— mutual exclusivity. Here, we explore the possibility that mutual-exclusivity phenomena do not reflect a bias but rather information encoded in the message. Learners might effectively be told when (and when not) to assume that word meanings are mutually exclusive. In three experiments ( N = 106 from across the United States; ages 2 years, 0 months−2 years, 11 months), we show that 2-year-olds only assumed that novel words have distinct meanings if the words were spoken with focus, an information-structural marker of contrast. Without focus, we found no mutual exclusivity; novel words were understood to label familiar objects. These results provide a novel account of mutual exclusivity and demonstrate an early emerging understanding of focus and information structure.

  PublisherDOI

Frequent coauthors

• David Barner

  28 shared

• Rose M. Schneider

  University of California, San Diego

  21 shared

• Erik Brockbank

  Stanford University

  19 shared

• Gábor Bródy

  Brown University

  16 shared

• Sam Whitman McGrath

  12 shared

• Jacob Russin

  Hologic (Germany)

  12 shared

• Rebecca Doherty

  John Innes Centre

  10 shared

• Daniel Smits

  Brown University

  8 shared

Labs

• Brown Language and Thought LabPI

Education

• PhD, Psychology

  Harvard University

  2015