Christine Cuskley
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Christine Cuskley has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
Vittorio Loreto
Department of Physics, University of Rome
4 shared publicationsFrancesca Colaiori
Istituto dei Sistemi Complessi (ISC-CNR), via dei Taurini 19, I-00185 Rome, Italy; Dipartimento di Fisica, Sapienza Università di Roma, P.le A. Moro 2, I-00185 Rome, Italy
3 shared publicationsFrancesca Tria
ISI Foundation, Via Alassio 11/c, Turin, Italy
3 shared publicationsMartina Pugliese
Dipartimento di Fisica, Sapienza Università di Roma, P.le A. Moro 2, I-00185 Rome, Italy
3 shared publications- Rachael Bailes
Centre for Behaviour and Evolution, Linguistics Section, Newcastle University, Newcastle upon Tyne, UK
3 shared publications Simon Kirby
Centre for Language Evolution, University of Edinburgh, Edinburgh, UK
3 shared publications
Latest Publications
- 1 July 2026
- Published by Cambridge University Press (CUP) in Behavioral and Brain Sciences
- Vol. 49, e207
For robust research, center values, not technology
Large language models are interesting, but linguistics and cognitive science should be cautious about centering any new technology as a magic bullet. Doing so reinforces the historically “narrow focus” of linguistics identified in the target article, rather than expanding our understanding of human language. We argue that centering values instead of technology is the best way to future-proof scholarly work and to keep finding out new things about language.
- 1 July 2026
- Published by Cambridge University Press (CUP) in Behavioral and Brain Sciences
- Vol. 49, e207
For robust research, center values, not technology
Large language models are interesting, but linguistics and cognitive science should be cautious about centering any new technology as a magic bullet. Doing so reinforces the historically “narrow focus” of linguistics identified in the target article, rather than expanding our understanding of human language. We argue that centering values instead of technology is the best way to future-proof scholarly work and to keep finding out new things about language.
- Open Access
- 20 May 2024
- Published by Center for Open Science in PsyArXiv
The limitations of large language models for understanding human language and cognition
Researchers have recently argued that the capabilities of Large Language Models (LLMs) can provide new insights into longstanding debates about the role of learning and/or innateness in the development and evolution of human language. Here, we argue on two grounds that LLMs alone tell us very little about human language and cognition in terms of acquisition and evolution. First, any similarities between human language and the output of LLMs are purely functional. Borrowing the “four questions” framework from ethology, we argue that what LLMs do is superficially similar, but how they do it is not. In contrast to the rich multimodal data humans leverage in interactive language learning, LLMs rely on immersive exposure to vastly greater quantities of unimodal text data, with recent multimodal efforts built upon mappings between images and text. Second, turning to functional similarities between human language and LLM output, we show that human linguistic behavior is much broader. LLMs were designed to imitate the very specific behavior of human writing; while they do this impressively, the underlying mechanisms of these models limit their capacities for meaning and naturalistic interaction, and their potential for dealing with the diversity in human language. We conclude by emphasising that LLMs are not theories of language, but tools that may be used to study language, and that can only be effectively applied with specific hypotheses to motivate research
- 20 May 2024
- Published by Center for Open Science in PsyArXiv
Open AccessThe limitations of large language models for understanding human language and cognition
Researchers have recently argued that the capabilities of Large Language Models (LLMs) can provide new insights into longstanding debates about the role of learning and/or innateness in the development and evolution of human language. Here, we argue on two grounds that LLMs alone tell us very little about human language and cognition in terms of acquisition and evolution. First, any similarities between human language and the output of LLMs are purely functional. Borrowing the “four questions” framework from ethology, we argue that what LLMs do is superficially similar, but how they do it is not. In contrast to the rich multimodal data humans leverage in interactive language learning, LLMs rely on immersive exposure to vastly greater quantities of unimodal text data, with recent multimodal efforts built upon mappings between images and text. Second, turning to functional similarities between human language and LLM output, we show that human linguistic behavior is much broader. LLMs were designed to imitate the very specific behavior of human writing; while they do this impressively, the underlying mechanisms of these models limit their capacities for meaning and naturalistic interaction, and their potential for dealing with the diversity in human language. We conclude by emphasising that LLMs are not theories of language, but tools that may be used to study language, and that can only be effectively applied with specific hypotheses to motivate research
- Open Access
- 31 March 2023
- Published by Center for Open Science in PsyArXiv
Communication games: Modeling language evolution through dyadic agent interactions
This chapter focuses on how dyadic agent-based communication games are used to investigate language evolution. We start by putting communication games in brief historical and interdisciplinary context: similar approaches have been used across a variety of disciplines for close to 50 years. These have often existed in parallel, but nonetheless demonstrate a kind of convergent cultural evolution that speaks to the effectiveness of the method: dyadic communication games are a powerful tool for investigating language evolution and emergence. Focusing on the Lewis Signaling Game (LSG) and the Naming Game (NG), we make the similarities between different modeling traditions explicit by reducing them to their skeletal dynamics. Specific examples studying the role of sensory biases in the evolution of (color) categories are used to illustrate how game variations can lead to convergent but also complementary results. We further discuss strategies for implementing communication games including general strengths and weaknesses of the method, as well as how the approach has been extended to model and analyze various phenomena relevant to language evolution.
- 31 March 2023
- Published by Center for Open Science in PsyArXiv
Open AccessCommunication games: Modeling language evolution through dyadic agent interactions
This chapter focuses on how dyadic agent-based communication games are used to investigate language evolution. We start by putting communication games in brief historical and interdisciplinary context: similar approaches have been used across a variety of disciplines for close to 50 years. These have often existed in parallel, but nonetheless demonstrate a kind of convergent cultural evolution that speaks to the effectiveness of the method: dyadic communication games are a powerful tool for investigating language evolution and emergence. Focusing on the Lewis Signaling Game (LSG) and the Naming Game (NG), we make the similarities between different modeling traditions explicit by reducing them to their skeletal dynamics. Specific examples studying the role of sensory biases in the evolution of (color) categories are used to illustrate how game variations can lead to convergent but also complementary results. We further discuss strategies for implementing communication games including general strengths and weaknesses of the method, as well as how the approach has been extended to model and analyze various phenomena relevant to language evolution.
- Open Access
- 14 December 2022
- Published by Center for Open Science in PsyArXiv
The evolution of linguistic iconicity and the cross-modal cognitive suite
This chapter will focus on the deep evolutionary history of the cognitive capacities underlying linguistic iconicity. The complex capacity for linguistic iconicity has roots in a more general cross-modal ability present throughout the animal kingdom, cross-modal transfer. Cross-modal transfer is the ability to make basic inferences about sensory properties of an object in multiple modalities based on experience from only one. This situates iconicity as a fundamentally cross-modal phenomenon; part of a broader, uniquely human cross-modal cognitive suite which includes relatively rare phenomena like synesthesia, alongside more ubiquitous phenomena like sensory metaphor and cross-modal correspondences. Evidence suggests the evolutionarily deep capacity for cross-modal transfer was honed into more sophisticated capacities underlying iconicity by an evolutionary ratchet of increased prosociality during human self-domestication. This period provided strong selective pressures for increasingly complex cross-sensory communication, and eventually, the predominantly arbitrary symbolic systems that underpin modern human language.
- 14 December 2022
- Published by Center for Open Science in PsyArXiv
Open AccessThe evolution of linguistic iconicity and the cross-modal cognitive suite
This chapter will focus on the deep evolutionary history of the cognitive capacities underlying linguistic iconicity. The complex capacity for linguistic iconicity has roots in a more general cross-modal ability present throughout the animal kingdom, cross-modal transfer. Cross-modal transfer is the ability to make basic inferences about sensory properties of an object in multiple modalities based on experience from only one. This situates iconicity as a fundamentally cross-modal phenomenon; part of a broader, uniquely human cross-modal cognitive suite which includes relatively rare phenomena like synesthesia, alongside more ubiquitous phenomena like sensory metaphor and cross-modal correspondences. Evidence suggests the evolutionarily deep capacity for cross-modal transfer was honed into more sophisticated capacities underlying iconicity by an evolutionary ratchet of increased prosociality during human self-domestication. This period provided strong selective pressures for increasingly complex cross-sensory communication, and eventually, the predominantly arbitrary symbolic systems that underpin modern human language.
- Open Access
- 26 November 2022
- Published by Center for Open Science in PsyArXiv
The burden for high-quality online data collection lies with researchers, not recruitment platforms
A recent article in Perspectives on Psychological Science (Webb & Tangney, 2022, W&T) reported a study where just 2.6% of participants recruited on Amazon’s Mechanical Turk were deemed "valid". They highlight some well-established limitations of Mechanical Turk, but their takeaways—the claims that it is "too good to be true," and that it captured "only 14 human beings … [of] N=529"—are radically misleading, yet have been repeated widely. We outline issues with how W&T handled data collection and how some design choices were unsuitable for this platform. Negative attitudes towards Mechanical Turk sometimes reflect a fundamental misunderstanding of what the platform offers and how it should be used in research. We point to substantial research from that details strategies for effective recruitment on Mechanical Turk, and key elements of experimental design for online research generally. We stress that Mechanical Turk is not suitable for every kind of study. Effective use requires specific expertise and design considerations. Like all tools used in research—from advanced hardware to specialist software—the tool itself places constraints on what you should use it for. Ultimately, high-quality data requires careful experimental design. This is the responsibility of the researcher, not the crowdsourcing platform.
- 26 November 2022
- Published by Center for Open Science in PsyArXiv
Open AccessThe burden for high-quality online data collection lies with researchers, not recruitment platforms
A recent article in Perspectives on Psychological Science (Webb & Tangney, 2022, W&T) reported a study where just 2.6% of participants recruited on Amazon’s Mechanical Turk were deemed "valid". They highlight some well-established limitations of Mechanical Turk, but their takeaways—the claims that it is "too good to be true," and that it captured "only 14 human beings … [of] N=529"—are radically misleading, yet have been repeated widely. We outline issues with how W&T handled data collection and how some design choices were unsuitable for this platform. Negative attitudes towards Mechanical Turk sometimes reflect a fundamental misunderstanding of what the platform offers and how it should be used in research. We point to substantial research from that details strategies for effective recruitment on Mechanical Turk, and key elements of experimental design for online research generally. We stress that Mechanical Turk is not suitable for every kind of study. Effective use requires specific expertise and design considerations. Like all tools used in research—from advanced hardware to specialist software—the tool itself places constraints on what you should use it for. Ultimately, high-quality data requires careful experimental design. This is the responsibility of the researcher, not the crowdsourcing platform.
- Open Access
- 26 August 2022
- Published by Center for Open Science in PsyArXiv
Non-Arbitrariness across modalities: Bootstrapping the emergence and development of questions
Over the last decade, language evolution research has been the driving force behind a renewed interest in non-arbitrary mappings (Cuskley & Kirby, 2013), as well as multimodal theories of language development (Vigliocco et al., 2014). These theories have focused largely on lexical iconicity in language systems. Here, we begin to extend this line of research by considering how non-arbitrariness plays a role in language development and emergence beyond the lexicon, moving from lexical to pragmatic meaning across different modalities. In other words, we focus on non-arbitrariness in situated language (Murgiano et al., 2021; Özyürek, 2021). Specifically, we consider the role of indices of dominance in scaffolding the development of questions. By surveying the state of the art in research on prosody, gesture, and signing, we outline evidence for cross-linguistically attested, iconic gestural scaffolding of questions across signed and spoken languages. This supports development of indexical use of pitch in question prosody, which has analogies in non-arbitrary dominance signaling. We conclude by synthesizing this evidence, presenting a novel non-arbitrary bootstrapping theory for the emergence of questions.
- 26 August 2022
- Published by Center for Open Science in PsyArXiv
Open AccessNon-Arbitrariness across modalities: Bootstrapping the emergence and development of questions
Over the last decade, language evolution research has been the driving force behind a renewed interest in non-arbitrary mappings (Cuskley & Kirby, 2013), as well as multimodal theories of language development (Vigliocco et al., 2014). These theories have focused largely on lexical iconicity in language systems. Here, we begin to extend this line of research by considering how non-arbitrariness plays a role in language development and emergence beyond the lexicon, moving from lexical to pragmatic meaning across different modalities. In other words, we focus on non-arbitrariness in situated language (Murgiano et al., 2021; Özyürek, 2021). Specifically, we consider the role of indices of dominance in scaffolding the development of questions. By surveying the state of the art in research on prosody, gesture, and signing, we outline evidence for cross-linguistically attested, iconic gestural scaffolding of questions across signed and spoken languages. This supports development of indexical use of pitch in question prosody, which has analogies in non-arbitrary dominance signaling. We conclude by synthesizing this evidence, presenting a novel non-arbitrary bootstrapping theory for the emergence of questions.
- Open Access
- 10 January 2022
- Published by Center for Open Science in PsyArXiv
The Cultural Evolution of Language
Language is one of only a handful of human cultural systems that is both unique to our species, and universal. This chapter will focus on the cultural evolution of language, situating this alongside the phylogenetic and developmental timescales which also feed into the evolution of language. The chapter begins by outlining the relationship between the emergence of human language and the language faculty and the more rapid, ongoing processes of language change, which are often framed as predominantly cultural. In particular, previous work has emphasised how these timescales interact, and how cultural factors in particular shape which aspects of language exhibit broad cross-cultural variation or stability. This is followed by detailed evidence for this relationship from three domains, focusing on the role of cultural evolution in language as observed in natural language (both historical corpora and cross linguistic data), the cultural evolution of language in agent-based models, and finally, experimental studies of the cultural evolution of language. We conclude that the study of the cultural evolution of language forms an important data-rich model for the study of the evolution of cultural systems more generally, while also providing key insights into the specific dynamics of this uniquely human behaviour.
- 10 January 2022
- Published by Center for Open Science in PsyArXiv
Open AccessThe Cultural Evolution of Language
Language is one of only a handful of human cultural systems that is both unique to our species, and universal. This chapter will focus on the cultural evolution of language, situating this alongside the phylogenetic and developmental timescales which also feed into the evolution of language. The chapter begins by outlining the relationship between the emergence of human language and the language faculty and the more rapid, ongoing processes of language change, which are often framed as predominantly cultural. In particular, previous work has emphasised how these timescales interact, and how cultural factors in particular shape which aspects of language exhibit broad cross-cultural variation or stability. This is followed by detailed evidence for this relationship from three domains, focusing on the role of cultural evolution in language as observed in natural language (both historical corpora and cross linguistic data), the cultural evolution of language in agent-based models, and finally, experimental studies of the cultural evolution of language. We conclude that the study of the cultural evolution of language forms an important data-rich model for the study of the evolution of cultural systems more generally, while also providing key insights into the specific dynamics of this uniquely human behaviour.
- 22 April 2021
- Published by Springer Nature
- p. 3097-3108
Field of Linguistics, The
- Open Access
Smooth Signals and Syntactic Change
A large body of recent work argues that considerations of information density predict various phenomena in linguistic planning and production. However, the usefulness of an information theoretic account for explaining diachronic phenomena has remained under-explored. Here, we test the idea that speakers prefer informationally uniform utterances on diachronic data from historical English and Icelandic. Our results show that: (i) the information density approach allows us to predict that Subject and Object type will affect the frequencies of OV and VO in specific ways, creating a complex Constant Rate Effect, (ii) the bias towards information uniformity explains this CRE and may help to explain others, and (iii) communities of speakers are constant in their average target level of information uniformity over long periods of historical time. This finding is consistent with an understanding of this bias which places it deep in the human language faculty and the human faculty for communication.
Open AccessSmooth Signals and Syntactic Change
A large body of recent work argues that considerations of information density predict various phenomena in linguistic planning and production. However, the usefulness of an information theoretic account for explaining diachronic phenomena has remained under-explored. Here, we test the idea that speakers prefer informationally uniform utterances on diachronic data from historical English and Icelandic. Our results show that: (i) the information density approach allows us to predict that Subject and Object type will affect the frequencies of OV and VO in specific ways, creating a complex Constant Rate Effect, (ii) the bias towards information uniformity explains this CRE and may help to explain others, and (iii) communities of speakers are constant in their average target level of information uniformity over long periods of historical time. This finding is consistent with an understanding of this bias which places it deep in the human language faculty and the human faculty for communication.