Bridging Minds and Models in an Understudied Language: Iconicity Rating Correlations in Turkish

11:00 - 12:30

Language Models as Tools for Psycholinguistics

Room: HSZ - N2

Chair/s:

Cosimo Iaia, Jack E Taylor

The development of powerful computational language models in recent years has seen an increasing application of such models in psychology and psycholinguistics. Both Distributional Semantic Models (e.g., word2vec, GloVe, etc) and Large Language Models (e.g., GPT, BERT) have been applied in two main ways: (1) as models of human language processing, and (2) as tools for generating measures that are relevant to psycholinguistic theories and hypotheses. However, the distinction between these two applications of language models is not always clear, and both applications are limited by fundamental differences in language processing between models engineered to achieve human-like performance, and the processes actually used in human language. Nevertheless, language models have demonstrated strong potential for providing insight into language processes. This symposium brings together five talks to address recent developments in the use of language models as tools for psycholinguistics, and the degree to which such models provide comparisons and outputs that are meaningful for progress in the field. The first talk will set a theoretical foundation for the symposium, evaluating caveats of comparing Large Language Models to humans, and outlining how meaningful comparisons require rigorous experimental methods. The second talk explores whether humans and language models (both Large Language Models and Distributional Semantic Models) represent abstract meaning in a similar way, while also highlighting differences emerging between the two systems. The third talk shows how Large Language Models can be used to generate new iconicity ratings for Turkish, providing a new avenue for investigating semantic dimensions in otherwise understudied languages. The fourth talk evaluates how well estimates of word frequency and familiarity derived from Large Language Models can explain children’s reading times. Finally, in the last talk, Distributional Semantic Models are applied to provide insight into the learning of morphology, showing that natural text provides sufficient information to learn the meanings of prefixes and suffixes. Together, these talks highlight the ongoing potential of language models as tools for psycholinguistics. However, these talks also provide opportunity for important discussion on the caveats of this approach, and on the scientific applications language models can support.

Submission 459

Bridging Minds and Models in an Understudied Language: Iconicity Rating Correlations in Turkish

SymposiumTalk-03

Presented by: Elif Ecem Caliskan

Elif Ecem Caliskan ¹, Irmak Kalkan ¹, Basak Dagistan ¹, Dilara Cil ¹, Xenia Kudláčková Schmalz ²

¹ Ludwig Maximilian University of Munich, Germany

² Chair of General Psychology, Technical University Dresden, Germany

Iconicity refers to a resemblance between orthographic word form and meaning, facilitating word learning and recognition. Psycholinguistic science has long been shaped by Eurocentric biases (Share, 2021), with most large-scale evidence for iconicity originating from English and other Indo-European languages (Perry et al., 2015; Blasi et al., 2022). We address this gap by examining correlates of iconicity in Turkish, a morphologically rich and underrepresented non-Indo-European language. Additionally, we compare whether large language models capture similar patterns to human intuitions. Native Turkish speakers (target N ≈ 240) rate 900 words for their iconicity. We test whether iconicity varies across lexical categories (e.g., adjectives > nouns), as has been shown in English and Spanish, and how it relates to other semantic factors, including age of acquisition, imageability ratings, orthographic neighborhood structure (OLD20), semantic neighborhood density, and orthography–semantics consistency. After collecting human ratings, we generate parallel estimates from Gemma 3 27B Instruct and GPT-4-o-mini (API) large language models (LLM) and assess human–LLM correspondence via Pearson correlations and mixed-effects models. We predict that words with orthographic neighbors that are also semantically similar will receive higher iconicity ratings and that iconicity will correlate positively with imageability and negatively with age of acquisition. With LLM data, we assess correlations with human behavioral data and similarities in the correlation patterns between the different semantic variables. By integrating behavioral and model-based data, the study contributes both conceptually, expanding iconicity research beyond Indo-European languages, and methodologically, by testing how symbolic and statistical systems converge in representing sound–meaning resemblance.

Bookmark