Menashe, B. | Search bioRxiv

bioRxiv2025

Expectation management in humans and LLMs

Mirative markers, such as "surprisingly", explicitly encode a violation of expectations. Such markers are used for expectation management during communication. Sensitivity to mirative markers relies on two abilities: i) updating expectations upon recognizing a mirative marker, and ii) identifying expectation violations warranting the use of a mirative marker. In this study, we compared sensitivity to mirative markers in humans and large language models (LLMs). In part 1, we used a sentence-completion task, where humans and LLMs were presented with sentence fragments and asked to continue them. Results show that for both humans and LLMs, the presence of a mirative marker significantly increased response entropy and decreased top-response probability, in line with theoretical accounts of mirative processing. In part 2, we created a novel task, mirative polarity selection, where humans and LLMs are presented with a sentence pair and asked to select whether it was connected by a mirative marker ("surprisingly") or an anti-mirative marker ("unsurprisingly"). Results show that LLMs perform at an impressive human-equivalent level. We conclude that both humans and LLMs use mirative markers as cues for calibrating their subsequent expectations during sentence processing.

neuroscience↗

bioRxiv2020

A general role for ventral white matter pathways in morphological processing: going beyond reading

The ability to recognize the structural components of words, known as morphological processing, was recently associated with the bilateral ventral white matter pathways, across different writing systems. However, it remains unclear whether these associations are specific to the context of reading. To shed light on this question, in the current study we investigated whether the ventral pathways are associated with morphological processing in an oral word production task that does not involve reading. Forty-five participants completed a morpheme-based fluency task in Hebrew, as well as diffusion MRI (dMRI) scans. We used probabilistic tractography to segment the major ventral and dorsal white matter pathways, and assessed the correlations between their microstructural properties and performance on the morpheme-based fluency task. We found significant correlations between morpheme-based fluency and properties of the bilateral ventral tracts, suggesting that the involvement of these tracts in morphological processing extends beyond the reading modality. In addition, significant correlations were found in the frontal aslant tract (FAT), a dorsal tract associated with oral fluency and speech production. Together, our findings emphasize that neurocognitive associations reflect both the cognitive construct under investigation as well as the task used for its assessment. Lastly, to elucidate the biological factors underlying these correlations, we incorporated the composite hindered and restricted model of diffusion (CHARMED) framework, measured in independent scans. We found that only some of our findings could be attributed to variation in a CHARMED-based estimate of fiber density. Further, we were able to uncover additional correlations that could not be detected using traditional dMRI indices. In sum, our results show that the involvement of the ventral tracts in morphological processing extends to the production domain, and demonstrate the added value of including sensitive structural measurements in neurocognitive investigations.

neuroscience↗