Neural Sensitivity to Word Frequency Modulated by Morphological Structure: Univariate and Multivariate fMRI Evidence from Morphologically Complex Words

A fundamental question in psycholinguistics concerns whether morphological decomposition is obligatory during visual word recognition or whether whole-word access can occur under conditions of high frequency and familiarity. The present study aimed to examine how morphological complexity and lexical frequency jointly influence neural representations during visual word recognition. We addressed this question using rapid event-related fMRI with both univariate and multivariate analyses. Twenty-five native Korean speakers performed lexical decisions on simple and inflected nouns that varied parametrically in surface frequency (token frequency of the full word form) and base frequency (cumulative stem frequency). Korean’s transparent agglutinative morphology enabled a principled dissociation of these frequency measures. MVPA failed to decode morphological condition from activation patterns in any region, suggesting that simple and inflected forms are not represented as discrete neural categories. Crucially, RSA demonstrated robust encoding of surface frequency—but not base frequency—in inferior frontal gyrus (IFG) pars opercularis and supramarginal gyrus (SMG), with significantly stronger correlations for inflected than simple nouns. Univariate analyses confirmed this pattern: surface frequency interacted with morphological condition in inferior parietal lobule (IPL), such that higher surface frequency increased activation selectively for inflected forms, while base frequency showed no reliable effects. These results challenge obligatory decomposition and support a distributional framework in which lexical statistics guide the processing of morphologically complex words. Taken together, our findings suggest that morphological complexity modulates neural sensitivity to whole-word rather than stem-level statistics, underscoring how structural and statistical factors interact to support morphological processing.