Neural representation of phonological wordform in temporal cortex

Abstract

While the neural bases of the earliest stages of speech categorization have been widely explored using neural decoding methods, there is still a lack of consensus on questions as basic as how wordforms are represented and in what way this word-level representation influences downstream processing in the brain. Isolating and localizing the neural representations of wordform is challenging because spoken words activate a variety of representations (e.g., segmental, semantic, articulatory) in addition to form-based representations. We addressed these challenges through a novel integrated neural decoding and effective connectivity design using region of interest (ROI)-based, source-reconstructed magnetoencephalography/electroencephalography (MEG/EEG) data collected during a lexical decision task. To identify wordform representations, we trained classifiers on words and nonwords from different phonological neighborhoods and then tested the classifiers' ability to discriminate between untrained target words that overlapped phonologically with the trained items. Training with word neighbors supported significantly better decoding than training with nonword neighbors in the period immediately following target presentation. Decoding regions included mostly right hemisphere regions in the posterior temporal lobe implicated in phonetic and lexical representation. Additionally, neighbors that aligned with target word beginnings (critical for word recognition) supported decoding, but equivalent phonological overlap with word codas did not, suggesting lexical mediation. Effective connectivity analyses showed a rich pattern of interaction between ROIs that support decoding based on training with lexical neighbors, especially driven by right posterior middle temporal gyrus. Collectively, these results evidence functional representation of wordforms in temporal lobes isolated from phonemic or semantic representations.

Keywords: Neural decoding; Neural representation; Phonology; Recognition; Spoken word.

Fig. 1

Regions of interest (ROIs) visualized over an inflated averaged cortical surface. Lateral (top) and medial (bottom) views of the left and right hemisphere are shown. ROI names are generated based on the location of the centroid vertex for each ROI in the Desikian-Killiany atlas parcellation of the average subject brain. For further description of the ROIs, see Table S2 (Online Supplementary Material)

Fig. 2

Bonferonni-corrected significant (corrected alpha = 0.05) transfer decoding clusters after training with word or nonword neighbors. (A) Clusters of significant above chance transfer decoding accuracy for words-only (blue solid) or nonwords-only (red dotted) conditions. (B) Clusters of transfer decoding accuracy which is above chance and significantly differs between training conditions. Better transfer decoding for words-only condition indicated by blue solid bars; no clusters of better transfer decoding for nonwords-only condition were observed. The vertical dotted line in both panels indicates the offset of the auditory stimuli

Fig. 3

Effective connectivity analysis of transfer decoding regions of interest (ROIs). Effective connectivity between the five ROIs that supported reliable transfer decoding of hub words after training with word neighbors in the interval of 250–550 ms after the onset of the word. Lighter arrows indicate significant one-way directed Granger causation between ROIs. Darker green arrows indicate significant reciprocal connectivity. Significance is based on FDR-corrected binomial testing with an alpha of 0.05