Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Characterizing word processing pathway using frequency tagging
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Kishen Senziani1, Berk Gercek1, Nina Kazanina1; 1University of geneva
Despite decades of neuroimaging and electrophysiological research aimed at characterizing the spatiotemporal pathways of visual word recognition in the human brain, consensus regarding the exact anatomical and temporal trajectories of this process remains controversial. Neuroimaging studies (Salmelin 2007; Samoylov et al., 2021) have identified a tripartite sequence during a silent reading task: 1) basic visual feature analysis near the occipital midline at ~100 ms; 2) letter-string feature analysis near the left ventral occipitotemporal cortex (vOT) at ~150ms; 3) activation of the left temporal cortex between 200 and 600ms linked to high-order linguistic process. Our objective is to re-evaluate this cascade using magnetoencephalography (MEG) and the frequency-tagging technique, capitalizing on its high signal-to-noise (SNR). By flickering stimuli at specific frequencies, the frequency-tagging entrains neural population activity within active processing regions, designed to optimize the SNR while maintaining high spatiotemporal precision (Norcia et al., 2015). While this method has shown is efficacity for studying some linguistic processing, its viability vis-a-vis pathway identification remains validated. We recorded MEG data from 30 healthy, native English-speaking adults using a 306-channel MEG system; a T1 structural MR scan was also performed for precise anatomical localization. The regions of interest were defined using a standard reading localizer protocol (Fedorenko et al., 2010) consisting of 12-item sequences of words or unpronounceable words presented at 200 ms per item. During the main MEG experiment, participants read written stimuli flickering in the center of the screen at one of two frequencies, 6 Hz or 7.058 Hz. The stimuli belonged to two conditions, WORD (e.g. “wary”) or NONWORD (consonant strings, e.g.” hkwj”). Each trial was 28.3 s long and presented a sequence of 10 stimuli from the same condition, 2.83 s/stimulus. To ensure attention, a behavioral catch-query (a probe word that the participant had to respond to with a button press) was embedded in 10% of the trials. Spectral analyses averaged across all channels demonstrated a highly significant peak (SNR > 5) at F1or F2 for both experimental conditions. Sensor-space topomaps captured surface level dynamics, revealed an early occipital activation at ~100ms for both conditions; however, by 350ms, the WORD condition elicited a transient dorsal activation propagating toward the left frontal cortex that was absent in NONWORD condition, which demonstrated only an isolated activation in the vOT. To map distinct cortical region activity from our MEG signal, we performed an anatomically constrained source time-course analysis using individual T1 MR scan. This source-space reconstruction provided further details about the tripartite sequence. While no significant activation differences were observed between WORD and NONWORD near the occipital midline at 100 ms, the NONWORD condition elicited a slightly higher activation in left vOT at 150 ms. Furthermore, confirming our sensor-space analysis, the NONWORD condition elicited a higher sustained activation in the left temporal cortex between 250-600ms. To conclude, our study demonstrates that frequency tagging achieves the requisite SNR to clearly differentiate real words from consonant strings within predefined language networks and establishes the methodological foundation for utilizing frequency tagging to study pathways of word processing.
Topic Areas: Methods,