Single-Node Wilson--Cowan Model Accounts for Speech-Evoked $γ$-Band Deficits in Schizophrenia

TL;DR

The paper develops a cochlea-inspired front end coupled to a Wilson–Cowan E/I model to simulate speech-evoked gamma activity across Healthy, SCZ–speech, and SCZ–semantics conditions in six languages. It demonstrates that $ERSP_\gamma$ and $\gamma\%$ follow a consistent hierarchy (Healthy > SCZ–speech > SCZ–semantics) robust to equal-energy control and input-gain perturbations, and that network dynamics align with a reduced single-node model. Pharmacology-like perturbations along the E/I axis produce bidirectional effects on gamma metrics, supporting an operating-point interpretation of schizophrenia-related gamma deficits without evidence of Hopf bifurcations. The framework provides a testable, cross-language mechanism linking speech processing to E/I balance, with potential applications in computational psychiatry and cross-population studies.

Abstract

Cortical gamma ($γ$)-band activity reflects local excitation-inhibition (E/I) balance. In schizophrenia (SCZ), reduced task-evoked gamma suggests altered E/I dynamics, but it is unclear whether differences stem from input properties or systematic shifts in E/I operating point and gain. We coupled a cochlear-inspired speech front end to a Wilson-Cowan E/I model to simulate gamma responses across three conditions: Healthy, SCZ-speech, and SCZ-semantics. Metrics included event-related spectral perturbation (ERSP$_γ$) and threshold-time fraction ($γ%$). A stable hierarchy emerged: Healthy(speech/semantics) $>$ SCZ(speech) $>$ SCZ(semantics), robust under equal-energy control and gain perturbations. Network dynamics coincided with single-node solutions, supporting interpretability. Pharmacological analogs showed bidirectional effects: reduced inhibition lowered $γ$, while reduced excitation increased $γ$, with no self-sustained oscillations. Findings indicate SCZ gamma deficits align more with shifts in E/I operating point and gain than input differences. This pipeline provides a testable, reusable mechanistic framework for speech-evoked gamma and a baseline for cross-population studies.

Figures (5)

• Figure 1: Metric equivalence between the multi-channel network and its single-node reduction across six languages. Each panel shows scatter plots of $\gamma\%$ and ERSP$_\gamma$ from the multi-channel model versus its single-node equivalent. Each dot corresponds to one utterance/segment; the solid line is $y=x$. Points lie essentially on the identity line in all languages, confirming numerical equivalence of the two formulations.
• Figure 2: Comparison of primary metrics across experimental conditions for six languages. Each panel shows ERSP$_\gamma$ in dB (left subplot) and $\gamma\%$ (right subplot) for Healthy (speech/semantics), SCZ (speech) and SCZ (semantics). Bars indicate means with s.e.m. and dots show individual utterances/segments. In all tested languages (Mandarin, English, Japanese, German, Spanish, and Arabic), the ordering Healthy(speech/semantics) $>$ SCZ(speech) $>$ SCZ(semantics) is strictly preserved.
• Figure 3: Equal-energy control across conditions for six languages. For each stimulus, the external drive was RMS-normalized before simulation. Left subpanel: $\gamma\%$ (time-above-threshold) for each condition. Right subpanel: ERSP$_\gamma$ in dB for each condition. Planned contrasts Healthy $>$ SCZ and SCZ(speech) $>$ SCZ(semantics) remain highly significant across all tested languages, indicating that the ordering is independent of raw input energy.
• Figure 4: Sensitivity of $\gamma\%$ and ERSP$_\gamma$ to multiplicative perturbations of input gain for six languages. Each panel shows mean $\pm$ s.e.m. across utterances/segments as a function of the gain factor $j$ under both the common-scale and relative-scale schemes. Across all $j$ and in all languages, the ordering Healthy(speech/semantics) $>$ SCZ(speech) $>$ SCZ(semantics) is preserved.
• Figure 5: Pharmacological perturbations across six languages. Each figure shows the combined results of all languages. Left subpanel: $\gamma\%$ (time-above-threshold) for each condition. Right subpanel: ERSP$_\gamma$ in dB for each condition. Significant contrasts between Healthy $>$ SCZ and SCZ(speech) $>$ SCZ(semantics) across all languages are observed, demonstrating robustness in these effects across languages.