Phonetic models predict EEG better, but only when speech is intelligible.

Evidence

Interestingly, the phonetic models show an improvement in their EEG predictions unseen for the ones based only on acoustical information (Fig. 3 ). Importantly, this effect is driven by the whole distribution and not only by single subjects (Jack-knife analysis, two-way repeated measures ANOVA, Greenhouse-Geisser sphericity correction, df = 1.8, F = 66.6, p < 0.0005). — Dijk, Pim van, Physiology, Psychoacoustics and…, 4371

TRF models were fit using linear regression on the standard for each speech representation (Env , Sgram , Ph , Fea , and FS ) in the primed and non-primed conditions. These models were then compared in terms of their ability of predicting the EEG data (see Methods ). Interestingly, the phonetic models show an improvement in their EEG predictions unseen for the ones based only on acoustical information (Fig. 3 ). Importantly, this effect is driven by the whole distribution and not only by single subjects (Jack-knife analysis, two-way repeated measures ANOVA, Greenhouse-Geisser sphericity correction, df = 1.8, F = 66.6, p < 0.0005). — Dijk, Pim van, Physiology, Psychoacoustics and…, 4415

Categorization

• QUESTION: Which phonetic representations (Ph vs Fea) drive the improvement, and does this reflect top-down phoneme decoding rather than low-level feature extraction?
• corroborated: 7 independent passages support this claim — Dijk, Pim van, Physiology, Psychoacoustics and…, 4371; Dijk, Pim van, Physiology, Psychoacoustics and…, 4415
• XREF: connects to: Categorical phonetic features encode EEG information the speech envelope lacks — Dijk, Pim van, Physiology, Psychoacoustics and…, 4429; Phonemic and articulatory-feature speech representations can be encoded as multivariate EEG-regression predictors — Dijk, Pim van, Physiology, Psychoacoustics and…, 4371; Temporal response functions predict EEG from speech representations with cross-validated correlation

Broader Connections

• Categorical phonetic features encode EEG information the speech envelope lacks #expands (0.85) — Dijk, Pim van, Physiology, Psychoacoustics and…, 4429
• Phonemic and articulatory-feature speech representations can be encoded as multivariate EEG-regression predictors #expands (0.80) — Dijk, Pim van, Physiology, Psychoacoustics and…, 4371
• Temporal response functions predict EEG from speech representations with cross-validated correlation #corroborates (0.78)