Emergence of the cortical encoding of phonetic features in the first year of life

Abstract

Even prior to producing their first words, infants are developing a sophisticated speech processing system, with robust word recognition present by 4-6 months of age. These emergent linguistic skills, observed with behavioural investigations, are likely to rely on increasingly sophisticated neural underpinnings. The infant brain is known to robustly track the speech envelope, however previous cortical tracking studies were unable to demonstrate the presence of phonetic feature encoding. Here we utilise temporal response functions computed from electrophysiological responses to nursery rhymes to investigate the cortical encoding of phonetic features in a longitudinal cohort of infants when aged 4, 7 and 11 months, as well as adults. The analyses reveal an increasingly detailed and acoustically invariant phonetic encoding emerging over the first year of life, providing neurophysiological evidence that the pre-verbal human cortex learns phonetic categories. By contrast, we found no credible evidence for age-related increases in cortical tracking of the acoustic spectrogram.

Fig. 1. Increasing low-frequency EEG tracking of phonetic features but not sound spectrogram in the first year of life.

A Schematic diagram of the analysis paradigm. Multivariate Temporal Response Function (TRF) models were fit to describe the forward relationship between stimulus features and the low-frequency EEG signal recorded from infants (4, 7, and 11 mo) and adults. TRF models were fit for acoustic spectrogram (S; green) and phonetic feature categories (F; orange) separately. EEG prediction correlations (Pearson’s correlation) were calculated for the TRF_S and TRF_F models with cross-validation. B EEG prediction correlations of the TRF_S and TRF_F models (average across all channels) for the lowΔ-band (0.1–1 Hz), Δ-band (1–4 Hz), and Θ-band (4–8 Hz; violin plots with mean value across participants). Stars indicate significant effects of age in infants (one-way repeated measures ANOVA; *p ≤ 0.05; **p ≤ 0.01). Statistically significant effects were measured for S in lowΔ-band (p = 0.002) and Δ-band (p = 0.002), but not Θ-band (p = 0.580). Statistically significant effects were measured for F in Δ-band (p = 0.013)and Θ-band (p = 0.004), but not lowΔ-band (p = 0.427). The violin plots show the result distributions and the mean value. C Topographical patterns of the EEG prediction correlations in infants and adults for the F and S models. D Individual-participant trajectories of the EEG prediction correlations for the longitudinal infant cohort. Colours indicate increasing vs. decreasing patterns with age (red and blue respectively). The figure was built using MATLAB software.

Fig. 2. Low-frequency (Δ-band) cortical encoding of categorical phonetic features in the first year of life.

A Schematic diagram of the analysis paradigm. Forward TRF models were fit to describe the relationship between speech features (including nuisance regressors) and low-frequency EEG signals. Speech features included the acoustic spectrogram (S), half-way rectified envelope derivative (D), visual motion (V), and phonetic features (F). B (Left) Hypotheses: The cortical encoding of phonetic feature categories was expected to progressively increase across the first year of life. Hp1-3: phonetic encoding emerging from 11, 7, and 4 months of age respectively. (Right) Phonetic feature encoding measured as the EEG prediction correlation gain when including phonetic features in the TRF (violin plots with mean value across participants). Only the frequency bands showing significant effects of age for TRF_F were studied. Stars indicate significant effects of age in infants (one-way repeated measures ANOVA; *p ≤ 0.05; **p ≤ 0.01, ***p ≤ 0.001). Statistically significant effects were measured in the Δ-band (p = 0.003) and in the Θ-bands (p = 0.009). C TRF weights corresponding to phonetic features for the TRF in Δ- and Θ-bands (1–8 Hz). The figure was built using MATLAB software.