Sleeping neonates track transitional probabilities in speech but only retain the first syllable of words

Abstract

Extracting statistical regularities from the environment is a primary learning mechanism that might support language acquisition. While it has been shown that infants are sensitive to transition probabilities between syllables in speech, it is still not known what information they encode. Here we used electrophysiology to study how full-term neonates process an artificial language constructed by randomly concatenating four pseudo-words and what information they retain after a few minutes of exposure. Neural entrainment served as a marker of the regularities the brain was tracking during learning. Then in a post-learning phase, evoked-related potentials (ERP) to different triplets explored which information was retained. After two minutes of familiarization with the artificial language, neural entrainment at the word rate emerged, demonstrating rapid learning of the regularities. ERPs in the test phase significantly differed between triplets starting or not with the correct first syllables, but no difference was associated with subsequent violations in transition probabilities. Thus, our results revealed a two-step learning process: neonates segmented the stream based on its statistical regularities, but memory encoding targeted during the word recognition phase entangled the ordinal position of the syllables but was still incomplete at that age.

Figure 1

Experimental protocol. (a), The experiment started with a resting state period (1 mn), followed by a random stream (2 mn) then the learning stream was presented (3 mn). This long Structured stream was followed by a test phase including 8 blocks of 16 triplets presented in isolation with a jittered ISI (2–2.5 s). Each block was separated from the next by short Structured streams (30 s) to maintain learning. A Random stream and a Resting state period were again recorded at the end of the experiment to control for the effect of time, and notably infant vigilance, on EEG recordings. (b), Possible types of test words. Test words could have violations in the TPs between the second and third syllables, in the ordinal position of the syllables (1 2 3 vs. 2 3 1), or both.

Figure 2

Neural entrainment to the syllabic rate (4 Hz) and the word rate (1.33 Hz) during the three periods (Resting state, Random stream, and Structured stream). (a), SNR for the power. In light gray, the entrainment for all electrodes. In red, the mean over the electrodes showing significant entrainment (p < 0.05, one-sided t-test, FDR corrected) at the syllabic rate. In blue, the mean over the electrodes showing significant entrainment (p < 0.05, one-sided t-test, FDR corrected) at the word rate. The topographies represent the entrainment in the electrodes space at the word rate and at the syllabic rate. Asterisks indicate the electrodes showing enhanced neural activity (cross: p < 0.05, one-sided t-test, FDR corrected; dot: p < 0.05, one-sided t-test, without FDR correction). (b), Same as (a) for ITC.

Figure 3

Neural entrainment. (a), SNR for the power at the Syllable and Word rate during the three conditions (RS = resting state, RND = random stream, STR = structured stream). Asterisks represent Bonferroni corrected p-values *** < 0.001, ** < 0.01 * < 0.05. (b), Same as (a) for ITC. (c) Time course of entertainment based on power computed on 120 s time windows. The times on the x-axis correspond to the center of the time windows. Error bars represent standard errors. The red line on the top indicates when the power at the Syllabic rate (4 Hz) was larger than the null hypothesis 0 (p < 0.05, one-sided t-test, corrected by FDR). The blue line on the bottom indicates when the power at the Word rate (1.33 Hz) was larger than the null hypothesis 0 (p < 0.05, one-sided t-test, corrected by FDR). (d), Same as (c) for ITC.

Figure 4

Responses to triplets in isolation. (a) Grand-average response to ABx and BCx triplets over the early frontal positive cluster (p = 0.0152) obtained from the cluster-based permutation analysis. The thick lines correspond to ABx (gray line) and BCx (orange line) conditions and the thin lines to the sub-conditions (Words and Edge-Words vs. Part-words and Non-words). Shaded areas correspond to the standard error across neonates. The time zero corresponds to the onset of the test word. Vertical lines signal the onset of each syllable and the end of the word. The topography shows the difference ABx-BCx during the time window where significant differences were observed (gray line under the plot). (b) Same as (a), but for the early negative cluster over left-temporal posterior electrodes (p = 0.0324). (c) Same as (a), but for a late positive cluster over frontal-left electrodes (p = 0.0142). (d) Same as (a), but for a later positive cluster over frontal electrodes (p = 0.0020). (e) Time progression of the ERP early effects (a, and b) over the 8 test blocks. (f) Time progression of the average of the two late ERP effects (c, and d) over the 8 test blocks.

Figure 5

Adult behavioral experiment. The distribution of the scores for all trials and participants are represented per condition. Asterisks represent p-values of the comparisons done two by two (Tuckey tests) *** < 0.001, ** < 0.01. W = Words (A_iB_iC_i), P = Part-Words (B_iC_iA_k), E = Edge-Words (A_iB_iC_k), N = Non-words (B_iC_iA_i).