Data analyses
To verify auditory pathway integrity, wave V in the click ABR was identified automatically as the major positive peak between 7.70 to 9.90 ms62 and, subsequently, reviewed by visual inspection for correctness. Wave V peak amplitude and latency were then quantified. Analysis of the FFR was based on recent guidelines and our protocols for the neonate FFR39,49,53. First, to avoid that group differences in the FFR could be explained by spontaneous neuroelectric fluctuations, the EEG at the pre-stimulus period was analyzed. For that, the root mean square (RMS) of the EEG amplitude of the 40 ms before stimulus onset (i.e., -40 to 0 ms) was computed and defined as the pre-stimulus RMS. Next, to account for the neural transmission delay from stimulus reception at the cochlea until neural phase locking onset in the central nervous system, the response’s lag (i.e., neural lag) was estimated as the time shift with which the correlation between the stimulus and the response waveforms was maximum. All successive FFR parameters computed in the present study were calculated from the frequency-following response onset, which is the stimulus onset plus the individual neural lag.
To investigate the stimulus encoding, the Fast Fourier Transform (FFT) was applied. The FFT is an algorithm that transforms the EEG signal in the time domain to a representation in the frequency domain. The lowest frequency of a periodic sound such as speech is called fundamental frequency (F0). Since pitch, defined as a perceptual correlate of the F063, is a sound attribute essential for language acquisition64,65, the analysis was centered on the stimulus F0.
The stimulus F0 spectral amplitude and its normalization were reported as indexes of response strength for the consonant transition and for the vowel regions independently, due to the different complexity in the spectral components of each region49. To carry out the spectral amplitude measurement, the mean over a 10 Hz frequency window around the stimulus F0 (i.e., ±5 Hz to 113 Hz) was retrieved. Its normalization, termed signal-to-noise ratio (SNR), was calculated as SNR = 10*log10(Signal spectral power/Noise spectral power)39, being the noise spectral power defined as the mean over two 28 Hz frequency windows located at each side of the signal window. All signal analyses were carried out using Matlab scripts (R2019b, Mathworks) created by Intelligent Hearing Systems (Miami, FI, EEUU) and in our laboratory.