Is speech only noise for newborn infants? Electrophysiological responses to events detected within continuous speech

Converging evidence shows that newborns prefer speech over other sounds of similar complexity. It is also widely accepted that infants are able to discriminate phonemes within an oddball paradigm. However, little is known about their ability to extract and process moderately salient events embedded in the variance of continuous speech. One methodological problem for studying this issue with event-related brain potentials (ERP) is how one can define salient events within speech with a precise onset. We tested a possible solution by extracting speech events using a skewness based event detection algorithm, which is sensitive to energy changes between frequency bands. The events were classified into four groups: high-frequency, low-frequency, stop-consonant and phrase-onset events. Eight sentences were presented in two speech modes (infant directed [ID] and adult directed [AD]), 25 times, each to adults and newborn infants. Significant (compared to zero; p<0.01 in 15 consecutive time points) ERP responses were obtained for all four event categories in adults, while only the ERP to stop consonants failed to elicit a significant response in newborn infants. When the trials were split by speech mode, in newborns, significant ERPs were obtained with ID speech and the difference between the ID- AD-responses was also significant. In contrast, no significant differences were found between the responses to ID- and AD-speech events in adults. These results suggest that ID speech enhances the processing of speech events in the newborn brain. Further studies are needed to assess the limitations of the method and to determine what kind of speech events can be studied with this approach.