German infants’ discrimination of the English /æ/-/ɛ/ contrast: Evidence from a cross-sectional and a longitudinal study

Young infants can discriminate many non-native sounds, but the discrimination ability is thought to decrease within the first year of life due to perceptual attunement. However, most studies tested infants' perception cross-sectionally, without examining within-group change. To this end, the current study tested German infants' discrimination of the English /æ/-/ɛ/ contrast both cross-sectionally and longitudinally using the visual habituation technique. In Experiment 1, 96 German-learning infants were tested cross-sectionally at 5-6, 8-9 and 12-13 months. Linear mixed-effects models revealed that while the 5-6-month-olds did not discriminate the contrast, the 8-9- and 12-13-month-olds showed signs of discrimination only when they were habituated with /ɛ/, in line with previous findings suggesting that changes from central to peripheral vowels in the F1/F2 vowel space are more noticeable than in the reverse direction. Moreover, the 8-9-month-olds showed a novelty preference, while the 12-13-month-olds showed a familiarity preference. In Experiment 2, the infants tested at 5-6 months in Experiment 1 were tested again at 8-9 and 12-13 months. Fifteen infants completed the three experiments. Here, only the 12-13-month-olds discriminated the contrast by showing a novelty preference but only when habituated with /æ/. Overall, both experiments showed gradual development of discrimination ability across the first year, which challenges the assumptions of perceptual attunement. We propose that the perceptual sensitivity for a non-native vocalic contrast can improve during development. The change in perceptual asymmetry tells us that the direction of asymmetry is not universal and can be altered by linguistic experience. The change from novelty to familiarity preference may be due to the emerging preference for the more native-like vowel as well as the effect of repeating experiments with the same infants. In sum, our cross-sectional and longitudinal results overlap broadly, but the potential effect of repeating experiments must be considered when interpreting longitudinal studies.