C03

Effects of variable input on word learning and word recognition in infants

PI(s): Prof. Dr. Barbara Höhle & Prof. Dr. Adamantios Gafos

The project explores effects of phonetic and acoustic variability in the speech input on the basis of which phonological categories in early language acquisition are established. The main hypothesis pursued is that not all types of variability are equal: some types of variability support phonological category building whereas others do not support or even hinder their establishment. A major part of our experimental work consists in word learning experiments with 14-month-olds and more specifically word learning experiments where the presented stimuli are minimal pairs, that is, pairs of words differing in only one phoneme. Our focus is the hypothesis that only specific types of variability among acoustic parameters in these minimal pairs serve to highlight the phonologically relevant dimensions on the basis of which categories are built.

In a first research component, we have confirmed results from prior work that the presentation of the words in the to-be-learnt minimal pair from multiple speakers results in learning whereas presentation of the minimal pair by a single speaker does not result in a successful learning outcome. A subsequent study established that the advantage of variability in learning the minimal pair in the multiple speaker setting derives from properties of the acoustic signal; when variability was presented in the visual forms of the stimuli (rather than in their acoustic forms), learning was unsuccessful.

In a third research component, we are exploring the hypothesis that phonological categories in natural speech observe specific relations of covariability among the acoustic cues that express the phonological contrasts in these categories. Two related aims are pursued here. The first is to better understand if and how sensitivity to these co-variance relations depends on age; the second is whether stimuli with variability that conforms to these natural relations found in speech are learned better than stimuli which contain the same degree of variability but conform to different relations from those found in natural speech.

In pursuing the above aims, the project employs a number of different experimental methods used in language development research paradigms: eye-tracking coupled with habilitation and familiarization techniques, word learning experiments, and analyses of pupil dilation changes in response to paired acoustic-visual stimuli.