Variation in reduction strategies of Dutch word-initial consonant clusters (2003)

Abstract

The main goal of this thesis is to investigate between-individual and within-individual variation in consonant cluster reduction strategies (C 1C2 is realized as C1 or C2) among young children. The empirical base of the study is a Dutch database with over 9,000 instances of C1 and C2 realizations of 23 word-initial consonant clusters from 45 children aged between two and three years old. A major contribution of the thesis concerns the methodology of the study, in which both between- and within-child variation as well as between- and within-cluster variation are tested in a systematic way. One of the findings is that within-child variation is very limited, whereas between-child variation occurs. Thus, although an individual child is consistent in his/her reduction strategy, there is a difference between children in the way clusters are typically reduced. Some clusters are always reduced in one particular way (e.g. /sp/, /sk/, /pl/, /pr/, /bl/, /br/, /tr/, /dr/, /kl/, /kr/, /fr/, /xr/, /tw/, /sw/). In other clusters, either target segment can be realized (e.g. /st/, /sx/, /sm/, /sn/, /sl/, /kn/, /fl/, /xl/). It was furthermore found that there are implicational relationships between some clusters, i.e., realizing C2 in cluster x implies realizing C2 in cluster y, but not vice versa. Neither the difference between stable and variable clusters, nor the implicational relationships between some clusters, can be explained by the feature [sonorant] alone. Instead, the variation in reduction strategies is the result of a difference in prosodification, which is caused by the fact that some children tend to avoid onsets with similar structure in terms of sonority distance and place of articulation. The perception of clusters was tested in a pilot study. The results suggest that it is not always correct to assume that a child has clusters underlyingly. In future research, a stronger focus must be put on the status of the child's prosodic and segmental representation. This can be done only if studies target both production and perception.