@article {341, title = {Morphological information and acoustic duration in Dutch compounds}, year = {2006}, publisher = {Nederlandse Vereniging voor Fonetische Wetenschappen}, address = {Nijmegen, The Netherlands}, abstract = {

Recent literature demonstrates that articulatory salience in speech (e.g. acoustic duration and loudness) is sensitive to the amount of information carried by phonemes, syllables and words. The more predictable (i.e. less informative) a linguistic unit is in its lexical or phonological environment, the less salient its realization. Examples of this phenomenon, especially common in spontaneous speech, include acoustic reduction of highly frequent functional words, durational shortening or deletion of predictable discourse markers, and longer articulation of phonemes with higher contribution to word recognition.

We tested whether the amount of information supplied by morphological units adds to other (phonetic, prosodic and lexical) domains of predictability and modulates the acoustic duration of affixes. This research focused on the interfixes -s- or -e(n)- in Dutch compounds. The selection of the interfix is not determined by rules, but depends on probabilistic characteristics of the left and right constituent families (sets of compounds sharing the left/right constituent with the target). The goal was then to detect the impact of families in the interfix articulation.

The study was based on two datasets collected from the "Library of the Blind" component of the Spoken Dutch Corpus: 1156 tokens containing the interfix -s- and 787 tokens containing the interfix -e(n)-. The dependent variables of the study were acoustic durations of the interfixes, and, for the interfix -e(n)-, the number of segments in the interfix. The acoustic duration of phonemes was determined with the help of an ASR, while the presence of [n] in the interfix was established by two phoneticians. We report the correlation of acoustic salience of the interfix and the amount of information in both positional families, as well as the distribution of interfixes in the left family. Moreover, we demonstrate that a number of durational effects induced by phonetic and prosodic factors and so far only observed under laboratory conditions is also found in the genre of lively read aloud speech.

}, author = {Victor Kuperman} }