ISCA Archive SWAP 2000
ISCA Archive SWAP 2000

Does morphological information influence phonetic categorization?

Kerstin Mauth

Results from phonetic categorization experiments show that listeners use lexical information as well as semantic and syntactic information from sentences to make phonemic decisions about ambiguous sounds. This series of phonetic categorization experiments was constructed to test whether preceding sentential context can influence the perception of inflectional morphemes. The relevant morpheme was the verbal 3rd person singular marker -t in Dutch. Listeners were presented with two different types of sentence constructions: (A) Vraag jij of Jan morgen gaat? 'Do you ask whether Jan leaves tomorrow?' (B) Zie jij nog wel eens een plaat? 'Do you now and then see a record?' In all conditions the final consonant was a stop plosive that varied in place of articulation along a continuum from [t] to [k] in which the [k] endpoints always formed nonwords. The main question was whether the shift in the categorization function towards the [t] endpoint would be any different for VP's as compared to NP's. Is people's perception influenced by the fact that in the VP the very last consonant [t] constitutes an inflectional morpheme which is predictable from the context? If yes, this might be an indication of a decompositional process being at work. Because the lexical shift that was expected for both sentence constructions where the [t] endpoints were real words (gaat vs. plaat) might mask a morphological effect, two other conditions were included. The existing verbs and nouns at the end of the sentences were replaced by nonwords which had a real noun embedded within them (vla 'custard' + t = vlaat). Would listeners be more willing to label an ambiguous sound as [t] when on the basis of the context (vraag jij of Jan morgen vlaat?), they can decompose the last nonword into two meaningful units than when this decomposition wouldn't make sense (zie jij nog wel eens een vlaat?)? Would this difference go away when the last word doesn't include an embedded noun (e.g., snaat) which would make a decomposition in either context impossible? Preliminary results show that when existing words are compared with nonwords, the lexicality effect for the verbal construction is significantly larger than for the nominal construction. The next step is to compare nonwords with possible words like vlaat.

Cite as: Mauth, K. (2000) Does morphological information influence phonetic categorization? Proc. Spoken Word Access Processes (SWAP), 27-30

  author={Kerstin Mauth},
  title={{Does morphological information influence phonetic categorization?}},
  booktitle={Proc. Spoken Word Access Processes (SWAP)},