Processing of Phonemic Consonant Length: Semantic and Fragment Priming Evidence from Bengali

Sandra Kotzor,Allison Wetterlin,Adam C Roberts,Aditi Lahiri
DOI: https://doi.org/10.1177/0023830915580189
Abstract:Six cross-modal lexical decision tasks with priming probed listeners' processing of the geminate-singleton contrast in Bengali, where duration alone leads to phonemic contrast ([pata] 'leaf' vs. [pat:a] 'whereabouts'), in order to investigate the phonological representation of consonantal duration in the lexicon. Four form-priming experiments (auditory fragment primes and visual targets) were designed to investigate listeners' sensitivity to segments of conflicting duration. Each prime derived from a real word ([k(h)[symbol: see text]m]/[g(h)en:]) was matched with a mispronunciation of the opposite duration (*[k(h)[symbol: see text]m:]/*[g(h)en]) and both were used to prime the full words [k(h)[symbol: see text]ma] ('forgiveness') and [g(h)en:a] ('disgust') respectively. Although all fragments led to priming, the results showed an asymmetric pattern. The fragments of words with singletons mispronounced as geminates led to equal priming, while those with geminates mispronounced as singletons showed a difference. The priming effect of the real-word geminate fragment was significantly greater than that of its corresponding nonword singleton fragment. In two subsequent semantic priming tasks with full-word primes a stronger asymmetry was found: nonword geminates (*[k(h)[symbol: see text]m:a]) primed semantically related words ([marjona] 'forgiveness') but singleton nonword primes (*[ghena]) did not show priming. This overall asymmetry in the tolerance of geminate nonwords in place of singleton words is attributed to a representational mismatch and points towards a moraic representation of duration. While geminates require a mora which cannot be derived from singleton input, the additional information in geminate nonwords does not create a similar mismatch.
What problem does this paper attempt to address?