15th International Congress of Phonetic Sciences (ICPhS-15)
The FUL model (Featurally Underspecified Lexicon) of speech perception claims that the listener extracts features from the acoustic signal and matches them to a phonologically underspecified lexicon. No syllables or segments are computed from the signal nor are word-variants listed in the lexicon. FUL predicts an asymmetry in the mapping between the signal and the mental representation depending on whether the lexical representation under consideration is underspecified or not. Thus [LABIAL] or [DORSAL] from the signal would not mismatch with an underspecified [CORONAL] but the opposite is not true. This holds even for a language like Bengali with two coronal consonants, dental and retroflex, since the feature distinguishing them is not dominated by [CORONAL] and hence not specified. Experimental results support this hypothesis.
Bibliographic reference. Lahiri, Aditi / Reetz, Henning (2003): "Retroflexes and dentals in the FUL-model", In ICPhS-15, 301-304.