A central issue in speech recognition is how contrastive phonemic information is stored in the mental lexicon. Using the mismatch negativity, a coomponent of the event-related brain activity which is sensitive to language-specific phoneme representations, we have shown that the human brain refers to abstract underspecified representations where not all phonemic features are stored. This has been shown for the coronal place of articulation in vowels as predicted by the Featurally Underspecified Lexicon-model (FUL). The planned experiments will build on our previous research to (i) extend our observations for vowels in isolation to those in more complex phonological contexts with consonantal onsets and codas; (ii) study the effect of lexical status on the mismatch negativity when investigating the segmental level of phonology; and (iii) ascertain further neurobiological evidence for the underspecification of other phonological features related to the manner of articulation like stridency, nasality and voicing, which are regularly used to contrast speech sounds in natural language.
