We focus on a complete analysis of obstruents within the FUL framework (Lahiri & Reetz 2002). This task includes the mapping of the acoustic signal to the lexicon via phonological features and a three-way matching logic. This approach allows us to deal with a certain amount of variation: speaker, dialectal and others. A second goal is to quantify the types of acoustic variation that can be handled, and examine when such variation is informative regarding dialectal differences or can explain historic developments. A computer simulation of the mapping of acoustic variability onto the underspecified lexical entries should give insight into the behavior of this representation.