A fairly extensive body of research supports the idea that chunking a continuous speech stream into discrete phonemes involves identifying specific acoustic characteristics, lining them up with phonemic features to identify segments, and building up word representations which then trigger lexical access. However, quantity-sensitive phonological systems add a further dimension to this task: that of incorporating information about segment duration into the parsing process. In Swiss German, duration is the cue to a phonemic distinction between long consonants (geminates) and short consontants (singletons). A difference in duration also implies a difference in the segment's position within the syllable: singleton duration maps onto an onset position within the syllable (CV.CV), whereas geminate duration maps onto a coda as well as an onset (CVC.CV). Previous research on Bengali (Lahiri and Marslen-Wilson 1992) suggests that the absolute duration of the consonant alone is not enough to distinguish geminates from singletons; this differentiation can only occur when the context of syllable structure is provided. The current study pits segmental-level against syllable-level identification by presenting participants with items whose consonant duration had been manipulated: words containing singletons were turned into non-words whose consonant had geminate duration, and vice versa. Behavioural and ERP results show a processing asymmetry: adding durational information does not disrupt lexical access, but subtracting it does. I argue that abstract syllable structure plays a decisive role in lexical access in quantity-sensitive consonant systems, and that the observed processing asymmetry indicates a bias towards parsing acoustic input into CV syllables.