We investigated the perceptibility of stop voicing in a domain-final neutralizing context in German that according to various phonological models is completely neutralized in favour of the voiceless category but that according to various empirical studies is distinguishable phonetically. A primary aim was to determine whether acoustic cues that were available for the stop voicing distinction were perceptible in a neutralizing context. A secondary aim was to assess whether voicing perception was influenced by phonotactic frequency and the potential for resyllabification. Nineteen listeners of a Standard German speaking variety made forced-choice judgments to synthetic stimuli spanning a voiced-voiceless continuum containing domain-final alveolar and velar stops in various neutralizing contexts that differed in terms of phonotactic probability and the potential for resyllabification. Our results showed that voicing information could be distinguished but that the perceptibility of this distinction also depended on statistical properties of phoneme sequences and whether a domain-final stop could potentially be perceptually resyllabified as domain-initial. Our general conclusion is that a categorical neutralization model is insufficient to account for stop voicing perception in German in a domain-final context: instead, voicing perceptibility in these contexts depends on an interaction between acoustic information and phonological knowledge which emerges as a generalization across the lexicon.