Recent research has shown that many words traditionally assumed to be homophones are not pronounced the same but exhibit systematic phonetic differences. This finding calls into question standard assumptions of established models of phonological representation and speech production. The proposed project aims at contributing to this emergent line of research by analyzing the acoustic realization of alleged homophones that have uses in more than one word class. More specifically, the aim is to compare the noun realization to the verb realization of noun-verb homophones, such as profit or notice. Since such noun-verb pairs are very frequent in English, this project targets a widespread characteristic of the English lexicon. From previous research it is known that the noun and verb pronunciation of these forms differ due to different sentence positions they occur in, resulting in nouns being more frequently subject to lengthening effects. Beyond this contextual influence, there is initial evidence of further variation between noun and verb realization; however, this has not been systematically explored yet. The first aim of this project is to empirically investigate these context-independent differences in acoustic realization, using speech data from corpora and from experimental elicitation. The second aim of the project is to test whether language users use phonetic detail of noun-verb homophones as a cue to word class; this assumption will be tested in perception experiments. The findings from both the production and the perception tasks will contribute to our understanding of the representation of phonetic detail for a substantial share of the English lexicon.