Some current views of phonology assume a single abstract representation for each lexical item, while others assume extensive encoding of fine-grained detail. Some proponents of the latter view have claimed that differences in lexical (token) frequency are manifested as differences in phonetic duration. This claim was investigated in three experiments measuring the phonetic durations of heterographic pairs of homophonous English nouns differing in token frequency. Homophonous pairs were grouped according to magnitude of frequency difference within pairs: large difference (time ∼ thyme), medium difference (pain ∼ pane), and no difference (son ∼ sun). Four participants read (a) words in a list in a frame sentence; (b) target items in composed sentences; and (c) pairs in contrast. No systematic differences of ratio of duration (more frequent/less frequent) were found for individual speakers or across speakers in (a) or (b). Preliminary results for (c) show differences in duration correlated with contrastive focus and final lengthening, but not lexical frequency. The lack of positive correlation between duration and frequency calls into question the hypothesis that greater frequency leads to shorter duration, and underlines the need for a better understanding of the locus of frequency effects in the lexicon and speech production.