This paper presents an exemplar-theoretic computational model of syllable frequency effects which yields simulation results in keeping with experimental results found in the literature. The argument posited here is that syllable duration variability is a function of segment duration variability for infrequent syllables. However syllable duration variability for frequent syllables cannot be predicted from segment duration variability. The simulation results support the hypothesis that frequent syllables are accessed as units whereas infrequent syllables are more likely to be produced on-line from exemplars of their constituent segments.