Abstract
Roelofs (2002) showed that by-item picture naming latencies in Santiago, MacKay, Palma, & Rho (2000) were linearly related to total number of segments across conditions, suggesting that structural effects of number of syllables and onset complexity might reflect a confound with phonological length. However, Roelofs failed to test the statistical reliability of this relationship with structural factors as covariates, and when we ran these and other analyses on our data, length effects were non-significant for two measures of length. We then discuss three additional Santiago et al. results favouring structural accounts but not length accounts, with number of syllables and onset complexity as the strongest structural factors, with a smaller effect (if any) of coda complexity, and no effect of vowel nucleus complexity. Finally, we argue that structure-sensitive phonological encoding mechanisms that may operate differently in different languages provide a better account of available evidence, including word production data that Roelofs (2002) claims support length accounts.