In the last few years, a number of connectionist models have been described that provide an explanation for age-of-acquisition (AoA) effects in verbal and nonverbal tasks. Further simulations and an empirical study were conducted to test two predictions that arise from these models. The first prediction is that AoA effects will be modulated by the nature of the mapping between representations; effects should be largest for arbitrary mappings (i.e., where there is no relationship between input and output representations such as that found in mapping semantics to phonology, e.g., picture naming), smaller for systematic mappings and minimal for componential representations (such as those found in mapping orthography to phonology for languages with relatively transparent orthographies). The second prediction from the models is that not only AoA but also frequency should influence performance and that these two variables might interact. These two predictions were tested in an empirical study. Performance on a set of 80 picturable nouns that orthogonally varied AoA and word frequency was directly compared in picture and word naming experiments. As predicted, there was a dramatic interaction between AoA and task, with substantial AoA effects in picture naming but not word naming. In addition, there was evidence for an interaction between AoA and frequency in both tasks. These results highlight the importance of the computational mapping between representations when considering both the likely performance in any task and the influence of any underlying factors. © 2006 Psychology Press Ltd.