A reanalysis of the numerosity judgment data described in T. J. Palmeri (19
97) showed that the mean latency exhibits clear deviations from the power f
unction as predicted by the component power laws (CMPL) theory of strategy
shifting (T. C. Rickard, 1997). The variance of the latency systematically
increases and then decreases with practice for large numerosities, a result
that is also consistent with the CMPL theory. Neither of these results are
predicted by existing versions of either the exemplar-based random walk or
the instance theories. These findings suggest that numerosity judgment. Li
ke other skills, reflects one at a time rather than concurrent execution of
algorithmic and memory retrieval strategies.