Hypothesized that temporal pauses in letter series are the major places at which recoding occurs, either in terms of a pronunciation code or a meaningful code. If pauses determine the units on which recoding is attempted, then there will be an optimal and several nonoptimal ways to locate pauses in a given series of codable trigrams. 3 immediate recall experiments, with 15, 12, and 16 undergraduates, confirmed this expectation, comparing optimal pausing with several nonoptimal ways of presenting the sequences. Analyses of transition error probabilities supported the view that: (a) pauses determine S's recall units, (b) recall is superior when a pause-bound group is recodable, and (c) subjectively coded groups are detectable even when no pauses occur in the series. (PsycINFO Database Record (c) 2006 APA, all rights reserved)