We tested the hypothesis that pigeons could use a cognitively efficient cod
ing strategy by training them on a conditional discrimination (delayed symb
olic matching) in which one alternative was correct following the presentat
ion of one sample (one-to-one), wheras the other alternative was correct fo
llowing the presentation of any one of four other samples (many-to-one). Wh
en retention intervals of different durations were inserted between the off
set of the sample and the onset of the choice stimuli, divergent retention
functions were found. With increasing retention interval, matching accuracy
on trials involving any of the many-to-one samples. The data suggest that
rather than learning each of the five sample-comparison associations indepe
ndently, the pigeons developed a cognitively efficient single-code/default
coding strategy.