Five experiments addressed the issue of how pigeons learn to discrimin
ate the relative frequency of stimuli. During a sampling period, three
different stimuli (keylights) were presented serially, in mixed order
, and with different frequencies. During a choice period, the stimuli
were presented simultaneously, and reinforcement was arranged for choo
sing the stimulus that was presented the least number of times during
the sample. The results showed that (a) the overall proportion of corr
ect choices was always above chance levels; (b) the likelihood of a co
rrect choice decreased with the serial position of the correct stimulu
s, a negative recency effect; (c) when the last three stimuli of the s
ample were constrained to be one of each kind, the negative recency ef
fect decreased but errors became more likely when the correct stimulus
occurred early in the sample, a negative primacy effect; (d) accurate
performance generalized to new and larger samples; and (e) under some
conditions the probability of a correct choice was independent of the
serial position of the correct stimulus. The serial position curves s
uggest that in a least frequent discrimination task, two processes det
ermine how the least frequent stimulus controls behavior: a passive de
cay process (the stimulus loses its effectiveness with time since its
last occurrence), and a residual salience process (when the stimulus o
ccurs in the first position it may decay to a higher asymptote than wh
en it occurs in later positions).