Seven pigeons were trained in a delayed conditional discrimination task where the sample stimulus consisted of four possible sequences of durations: each sequence could be short or long, and was composed by two light modalities (steady-flicker or vice versa). Once the discrimination criterion was met, the effect of a retention interval was assessed: 2 conditions were studied, according to the location of the retention interval, which could be located between the components that formed the sequence (Retention Interval Between Components, RIBC), or at the end of it (Retention Interval at the End of Sequences, RIES). The discrimination indexes were asymmetric for the long and short sequences (choose-short effect) in the RIES condition, not so in the RIBC condition. These results are consistent with various findings reported in the literature and they are discussed in terms of an information-processing model for tasks of temporal estimation.