The finding that some subjects learn about more than one set of relevant cues in a discriminative learning or concept identification task has been considered as strong counter‐evidence to a one‐look model of attention in learning (Trabasso & Bower, 1968). It is demonstrated, however, that one‐look models can predict multiple‐cue learning if it is assumed that subjects switch attention between trials, and that they store some information about past trials. Three submodels of a general one‐look model are presented and all are shown to fit the data of Trabasso & Bower (1968) as well as their own multiple‐look model. The success of the one‐look model argues against the use of solution‐type data as a simple means for determining the breadth of attention in the learning situation. The present interpretation of multiple‐cue learning also provides a framework within which to consider at least one important role that memorial processes play in concept identification and discriminative learning.