This study investigated two methods for establishing a systematic, selective, attending strategy in a memory task for children. One method was direct training of a specific strategy, employing instructions, fading, modeling, and prompts to direct the child's attention to the relevant features and to organize systematic looking behavior. The second method involved the design of the stimuli, making relevant features perceptually more (or less) salient. Observing behavior and short-term recognition were studied. Sixty 3 1/2-to-5 1/2-year-old children matched pictures from memory with either strategy training for systematic scanning or placebo practice followed by transfer. One-third of the subjects in each condition saw stimuli with relevant portions made perceptually salient, another third with irrelevant portions salient, and the rest with no portions salient. Strategy training enhanced systematic relevant observing behavior and facilitated recognition in both training and transfer. Stimulus saliency, when irrelevant, interfered in training for placebo subjects. Saliency directly influenced looking behavior only for young subjects in the early part of each trial. (Author/SB)