Mildly retarded, equal-MA, and equal-CA individuals viewed tachistoscopic presentations of single-element displays under a backward visual masking paradigm. Stimulus onset asynchrony and stimulus familiarity were varied to examine the process of selective attention within the context of a proposed model of attention and memory. The data revealed that, under minimal load conditions, the internal cue-selection component of selective attention for mildly retarded subjects is comparable to that of the nonretarded subjects when familiar stimuli are employed. Alternately, when unfamiliar stimuli are employed, the equal-MA and the retarded subjects are significantly less efficient relative to the equal-CA subjects. This inefficiency is attributed to the failure of the low-MA groups to apply a durable code to the products of the orienting component of selective attention.