Primacy and recency aspects of short term memory (STM) were investigated with 30 learning disabled (LD) and 30 normal children, all with a mean age of 113 months. The pretest experiment compared the serial position curve performance of LD and normal children when seven-digit series were presented visually or auditorially. The second experiment compared the effects of training encoding processes in the LD children who received 8.7 hours of training in one of three treatments--chunking (organizing serial input into units), verbal stringing (repeating orally the entire series as each new element is added), and hawthorne (a control treatment). It was concluded that the overall performances of LD children in STM were qualitatively similar to but quantitatively lower than normal children, while secondary memory for LD children was disproportionately more inefficient than primary memory. The training of rehearsal strategies did not produce a significant differential improvement in primacy performance, though chunking appeared to be a more effective strategy when used to organize visual stimuli (when presented slowly) and verbal stringing a more effective strategy for rehearsal of auditory stimuli. (Author/DB)