The short-term memory (STM) of 25 deaf and 20 hearing adults fluent in Australian Sign Language (Auslan) was tested using both free- and serial-recall versions of three tasks. On two tasks, where stimuli were presented as either written words or Auslan signs, hearing subjects performed significantly better than deaf subjects. This difference was attributed to the facility of the hearing subjects in translating these two classes of language-based stimuli into phonological codes, which have a preferred status in STM. On the third, language-free task, which was an adaptation of the Corsi Blocks test, the deaf and hearing subjects performed at comparable levels, indicating that differences in their STM became evident only with the introduction of language-based factors. Analyses restricted to the deaf subjects showed that performances on the language-based STM tasks correlated positively with scores on a reading comprehension test. Also, deaf subjects who reported an oral education outperformed their counterparts, who reported a total communication (oral plus signed English) education on the language-based STM tasks. Thus, for this diverse adult deaf sample, proficiency in STM for language-based material, skill in reading, and report of an oral rather than total communication education appear to covary.