Persistent Difference in Short-Term Memory Span Between Sign and Speech

Abstract

Short-term memory (STM) is thought to be limited in capacity to about 7 ± 2 items for linguistic materials and 4 ± 1 items for visuospatial information (Baddeley & Logie, 1999; Cowan, 2001). Recently, we (Boutla, Supalla, Newport, & Bavelier, 2004) challenged this dichotomy between linguistic and visuospatial STM by showing that STM capacity in users of American Sign Language (ASL) is also limited to about 4 or 5 items. This finding suggests that although longer spans appear for speech, spans are not necessarily longer for linguistic materials across all modalities. Wilson and Emmorey (2006) responded that because we evaluated span using digits for English speakers and letters for ASL signers, the difference we reported might have stemmed from stimulus selection rather than language modality. Here we address this claim by reporting an experiment in which we reexamined STM span in English speakers using letters and compared the outcome with results we and Wilson and Emmorey have obtained for ASL signers. It is important to note that the discrepancy between our previous results and those of Wilson and Emmorey is not in the obtained span in signers found by all parties to be around 4 to 5 items (Fig. la). Rather, Wilson and Emmorey disputed whether the digit span of 7 ± 2 in speakers is an appropriate benchmark for comparison with signers. Using letters to measure span in speakers, they found a span of only 5.3, comparable to that of signers. They suggested that there is no difference in span between the two languages. Here we show that their result is not due to their use of letters instead of digits. Rather, in selecting letters that are translations of one another in English and ASL, Wilson and Emmorey failed to control the stimuli in each language for phonological factors known to affect span size. One such crucial factor is phonological similarity. The finding that span is longer for digits than for letters in English speakers is not new (Cavanaugh, 1972). However, this difference has been attributed to the greater phonological similarity of letter names than digit names in English (Conrad & Hull, 1964; Mueller, Seymour, Kieras, & Meyer, 2003). In our previous study, we used letters with signers and digits with speakers to match stimuli in this important regard. Finger-spelled letters are less phonologically similar than number signs in ASL and therefore are more comparable to digits in English speakers. To demonstrate that there is nothing special about letters versus digits, other than the fact that many letter names are highly similar in English (e.g., bee, dee, ee, gee) and thus prone to produce shorter spans, we show here that when phonologically controlled letter materials are used with English speakers, the span of speakers returns to the typical 7 ± 2 range and continues to contrast with the span of signers.