Memory Span for Spoken Digits in Adults With Cochlear Implants or Typical Hearing: Effects of Age and Identification...

Short-term and working memory vary across individuals and life span. Studies of how cochlear implant (CI) users remember spoken words often do not fully disentangle perceptual influences from memory assessment because stimulus identification is rarely checked; instead, correct perception is assumed by using simple or practiced stimuli. Here, we examine whether the accuracy and speed of auditory digit identification are associated with short-term and working memory in adults with CIs as a function of age. Do perceptual errors and longer processing times depress memory span scores for such listeners? An intensive digit span assessment using manual responses is compared with traditional digit span, which uses only a few speech-based trials. We also assess how vocoder-based degradation impacts identification and digit span in older versus younger listeners with normal hearing (NH).

We measured forward and backward recall for spoken digits in 25 adults with CIs (M = 57 years; range = 20-82 years), 10 older adults with NH (> 55 years; M = 64 years), and 11 younger adults with NH (< 22 years). Listeners identified recorded digits one at a time by clicking numerals on a computer screen and then were cued to recall the list. In a comparison condition, listeners only identified each item. Listeners with NH also completed the tasks under four conditions of signal degradation implemented using a channel vocoder: four or eight channels crossed with 0- or 6-mm simulated tonotopic shift, administered in randomized blocks. A mixed design was used, with the effects of degradation examined within subject, across different listener groups, and as a function of age.

CI users were near perfect at identifying auditory digits (> 99% correct, on average), both in isolation and while experiencing a memory load. Perceptual accuracy, thus, did not affect digit span in these listeners. Identification times, however, were significantly slower under memory load. Age was only weakly related to span in CI listeners. Instead, individual differences in span in the CI group were better predicted by simple no-load identification times. Among listeners with NH, spans for vocoded digits that were not frequency shifted did not differ from spans for unprocessed stimuli. Frequency-shifted vocoded digits, however, yielded lower scores that could mostly, but not entirely, be accounted for by crediting for perceptual errors. Frequency-shifted vocoded digits were particularly challenging for older adult listeners with NH to identify. Traditional digit span correlated more strongly with our primary span measure for listeners with CIs than for listeners with NH.

Digit span can be studied independently of perception in many adult CI users. For both CI listeners and listeners with NH, stimulus degradation does not appear to significantly affect memory span, unless an effect on simple identification is also present. Auditory degradation that may slow, but which does not ultimately prevent identification, appears to have negligible impact on short-term and working memory spans.