A Common Serial Order Mechanism for Verbal Working Memory and Language Production: Evidence from Aphasia

• ¹ Rice University, United States
• ² University of Texas at Austin, United States

Serial order information is critical in our daily life, from recalling a phone number, or producing phonemes when saying a word, or writing letters when spelling. In the domain of working memory (WM), there is clear evidence for a dissociation of item-identity and serial-order WM (Majerus, 2018). The current study tests two hypotheses about verbal serial order working memory (SOWM). First, that there is a single SOWM capacity that is involved in both verbal and nonverbal WM. Second, that there is a common order processing capacity shared between verbal WM and language production. Both hypotheses are addressed in a multiple-case study approach, looking at patterns of associations and dissociations in serial order performance in patients with memory and language impairments following stroke. Three chronic aphasic patients (M.B., E.V., and S.H.) were identified with impaired forward and backward digit span (span < 4). They were then tested on verbal WM, including item probe (IP), order probe (OP), and serial reconstruction (SRC). Evidence of SOWM impairments includes poor performance on SRC and impaired OP relative to IP (henceforth referred to as adjusted order span). Patient M.B. exhibited impaired verbal SOWM with both digits (SRC = 3.08) and letters (SRC = 2.45, p’s < .05) as well as with the probe tasks (verbal adjusted order span = 50%, p = .03) when compared to ten age- and education-matched control subjects, whereas the other two patients showed no evidence of a verbal SOWM impairment. To test the single SOWM hypothesis, patients were given the same WM battery with nonverbal materials (visuospatial locations). None of the patients showed nonverbal SOWM impairments. Most critically, M.B. exhibited SOWM deficit in the verbal domain but not the nonverbal domain (SRC = 5.33, nonverbal adjusted order span = 59%, p’s > .05). This dissociation between verbal and nonverbal SOWM systems provides clear evidence against the single SOWM hypothesis. To test the hypothesis that verbal SOWM and language production share a common order processing capacity, we next analyzed phoneme/grapheme errors in a corpus of repetition and spelling tasks. Errors in these tasks were categorized into item, order, and omission errors. M.B. produced significantly more phoneme and grapheme order errors than would be expected by chance (p’s < .001 in Monte Carlo simulation), whereas the other two patients did not have a tendency to make order errors in these tasks. The association between M.B.’s order impairment in verbal WM, spoken production and spelling, along with the fact that E.V. and S.H. do not show order impairments in any of those domains, suggests a common serial order mechanism. In sum, our multiple-case study suggests a common mechanism for serial order between verbal WM and the language production system that is distinct from the serial order mechanism used in nonverbal WM.