Poster C6, Friday, August 17, 10:30 am – 12:15 pm, Room 2000AB
Does the modified Sternberg task measure verbal working memory function? Evidence from experimental pharmacology
Christopher Barkley1, Zhenhong Hu2, Ann Fieberg1, Lynn Eberly1, Susan Marino;1University of Minnesota, 2University of Florida
The modified Sternberg has been widely adopted for use as a verbal working memory (VWM) task (e.g. Ruchkin, 1991). However, there is an important structural difference between the currently used version of the task and the original (Sternberg, 1966): almost universally, subjects are no longer asked to perform serial recall at the end of each trial. The absence of a requirement to encode and recall serial order information is a fundamental difference between the modified Sternberg and other VWM tasks (e.g. reading span; Daneman & Carpenter, 1980); this raises the possibility that the modified Sternberg may not be an unambiguous test of VWM. In the current study, we investigated the widespread cognitive side effects associated with two commonly prescribed drugs: Topiramate (TPM), a broad-spectrum anti-seizure drug used to treat a range of conditions including epilepsy, obesity, and migraine; and lorazepam (LZP), a benzodiazepine used to treat anxiety and sleep disorders. In addition to further characterizing the incidence and severity of these drug-induced impairments, this study allowed us to assess the construct validity of the modified Sternberg as a test of VWM. Twenty-nine subjects received TPM (100, 150, or 200 mg), LZP (2 mg), and PBO in a randomized, double-blind, crossover study design. Four hours after drug administration, subject's accuracy and reaction time (RT) were recorded while they completed the modified Sternberg (three memory loads: 1, 3, 5 syllables); .5, 2.5, and 6 hours after drug administration they completed a neuropsychological (NP) battery containing tests of: VWM (digit span), verbal fluency (semantic, phonemic), and visual information processing and psychomotor speed (Trails A and B; SDMT). The severity of drug-induced impairment was calculated by comparing each treatment to PBO ((drug-PBO)/PBO) to generate relative change scores, thus normalizing across individual differences in unimpaired performance. Linear mixed effects models of relative change scores were constructed for all VWM and NP measures and adjusted for age, gender, education, treatment order, session number, and TPM dose group. TPM caused decreases in accuracy (but not RT) on the modified Sternberg (p<.0001) and impaired performance on all NP measures (p<.0001). LZP caused decreases in accuracy (p<.0001) and increases in RT (p<.0001) on the modified Sternberg, and impaired performance on Trails A and B, and SDMT (all p<.01). There was no significant difference between LZP and PBO on digit span or verbal fluency tasks. These LZP data suggest that, absent the serial recall requirement, the modified Sternberg may not be a true VWM task. If it were, one would expect impairments on the modified Sternberg to co-occur with deficits on VWM tasks (digit span) and declines in verbal fluency. Instead, performance on these tasks was spared, while LZP caused pronounced impairments on tasks that require the rapid processing of visual information (Trails A and B, SDMT). These data are consistent with the view that the modified Sternberg should perhaps be viewed as a visual recognition memory task rather than a test of VWM. If so, modified Sternberg data and the conclusions drawn from it should be interpreted with caution.
Topic Area: Phonology and Phonological Working Memory