To the Editor: In a recent article, Heather M. Conklin, B.S., and colleagues (1) reported that schizophrenia patients as well as their first-degree, nonafflicted relatives showed impairment on the backward digit recall portion of the Wechsler Digit Span task, but relatives did as well as nonpsychiatric comparison subjects on digits forward. The authors concluded that these results support a diathesis in schizophrenia for verbal working memory deficits. In pointing out that verbal working memory impairments may be an endophenotypic marker, like impaired smooth-pursuit eye movements, they unintentionally implied that this may be a fixed and untreatable feature of this disorder. However, our work and that of a few other investigators have suggested that it is possible to remediate these impairments.

More than 25 years ago, investigators began reporting that some cognitive impairments in schizophrenia could improve with training. The possibility of restoring lost elemental function through neurocognitive retraining has gained support from studies of brain plasticity in animals and recently from direct applications of training methods with patients (2, 3).

In a current U.S. Department of Veterans Affairs study, we used the Wechsler Digit Span task in the pre- and postneuropsychological assessment of a 6-month cognitive remediation and work rehabilitation program. Thirty-one subjects with schizophrenia or schizoaffective disorder were randomly assigned to neurocognitive training while participating in a work therapy program, and 34 received work therapy alone. Memory training was performed by using computer-based exercises designed for patients with traumatic brain injury but were modified to our specifications. Subjects began reading short lists of words or digits presented on-screen, followed by cues to retrieve the information. In a hierarchically arranged progression, tasks were made more difficult by shortening the exposure time, increasing the latency period, and increasing the number of items. Subjects had no direct training on digits backward, a task that requires manipulating remembered numbers so that they can be repeated in reverse order.

Results indicated improvement in the Wechsler Digit Span scaled score for the training group (paired t test=2.96, df=30, p<0.01), and an analysis of covariance revealed significantly greater improvement than in the control group (F=3.63, df=2, 61, p<0.05). Moreover, the most dramatic results occurred in the digits backward task. More than 60% of the subjects receiving training showed significant effect size improvements (0.2 standard deviations or greater) compared with 25% of those in the work-only condition (χ²=6.81, df=1, p<0.01), and 40% had large effect size improvements (0.8 standard deviations or greater) compared with 18% for those who received work therapy only. Thus, the neurocognitive training of memory had improved verbal working memory, generalizing from our training procedures to the unpracticed task of digits backward.

We are encouraged by these findings and expect refinements in remediation methods in the future. Although we have no disagreement with Ms. Conklin et al. regarding a diathesis for verbal working memory impairment in schizophrenia, we do not want to equate genetic vulnerability with irreversibility. Our findings suggest that remediation of verbal working memory deficits is possible in schizophrenia.