Item response biases on the logical memory delayed recognition subtest of the Wechsler Memory Scale-III

We hypothesized that the wording and sequential order of the WMS-III Logical Memory recognition questions may bias subjects toward correct or incorrect responses on specific items. Therefore, we classified each item according to one of three potential sources of bias (yeasaying to proper names, priming of "yes" responses by previous items with similar content, naysaying to unlikely occurrences) and administered the items to 31 subjects who were completely naïve to the story content. The items predicted to have correct endorsement biases were answered correctly at greater than chance frequencies, while items predicted to be biased toward incorrect answers were missed more frequently than expected by chance. The same sources of bias were tested in an independent clinical sample of 36 neurological patients who were administered the WMS-III in the standard manner. In these patients biases appeared robust enough to be detected in the performances of clinical patients during a neuropsychological evaluation. With further research, such biases may provide avenues for detecting malingering.