Manic (n = 26), schizophrenic (n = 26), and normal (n = 25) subjects were examined with a digit span distraction task and with a reality monitoring task. All subjects were tested twice at a 4-day interval, and a clinical assessment of thought disorder was conducted both times on the patients. We found that reality monitoring, distraction task performance, and clinical thought disorder were all quite stable at the retest interval. We further found that different patterns of correlational relationships between cognitive deficits and positive and negative thought disorders were present in the manic and schizophrenic samples. When we conducted a cross-temporal analysis of our data, we found that no cognitive deficits in mania predicted the severity of positive thought disorder over time, although the severity of thought disorder predicted distraction performance over time. In the schizophrenic subjects, distraction performance, but not reality monitoring, exerted a significant cross-temporal influence on positive thought disorder.