Recognition memory performance in schizophrenia has been shown to vary greatly across studies. To identify the conditions under which recognition memory is significantly impaired, we used a meta-analytic strategy to quantify the moderating effects of several cognitive and clinical variables. Eighty-four studies (from 1965 to July 2003) provided recognition memory data for both a schizophrenia and control group. The overall group comparison for recognition memory yielded a significant mean weighted effect size of d=0.76. Material specificity was the most significant cognitive variable found, with patients exhibiting greater impairment for figural than verbal recognition. A yes-no recognition format and auditory encoding also led to significantly greater effect sizes for recognition memory relative to forced-choice recognition tests and visual encoding, respectively. Furthermore, the effect size for recognition memory as measured by false alarm was smaller than the effect size as measured by hit rate or by d-prime and its related measures. Among clinical variables that were associated with higher effect sizes, chronicity was the most significant, but different trends linking poor performance to negative symptoms and general symptomatology were also observed. Thus, a recognition memory deficit moderated by both cognitive and clinical variables is clearly present in schizophrenia.