Schizophrenic individuals (n = 31), including paranoid and nonparanoid diagnostic subgroups, and normal controls (n = 20) participated in a semantic priming experiment involving a single-choice lexical decision task. For the automatic priming blocks, a 260-ms stimulus onset asynchrony (SOA) was used; for the controlled priming blocks, a 1,000-ms SOA was used. The paranoid subgroup showed significantly less priming than did the control group. The nonparanoid subgroup showed a decrease in priming compared with the control group that approached significance. There was an increased priming effect for the controlled compared with the automatic priming condition; this difference was not modulated by participant group. Nonsignificant semantic priming (equal to 0) occurred only for schizophrenic subgroups and only in automatic priming conditions.