The concept of bizarre delusions figures prominently in the diagnosis of schizophrenia under DSM-III-R criteria. Yet this concept may be difficult to apply in the clinical world. The interrater reliability of distinguishing bizarre versus non-bizarre delusions was estimated by compiling a sample of 40 delusional ideas and asking a group of psychiatrists (ranging from experts in the field to residents) to identify them as bizarre or non-bizarre. The interrater reliability of the 45 respondents was consistently low as measured by kappa statistic (kappa less than 0.40) within and across groups of raters, and regardless of whether or not a structured definition was applied. Implications for DSM-IV are discussed.