After randomly dividing a sample of 246 suicide gesture naval recruits & 246 controls into validation & cross-validation subsamples, a predictive cluster of 21 statements from the Recruit Temperament Survey was developed on the validation sample. Biographical variables & the RTS scale were entered into a multiple regression program. The RTS cluster was the most powerful predictor of suicide gesture, & this scale was found to be effective in differentiating suicide gesture recruits from controls.