A SAMPLE OF 48 SUBJECTS WAS DRAWN FROM INMATES AT THE UTAH STATE PRISON AND FROM COURT-COMMITTED INSTITUTIONALIZED PATIENTS AT THE FORENSIC UNIT OF UTAH STATE HOSPITAL. CLINICAL PSYCHOLOGISTS SELECTED 12 SUBJECTS FOR EACH OF THE FOLLOWING GROUPS: 'NORMAL' NONPSYCHOTIC, LATENT SCHIZOPHRENIC, RESIDUAL SCHIZOPHRENIC, AND SCHIZOPHRENIC-PSYCHOTIC. ALL SUBJECTS WERE MALE AND RANGED IN AGE FROM 19 TO 46 YEARS. SUBJECTS WERE ADMINISTERED THE RORSCHACH TWICE, ONCE WITH THE INSTRUCTION TO RESPOND AS THOUGH THEY WERE NORMAL WELL-ADJUSTD, AND ONCE WITH THE INSTRUCTION TO RESPOND AS IF THEY WERE SERIOUSLY MENTALLY ILL-PSYCHOTIC. A COUNTERBALANCED DOUBLE-BLIND DESIGN WAS USED. A INDEPENDENT CLINICAL PSYCHOLOGIST WAS TO DETERMINE WHICH PROTOCOL WAS OBTAINED UNDER THE INSTRUCTION 'APPEAR NORMAL' AND WHICH PROTOCOL WAS OBTAINED UNDER THE INSTRUCTION 'APPEAR MENTALLY ILL.' THE JUDGE WAS ABLE TO SORT PROTOCOLS INTO THE TWO CATEGORIES WITH AN 80 PERCENT ACCURACY RATE. A FISHER'S Z TEST CONFIRMED THE SATISFACTORY ABILITY OF THE JUDGE TO ASSESS PROTOCOLS AND SORT THEM AS PER INSTRUCTIONS, AND A CHI SQUARE TEST SHOWED THAT THE JUDGE'S ABILITY TO SORT THE PROTOCOLS DID NOT VARY SIGNIFICANTLY FROM ONE DIAGNOSTIC GROUP TO ANOTHER. OVERALL, THE VARIABLES, RATIOS, AND 'GROTESQUE' RESPONSES (INVOLVING THEMES OF DEPRESSION, CONFUSION, SEX, AND BLOOD) WERE SIGNIFICANTLY ALTERED AS A RESULT OF VARIED INSTRUCTIONS. THESE RESULTS INDICATE THE VULNERABILITY OF CERTAIN ASPECTS OF THE ROSCHACH TEST AS A RESULT OF SITUATIONAL INFLUENCES, BUT CONSISTENT WITH PREVIOUS STUDIES, A NUMBER OF OTHER VARIABLES CONSIDERED MORE STABLE DID NOT CHANGE AS A RESULT OF INSTRUCTIONS. TABLES AND 17 REFERENCES ARE INCLUDED.