IN THE FIRST STUDY, 60 INTRODUCTORY PSYCHOLOGY STUDENTS (24 MALES AND 36 FEMALES) WERE DIVIDED INTO 20 THREE-PERSON SAME-SEX GROUPS. EACH SUBJECT WAS GIVEN A BOOKLET CONTAINING A SUMMARIZED TRIAL TRANSCRIPT AND A SERIES OF QUESTIONNAIRES. SUBJECTS WERE ASKED TO SELECT A PRISON SENTENCE OF 1 TO 20 YEARS FOR THE DEFENDANT DESCRIBED IN THE TRANSCRIPT AND TO RECORD THEIR IMPRESSIONS OF THE DEFENDANT ON 13 BIPOLAR ADJECTIVES. GROUPS THEN ENGAGED IN 15-MINUTE DISCUSSIONS TO REACH A UNANIMOUS JUDGMENT ON THE DEFENDANT'S SENTENCE AND RELATED QUESTIONS. SUBJECTS WERE RETESTED AFTER DISCUSSION ON THEIR INDIVIDUAL SENTENCING JUDGMENTS AND IMPRESSIONS OF THE DEFENDANT AND WERE ASKED TO COMPLETE A 22-ITEM PUBLIC OPINION SCALE (A MEASURE OF AUTHORITARIANISM). A FAIRLY CONSISTENT PATTERN OF SHIFTS TOWARD LENIENCY IN SENTENCING JUDGMENTS FOLLOWING DISCUSSION EMERGED. THE ONLY CONDITION THAT FAILED TO PRODUCE A LENIENCY SHIFT WAS ONE IN WHICH FEMALE SUBJECTS JUDGED A DEFENDANT WHO WAS DESCRIBED IN A NEGATIVE MANNER. THIS COMBINATION ALSO PRODUCED THE HIGHEST INITIAL SENTENCE OF ANY CONDITION. THE REPLICABILITY OF LENIENCY SHIFTS OBSERVED IN THE FIRST STUDY WAS TESTED IN THE SECOND STUDY. USING A LARGE SAMPLE AND BOTH DEFENDANT-SALIENT AND VICTIM-SALIENT CONDITIONS (TRANSCRIPTS EMPHASIZING EITHER DEFENDENT CHARACTERISTICS OR VICTIM CHARACTERISTICS, AN INTERACTION BETWEEN THE SALIENCE VARIABLE AND DELIBERATION HAS PREDICTED SUCH THAT LENIENCY SHIFTS WOULD OCCUR ONLY IN DEFENDANT-SALIENT CONDITIONS. TH SUBJECTS INCLUDED 160 INTRODUCTORY PSYCHOLOGY STUDENTS (80 MALES 80 FEMALES) WHO FORMED 32 FIVE-PERSON SAME-SEX GROUPS. PROCEDURES USED IN THE SECOND STUDY WERE SIMILAR TO THESE IN THE FIRST STUDY. A SIGNIFICANT EFFECT FOR TYPE OF SALIENCE WAS REVEALED; SUBJECTS IN VICTIM-SALIENT GROUPS WERE MORE PUNITIVE THAN SUBJECTS IN DEFENDANT-SALIENT GROUPS. LENIENCY SHIFTS WERE OBSERVED, HOWEVER, AT BOTH LEVELS OF SALIENCE. SEX HAD A MARGINALLY SIGNIFICANT EFFECT ON PREDISCUSSION SENTENCING, WITH FEMALE SENTENCES SUBSTANTIALLY HIGHER THAN MALE SENTENCES. DECISION RULE DID NOT AFFECT SENTENCING AND DID NOT FIGURE IN ANY SIGNIFICANT INTERACTION WITH DELIBERATION. THE RESULTS ARE COMPATIBLE WITH THE POLARIZATION HYPOTHESIS OF MOSCOVICI AND ZAVALLONI. REFERENCES ARE INCLUDED. (DEP)