The purpose of this study was to explore cross-situational consistency in moral reasoning. Based on Kohlberg's developmental theory, which hypothesizes within-subject consistency of moral reasoning, adults' moral rationales were analyzed. The investigation centered on whether the moral reasoning utilized in response to abstract, remote, hypothetical moral dilemmas differed from the reasoning employed in more concrete situations. The design utilized paper and pencil instruments in group administered settings with 161 college students. The Ethical Reasoning Inventory and the revised Objective Assessment of Moral Development were used as the abstract and contextualized measures of moral reasoning. Respondent demographic characteristics, respondent past moral crisis experience, the format of the moral development instrument, and the contextual characteristics of the moral conflict situation were hypothesized as influencing moral reasoning. A within-subjects post-hoc analysis was undertaken utilizing an array of multivariate techniques. Discriminant function analysis was employed in building a predictive model of situationally influenced moral reasoning. Five significant findings emerged. First, a low--but significant--correlation existed between abstract and contextualized moral reasoning ability. Second, verbal ability was correlated positively with contextualized moral reasoning ability. This relationship held even when controlling for age, gender, ethnicity, marital status and past experience. Third, past moral conflict experience was inversely related to moral maturity on remote content dilemmas. Fourth, both male and female respondents were influenced by two contextual features, i.e. the gender and socioeconomic status (SES) characteristics of the moral dilemma model. Contexts involving male and female low SES models evoked less mature reasoning. Fifth, an 80% accuracy rate was obtained when using the canonical discriminant function to predict respondent level of moral maturity in contexts involving male and female low SES models.