The relationship between levels of moral reasoning and decisions in dilemmas of neonatal care was investigated in a sample of 452 pediatricians. Subjects included residents, faculty members, and practitioners recruited from a variety of university-affiliated and community hospitals. It was hypothesized that physicians whose moral reasoning was more fully developed would less actively treat particular cases. Such cases might include those where a patient's family requested such a limit (designated "negative family attitude") or the quality of life likely to be led after therapy was so low as to preclude even a minimal degree of human activity or social interaction (designated "unsalvageable prognosis"). The hypothesis was tested through the use of two questionnaires. The first questionnaire, devised by Crane, assessed the physician's reported degree of activism in treating six cases of infants born with severe defects. The structure of moral reasoning was measured by a second questionnaire, Rest's Defining Issue Test. Subjects were scored by the degree to which they use universal, ethical principles in resolving a series of moral dilemmas. Results of the absolute level of activism (Table 1) showed that among both residents and postresidents, the degree to which cases are actively treated depends, for salvageable patients, on the type of damage and on the possibility for research. Results involving moral reasoning showed a different pattern among residents and postresidents. Among residents, a significant correlation exists between principled reasoning and the absence of active treatment (r = - 0.41, Form A; r = - 0.23, Form B). As predicted, such correlations were strongest for cases of negative family attitude or of unsalvageable prognosis. The pattern of correlations among postresidents showed either no relationship to moral reasoning or the reverse of the residency pattern (r = -0.08, Form A; r = 0.30, Form B). The influence of the type of institution a resident operates within was assessed by analysis of variance. Inasmuch as moral reasoning and institutional type both had significant main effects (Form A), their magnitude differed. Institutional type accounted for 43% of the variation in mean activism scores whereas moral reasoning accounted for only 4%; however, because one could, a priori, expect institutional norms and customs to be powerful determinants of behavior, any additional, identifiable influence deserves attention. The structure of individuals' moral reasoning seems to be such an influence.