Inmates from the six federal corrections institutions of El Reno, Okla.; Atlanta, GA.; Danbury, Conn.; Texarkana, TX.; La tuna, TX.; and Fort Worth, TX; were included in the study. Measures used were illness complaint rates, disciplinary infraction rates, blood pressure, perception of crowding, sleep, inmate evaluations of housing, death rates (both violent and nonviolent), suicide rates, psychiatric commitment rates, self-mutilation, and suicide attempt rates. A questionnaire was administered to inmates and biographical information was obtained from them. Study findings supported several principal conclusions. First, high degrees of sustained crowding have a wide variety of negative psychological and physiological effects, including increased illness complaint rates, higher death and suicide rates, and higher disciplinary infraction rates. Second, large institutions produce much more severe negative psychological and physiological effects than do small institutions, as expressed in higher death, suicide, and psychiatric treatment rates. Finally, partitioning of open dormitories into privacy cubicles has a strong positive effect as indicated by the reduction or elimination of negative effects typically associated with open dormitories. The findings also indicate that there are substantial individual differences in responses to overcrowding as well as differences among racial and ethnic groups. It was also found that both the number of occupants in housing quarters (social density) and space per person (spatial density) contribute to crowding effects, with social density typically the most influential factor. It appears that once space-per-person levels of 50 square feet or higher are reached, the number of people living together and the space arrangement (single bunking, cubicling, segmenting into bays) may be the main factors determining the reaction to the housing. Recommendations concerning optimum housing arrangements policies are given and future research needs are discussed. Figures, references, and an appendix of data collection forms are provided.