Premature infants, visiting parents, and staff in two intensive care units (ICUs) were observed around the clock for 3 weeks to study the early experiences of these infants: types of medical and general care, adult to infant personal attentions, interactions, infant activities (by sensory mode), and factors affecting these patterns of care. Infant case histories included family, medical, and environmental data such as number of days on oxygen or in an incubator. In-depth interviews and epidemiological assessment of psychiatric symptomatology of persons identified in the observations as primary care providers were used to study the psychodynamics of the ICU environment. Data were analyzed longitudinally for each infant and cross-sectionally to identify trends along environmental, sensory mode, and interaction variables. For example, despite increases in respiratory capacity, infant vocalization continued to decrease significantly as the number of days increased with mechanical ventilation, in oxyhoods, incubators, and even in an ICU environment. In effect, initial suppression of vocalization for medical reasons was subsequently reinforced by the ICU environment. The amount of time parents spent with their infants significantly correlated inversely to the number of miles between home and hospital (range of 5 to 180 miles) and dropped to zero for infants transported more than 50 miles (three of 10 such infants became wards of the State during hospitalization). Staff-infant contact for medical care was 10.6% of observation time (12 30-minute observations per infant per day), 10.9% for general care, 15.4% personal attention simultaneous with care tasks, plus 3.8% additional personal attention. These extrapolate to 6.1 hours of infant contact per day. Personal attention correlated significantly with medical care, and at even a higher level with general care, personal attention independent of tasks, and staff rank order of preference for individual infants, thus differentiating between the nurturing care of preferred infants and the more impersonal care provided to infants lower in staff preference. Factors affecting preference were long term prognosis, family-infant and staff-family relations, staff sense of success or failure from working with an infant, familiarity, ethical issues, and the physical and perceptual accessibility of the infant. Persons identified in the observations as primary care providers exhibited a significantly higher degree of psychiatric symptomatology (psychosocial dysfunction, depression, general psychopathology, and psychosomatic symptomatology) than a representative sample of the general population. This symptomatology had links to their attachment to the high risk infants, to the care provider's life history, and to reasons for liking their work. Such findings indicate a need to expand the environmental model used to study the early experiences of high risk premature infants.