Developmentally normal twins, joined from the mid-sternum to the umbilicus in a face-to-face position, were observed in 90-minute sessions twice daily from postnatal days 12 to 16. Sleep-waking states were coded in 10-second epochs. Within-twin data were analyzed for bout lengths, interbout intervals, transition probabilities and 60-minute autocorrelation patterns. Between-twin data were studied for simultaneous occurrences of sleep-waking states and 60-minute cross-correlation patterns. Epochs per bout and interbout intervals for quiet and active sleep, active sleep epochs with rapid eye movements (REMs) and quiet and active waking states showed little difference between the twins over the study days. Probabilities of transition between quiet and active sleep and active wakefulness also showed the same patterns for each twin. However, the organization of temporal states differed markedly. Simultaneous occurrences of the same states were at or barely above chance levels, autocorrelation patterns of all states except epochs with REMs were not the same and cross-correlations indicated little influence of one twin's current state on the subsequent probability of that state's occurrence in the other twin. Thus, neither shared environment, identical genotype nor shared stimulation from co-twin motor activity appeared to synchronize the temporal organization of behavioral states in these twins. We concluded that the overall distributions of neonatal sleep-waking states seem to be determined primarily by age, whereas the temporal organization of states expresses the neonate's individuality.