The ability of Plasmodium falciparum isolates from 103 Papua New Guinea children with cerebral malaria and 158 children with uncomplicated malaria to form rosettes in vitro was studied. Of these, 81 isolates from cerebral malaria and 151 isolates from uncomplicated malaria grew to schizogony and were included in the rosetting analysis. Wide variation occurred in the level of rosette formation, with all isolates from both cerebral and uncomplicated malaria patients being able to form rosettes. No statistically significant difference existed between the geometric mean rosetting rate of isolates obtained from cerebral malaria and those from uncomplicated malaria (9% versus 8.6%, P = 0.27). The ability of acute sera to inhibit rosette formation was not significantly different between 18 cerebral malaria cases and 20 controls tested [mean reduction in rosetting rate 6.1% (SD 11.5) versus 8.4% (SD 12.3), P = 0.57]. The rosetting rate of cerebral malaria cases was not associated with the clinical outcome. Among the clinical and laboratory variables tested, only blood group and parasite density were significantly associated with rosetting. These data do not support the hypothesis that rosette formation is associated with cerebral malaria in Papua New Guinea, but indicate that rosetting is an intrinsic property of parasites occurring in all manifestations of the disease.