We conducted a prospective longitudinal study of fetal size in rural Papua New Guinea (PNG) involving 439 ultrasound-dated singleton pregnancies with no obvious risk factors for growth restriction. Sonographically estimated fetal weights (EFWs; N = 788) and birth weights (N = 376) were included in a second-order polynomial regression model (optimal fit) to generate fetal weight centiles. Means for specific fetal biometric measurements were also estimated. Fetal weight centiles from a healthy PNG cohort were consistently lower than those derived from Caucasian and Congolese populations, which overestimated the proportion of fetuses measuring small for gestational age (SGA; < 10th centile). Tanzanian and global reference centiles (Caucasian weight reference adapted to our PNG cohort) were more similar to those observed in our cohort, but the global reference underestimated SGA. Individual biometric measurements did not differ significantly from other cohorts. In rural PNG, a locally derived nomogram may be most appropriate for detection of SGA fetuses.