OBJECTIVE: To examine the contribution of fetal growth restriction ultrasound phenotypes to adverse perinatal outcomes at term. DESIGN: Retrospective population-based cohort study. SETTING: John Radcliffe Hospital, Oxford, UK where universal ultrasound at 35+1-36+6 weeks is performed. POPULATION: Congenital abnormalities and births before the scan were excluded. Singleton foetuses were categorised as five mutually exclusive phenotypes using a hierarchical approach: ISUOG fetal growth restriction (FGR), according to Delphi criteria; Constitutional small-for-gestational-age (SGA) (estimated fetal weight [EFW] < 10th centile); Appropriate-for-gestational-age (AGA) with either cerebroplacental ratio < 5th centile or umbilical artery >95th centile; AGA with slowing abdominal circumference growth velocity (ACGV < 10th centile); Normal AGA; METHODS: Univariate logistic regression was employed using normal AGA as the reference group to estimate odds ratio with 95% confidence intervals. Group differences for continuous variables were assessed using mean differences with confidence intervals through a generalised linear model. MAIN OUTCOME MEASURES: Stillbirth (SB); composite adverse outcome (CAO) (1+ of Grade 2-3 encephalopathy, cooling, ventilation > 24 h, or perinatal death); severe SGA at birth; neonatal unit admission; obstetric interventions. RESULTS: Among 45 179 pregnancies, 54 SBs (0.1%) and 253 CAOs (0.6%) occurred. Normal AGA foetuses at the 35+1-36+6 week scan accounted for 82% of all pregnancies and for 43 (79.6%) SBs and 205 (81%) with the CAO, yet only 37.3% of neonates born with severe SGA. The absolute risk of SB and CAO was similar in all groups (0.1%-0.2%). CONCLUSIONS: Term FGR and 'normal' babies have similar perinatal risks, presumably because of intervention. Despite a detection rate of 62.7% for severe SGA, most adverse outcomes occurred in pregnancies with a normal scan.