Retinal function was assessed by electroretinogram in 32 neonates randomly assigned to formulas of different ω-3 fatty acid content and in 10 infants fed human milk. All neonates had a birth weight of 1000-1500 g and were fed study diets from d 10 to 45 or discharge. Group A received formula containing predominantly 18:2 ω-6. Group B received a balanced mix of 18:2 ω-6 and 18:3 ω- 3. Group C was given a formula containing both essential fatty acids and supplemented with marine oil to provide 22:6 ω-3 content similar to that of human milk. The fatty acid composition of plasma and red blood cell (RBC) lipids were similar for all groups on entry but marked dietinduced differences were found after feeding the study diets. Group C was comparable to the human milk-fed group, but group A had lower 22:6 ω-3 and w-3 long-chain polyunsaturated fatty acids (LCPUFA) in plasma and RBC membranes. Cone function was not affected by dietary essential fatty acids. Rod electroretinogram thresholds were significantly higher for group A relative to the human milk-fed group and group C and significantly correlated with RBC w-3 LCPUFA (r = -0.63, p < 0.0001); 44% of the variance could be explained by RBC and plasma w-3 LCPUFA content. Rod electroretinogram amplitude was significantly lower for group A relative to the human milkfed group and group C and related to plasma 22:6 w-3 (r = 0.55) and total w-3 LCPUFA (r = 0.58) (both p < 0.0001); 42% of the variance was explained by plasma w- 3 LCPUFA, the ratio of ω-6/ω-3 LCPUFA in RBC, and gestational age at birth. Our results support an essential role for ω-3 fatty acids in retinal development.