The apparent retroconversion of docosahexaenoic acid (22:6n-3) to eicosapentaenoic acid (20:5n-3) and docosapentaenoic acid (22:5n-3) was studied in vivo, in rats and humans, after they ingested a single dose of triacylglycerols containing [13C]22:6n-3 ([13C]22:6-triacylglycerol), without 22:6n-3 dietary supplementation. The amount of apparent retroconversion and the distribution of the three n-3 polyunsaturated fatty acids (PUFAs) in plasma lipid classes were followed as a function of time by measuring the appearance of 13C in these PUFAs with gas-chromatography combustion-isotope ratio mass spectrometry. This [13C]22:6n-3 retroconversion, calculated by summing the amounts of [13C]22:5n-3 and [13C]20:5n-3 in plasma lipids, was lower in humans than in rats, reaching a maximum of approximately 9% of the total plasma [13C]22:6n-3 in rats, but only 1.4% in humans. The incorporation of [13C]22:6n-3 and [13C]22:5n-3 in lipid classes followed their endogenous distribution with a maximal accumulation in phospholipids, but a low incorporation into cholesterol esters (CEs), whereas [13C]20:5n-3 was equally present in phospholipids and CEs. The ratio of the amount of HDL-CE to HDL-phosphatidylcholine for [13C]20:5n-3 was higher than for [13C]22:6n-3, indicating a selectivity of the lecithin-cholesterol acyltransferase enzyme with regard to these PUFAs, which may be related to the differences in their biological properties after fish oil feeding. The occurrence of a weak basal 22:6n-3 retroconversion in humans supports feeding this pure PUFA in cases in which 20:5n-3 presents undesirable side effects and when specific alterations of blood lipids are expected.