Rats fed a semipurified diet supplemented with 3% (w/w) safflower oil [Saf, n-3 fatty acid deficient, high linoleic acid (18:2n-6)] through two generations exhibit decreased correct response ratios in a brightness-discrimination learning test compared with rats fed 3% perilla oil [Per, high alpha-linolenic acid (18:3n-3)]. This is associated with a decreased DHA (22:6n-3)-to-arachidonic acid (20:4n-6) ratio in brain lipids. In the first set of experiments, dietary oil was shifted from Saf to a mixture of 2.4% safflower oil plus 0.6% DHA after weaning (Saf-DHA), but all parameters measured in the learning test were essentially unchanged. Brain 22:6n-3 content of the Saf-DHA group reached that of the Per group but the levels of 20:4n-6 and docosatetraenoic acid (22:4n-6) did not decrease to those of the Per group at the start of the test. In the second set of experiments, dietary oil was shifted to a mixture of 0.6% safflower oil plus 1.2% oleic acid (OA) plus 1.2% DHA (Saf-OA-DHA group) with 18:2n-6 content comparable to that of the Per group. The Saf-OA-DHA group exhibited a learning performance similar to that of the Per group; brain 22:6n-3, 20:4n-6, and 22:4n-6 contents were also comparable to those of the Per group. These results indicate that the altered learning behavior associated with a long-term n-3 fatty acid deficiency is reversed by supplementing 22:6n-3 after weaning, when the levels of competing n-6 fatty acids in the diet and brain lipids are limited.