OBJECTIVE To evaluate the hypothesis that breast-feeding women who participate in relaxation training will have increased secretory IgA (sIgA) levels in their breast milk compared with women not receiving training. METHODS Nonrandomized control trial of a convenience sample. METHODS Women were recruited from the postpartum floor of a university teaching hospital. The intervention took place in the women's homes. METHODS Women in the first 48 hours after delivery who were planning to breast-feed their healthy newborn infants for at least 8 weeks were approached for enrollment. Women were excluded if they had previous experience with relaxation training. At 4 to 6 weeks postpartum, we enrolled 38 women still breast-feeding their infants. METHODS Women were allocated into 3 groups. Women in group 1 were taught relaxation and had breast milk samples collected before and after the teaching. Women in group 2 had conversation with similar breast milk sample collection, and women in group 3 had 1 breast milk sample collected. Women in group 1 were encouraged to practice the relaxation once or twice a day for 2 weeks, and a second visit was made to all mothers with repeated breast milk collections. Women who were still breast-feeding at 6 to 8 weeks after study end had a final breast milk sample collected. Breast milk was analyzed for secretory IgA levels. Stress was assayed using the Symptom Checklist-90-R and open-ended questions. RESULTS There was no difference in sIgA levels among the 3 groups at any time. Women who reported stress present between visit 1 and visit 2 increased their sIgA levels at the final sample collection (+0.16 g/L) compared with women who reported no stress (-0.09 g/L; P= .03). The ratings of success in relaxation in women in group 1 were related to the following sIgA levels in sample 4: poor relaxation, 0.67 g/L; fair relaxation, 0.41 g/L; good relaxation, 0.35 g/L; and very good, 0.30 g/L (P= .006). CONCLUSIONS Self-reported stress appears to increase breast milk sIgA levels. Success at relaxation was inversely related to sIgA levels in the group learning relaxation.