395 disorders responsive to folic acid therapy. Can Med Assoc J 1976; 115: 217-22. 19. Reynolds EH. Folic acid and neuropsychiatry. Farmaci Terapia 1985; 2: 163-68. 20. Carney MWP, Sheffield BF. Associations of subnormal serum folate and vitamin B12 and effects of replacement therapy. J Nerv Ment Dis 1970; 150: 404-12. 21. Reynolds EH. Anticonvulsant drugs, folate metabolism and mental symptoms. In: Dam M, Gram L, Penry JK, eds. Advances in epileptology. XIIth Epilepsy International Symposium. New York: Raven Press, 1981: 621-25. 22. Botez MI, Botez T, Leveille J, Bielmann P, Cadotte M. Neuropsychological correlates of folic acid deficiency: facts and hypotheses. In: Botez MI, Reynolds EH, eds. Folic acid in neurology, psychiatry and internal medicine. New York: Raven Press, 1979: 435-61. 23. Coppen A, Abou-Saleh MT. Plasma folate and affective morbidity during long-term lithium therapy. Br J Psychiatry 1982; 141: 87-89. 24. Coppen A, Chaudhury S, Swade C. Folic acid enhances lithium prophylaxis. J Affective Disord 1986; 10: 9-13. 25. Agnoli A, Andreoli V, Casacchia M, Cerbo R. Effect of S-adenosyl L-methionine (SAMe) upon depressive symptoms. J Psychiatr Res 1976; 13: 43-54. 26. Janicak PG, Lipinski J, Davies JM, et al. S-adenosylmethionine in depression: a literature review and preliminary report. Alabama J Med Sci 1988; 25: 306-13. 27. Bell KM, Plon L, Bunney WE, Potkin SG. S-adenosylmethionine treatment of depression; a controlled clinical trial. Am J Psychiatry 1988; 145: 1110-14. 28. Reynolds EH, Stramentinoli G. Folic acid, S-adenosylmethionine and affective disorder. Psychol Med 1983; 13: 705-10. Do changes in pattern of breast usage alter the baby’s nutrient intake? Twelve mother/baby pairs took part in a study of the difference in effect of two patterns of breast feeding—either feeding at one breast or at two breasts during a feed. Baseline measures were taken at 4 weeks, and the test patterns of feeding were followed for a week each, in random order. The two patterns of feeding led to differences in milk volume intake and mean feed fat concentration, but not in the baby’s net fat intake per 24 h. The results indicate that the breast-fed baby can regulate his fat intake quickly and thus mothers should be encouraged to practise "baby- led" feeding. Introduction The secretory output of human mammary epithelial cells is consistent when studied in vitro.1 However, since milk is a stored product, there is scope for inconsistency between the composition of the milk synthesised and that which the infant derives by suckling. Thus, during the course of a breast-feed there is a conspicuous increase in the fat concentration of the milk.2-4 Hytten’s plausible explanation for this pattern of change was based on a physical model of breast emptying;2 no secretory explanation has yet been proposed. It may be inferred from Hytten’s proposal that the more completely the breast is emptied, the higher is the fat concentration of the milk-a view supported by experiments showing that the fat concentration at the end of a breast-feed is directly related to feed length and feed volume 5 This finding underlies the hypothesis6 that with incomplete breast emptying (as might be caused by premature termination of feeds or poor positioning of the baby at the breast) the milk that the infant receives will, within weeks, become progressively lower in fat, so that the baby could become calorie deprived and fail to thrive. Here we describe how we tested this hypothesis in a within- subject, repeated measures, randomised, crossover study in which each mother acted as her own control. This design is sensitive to small experimental effects. The study has implications for the estimation of the breast-fed baby’s energy intake: if completeness of breast emptying affects the milk fat concentration, then estimates of nutrient intake will be influenced by feed management policies that affect breast emptying. Methods Subjects Twelve mother/baby pairs provided complete data during the baseline and both experimental periods. The mothers (seven primiparous, five multiparous) were enlisted through the National Childbirth Trust. The babies (eight female, four male) were all born at term, and their mean birthweight was 3-465 kg (SD 035). Data collection Breast milk output was assessed by weighing the infant before and after every feed taken during a period of 24 h (8 am to 8 am). The babies were weighed with a ’Sartorius 3826-MP8’ balance with data-averaging keyboard (74401-AWK) (Sartorius Ltd, Epsom). Milk samples expressed by mothers (at least 5 drops) before and after the feed (from the breast offered over the 12 hours from 8 am to 8 pm were analysed for fat content) by use of the creamatocrit technique by one investigator (J. C. I.), coefficient of variation for repeated samples was 1 .%. The fat concentration of the milk taken at each feed was defined as the mean of the pre-feed and post-feed fat values. This definition is based on the assumption that breast milk fat changes in a linear manner during the feed. The theoretical issues underlying this assumption have been addressed previously4,7 and it is generally accepted to be a reasonable approximation.8,9 An adjustment for non-linearity has recently been suggested,1O but when applied to the present data it did not materially affect the outcome (largely due to the within-subject study design employed). Net fat intake at a feed was taken as the product of feed volume and the fat concentration of the milk taken at that feed; total daytime (8 am-8 pm) fat intake was taken as the sum total of the individual feed values. Night-time (8 pm-8 am) fat intake was estimated as the product of the night-time volume intake and the mean of all ADDRESS Institute of Child Health, University of Bristol, Royal Hospital for Sick Children, St Michael’s Hill, Bristol BS2 8BJ, UK (M.W Woolridge, DPhil, J C Ingram, PhD, Prof J. D Baum, FRCP) Correspondence to Dr M. W. W