1. Twenty-five northern Thai mothers, breast-feeding their infants on demand, were studied in their homes for 24 h. All breast-feeds were test-weighed and pre- and post-feed expressed breast-milk samples (0·5 ml) taken at each feed.
2. The fat concentration of milk taken during a feed showed significant circadian variation, with maximum values between 16.00 and 20.00 hours and minimum values between 04.00 and 08.00 hours. Fat concentration at the start and at the end of a feed also varied significantly over 24 h.
3. Multiple regression analysis showed that the most important predictor of fat concentration at a feed was the length of time elapsed since the previous feed – the longer this interval, the lower the subsequent fat concentration. Other significant predictors were the fat concentration at the end of the previous feed, and the milk intake at the previous and at the current feed.
4. Fat concentration declined between feeds in proportion to the length of time between feeds, but the decline was less between sleep feeds than between waking feeds. This would appear to be a reflection of the lower post-feed fat concentration and higher pre-feed fat concentration of sleep feeds compared with waking feeds, when other variables relating to feeding pattern are taken into account.
5. The larger the milk intake at a feed, the greater was the increase in fat concentration from the start to the end of the feed. The change in fat concentration was less in feeds taking place during the sleep period than in daytime feeds.