Feeding behaviour in the weaning period is important theoretically and practically. The aim of the study was to develop appropriate observational codes for its description, to assess their reliability, and to examine the relationships between feeding behaviour, meal duration, and food intake. One hundred children aged 12–14 months were visited in their own homes, and two of each child's usual meals video-recorded and analysed using direct observation. Codes were developed that distinguished between the mother feeding her child directly and assisting her child's self-feeding, and between the child's behaviour when responding to being fed by the mother and when feeding themselves. All-occurrence sampling was used to record counts of these feeding acts during the meals. Two observers replicated coding of 40 randomly chosen meals to determine the reliability of these counts. Except for three codes which were used very infrequently (median counts of zero over the 200 meals), reliability was high with ρ > ·80. There was wide variation in the extent to which individual children fed themselves during meals, with only moderate consistency from meal to meal. Food intake was uncorrelated with meal duration, but correlated with the number of bites of food taken. Adjusted for the number of bites, longer meals were associated with a lower intake. When fed by the mother the child's food intake was greater than when they fed themselves, but the duration of the meals was little affected. The coding scheme is simple to use and generally reliable, and provides a means for relating more global measures of emotional or other characteristics of mealtime behaviour to feeding behaviour and nutritional intake.