Abstract There have been very few systematic observation studies of children's behaviour in British infant schools. This paper provides a descriptive account of children's behaviour over the three years of their infant schooling, when aged five to seven years. The results come from a longitudinal study of children entering reception classes in 33 Inner London Education Authority (ILEA) schools in September 1982. The observation sample consisted of four children from each school, two of whom had parents of Afro‐Caribbean origin and two who had white indigenous parents. As far as possible each pair comprised a boy and a girl, and were matched by nursery scores. The children were observed with a systematic observation schedule that describes classroom behaviour in relation to teachers, other children and on their own. Each child was observed with the five‐minute schedule (divided into continuous time intervals), six times on five observation days during the first year, six times on three separate days during the second year and nine times on two separate days during the third year. Inter‐observer agreements on observation categories were high, as were stabilities for the whole sample for each year. Most time was spent ‘on‐task’ (more than half of all observations), in individual task activities and in language work. Mathematical activities occurred relatively infrequently but, along with language work, increased over the three years. Children were rarely observed working together cooperatively on task activities ‐ despite the importance placed on this by educationists. From the child's point of view contact with teachers tended to be in a class situation, where children tended to be passive listeners. A measure of the child's active role in contact with teachers, e. g. asking questions, showed a decrease over the three years. Individual instruction‐‐recommended by the Plowden Report ‐ was found to occur infrequently, a result that matches recent findings in junior schools. ‘Off‐task’ behaviour occurred in less than 10 per cent of all observations and, contrary to a recent survey and press reports, disruptive behaviour in the shape of aggression between children and inappropriate behaviour towards teachers was very rare. Teacher behaviours ‐‐‐ often identified by North American researchers as of educational significance, e.g. criticism, praise, discipline, etc. ‐ also occurred rarely. It is acknowledged that systematic observations can only paint a picture of the prevalence of classroom behaviour, and it is difficult to interpret low‐frequency behaviours that may have important effects.