The effect of different classroom seating arrangements on children's on-task behaviour was examined by observations of two top junior classes of ten- to eleven-year-old children. In both classes mean on-task behaviour was lower for the first two weeks in which children sat around tables compared with the second two-week period spent in rows, and declined again when the children returned to tables for the final two weeks. An analysis of the data broken down into groups of children with low, average and high initial on-task behaviour showed that the rows condition had its most powerful effect on children with low initial on-task behaviour. There was little difference between conditions for the children with high initial on-task behaviour. The results for the average groups of children were similar to those for the classes as a whole. Rising baselines during the first tables condition confused the effect but there was clear evidence for declines in on-task behaviour in the final tables condition.