The response of twenty latency-aged children to residential treatment was measured by serial child-care ratings. Those who did not respond early made no gains during the latter half of hospitalization. A satisfactory outcome was best predicted by the total rating change within the first six months as a percentage of the treatment interval. Treatment success appeared to depend upon the extent to which the program modified the child's customary defences and behaviour within that period, irrespective of whether this resulted in an increase or reduction in disturbance. Possible reasons for lack of early change are discussed.