We describe modifications made to the Rivermead Behavioural Memory Test (RBMT) to make it appropriate for use with children. Results of a sample of 335 children aged 5 to 10 years, attending mainstream schools, are reported. Each child was given two of the four versions of the modified test for children together with a number of tests of memory and achievement from the Wechsler Intelligence Test for Children (Revised WISC-R), and the British Ability Scales (BAS). The children's version of the behavioural memory test (RBMTC) has good interrater and parallel from reliability. The validity of the RBMTC was established in two ways. First, with other tests of memory from the WISC and BAS. Second, a small (n = 36) group of children with severe epilepsy and memory difficulties, attending a residential school, were tested. The houseparents of these children were asked to complete rating scales on everyday memory performance. There was significant agreement between houseparents' observations and scores on the RBMTC (Spearman rank correlation = .71, p < .001), indicating that the test is indeed assessing everyday memory.