Item reliability for the Griffiths scales of mental development.

As a preliminary to the revision of the Griffiths scales, the 498 items constituting the original test were studied in terms of inter-observer agreement over the scoring, using videotapes of tests, given to children aged from 2 months to 7 years, and large scoring panels. Fifty-seven items failed to be administered or scored adequately often. One hundred and twelve were found to be unreliable on one in three administrations and only 23 were unreliable on two of three administrations. Among items scored sufficiently often, 88% of all administrations were reliable. Subscale differences were not as expected since the locomotor, personal--social and hearing and speech scales fared no worse than hand--eye coordination. Inter-observer agreement varied with the age of the children. Subscale weaknesses were examined in terms of the age ranges most concerned.