Summary. The findings reported here—which are part of a much larger longitudinal project—deal with the drawings of a man made by thirty‐seven children at yearly intervals throughout their junior school career, i.e., between the ages of 7 to 10 years. A total of 148 drawings was analysed according to Goodenough's instructions. The intelligence quotients obtained from the drawings were intercorrelated: (a) with each other; (b) with the Stanford‐Binet Intelligence Quotients and; (c) with W.I.S.C. Verbal, Performance and Full Scale Quotients. An analysis of the fifty‐one items of the Goodenough Scale was also carried out to investigate whether or not there is a regular, annual increase in scores over the four‐year period. The findings suggest: That for junior school children the longitudinal test‐retest reliability of the Goodenough Scale is too low for it to be regarded as a stable measure of intelligence. That the Scale's validity is even lower, average correlations with quotients obtained from the Stanford‐Binet Scale as well as from the W.I.S.C. Verbal, Performance and Full Scale being in the region of ·2. That its discriminating power between successive years is not entirely satisfactory.