The case is made for basing analysis of longitudinal measurement data on each individual's array or curve of scores, and employing methods of analysis which take account of shape of each individual's curve. DQs at 6 months and 18 months, and IQs at 3, 5, 8, 11, 14, and 17 years were available on 84 subjects from 6 months to 17 years, and 109 subjects from 6 months to 14 years, from a London longitudinal project. Scores at each age were converted to SD scores with M = 0, SD = 1. Polynomial equations were fitted to each subject's curve over periods: 6 months − 17 years; 3 years − 17 years; 3 years − 11 years. For whole sample fitted curves yielded highly significant error reduction across all periods, thus rendering untenable doctrine of IQ constancy. Over 6 months to 14 years, 54 per cent of subjects’ curves had a significant fit. Curves differ not only in linear slopes, but in degrees and forms of curvilinearity, particularly over longer periods. A visual method of classification of curves is also presented, which proved effective.
Implications of findings are discussed, including that changes in IQ scores cannot be regarded as random variation, but as following systematic trends which differ between subjects; need to regard isolated IQ measures on a subject as derived from a curve of unknown shape and slope; and dangers of treating as equivalent IQs of subjects at different ages. It is concluded that present methods of analysis are applicable to a variety of longitudinal measurement data.