3 Analyzing Developmental Change

Any cross-sectional study is a slice at one time point across the continuing ontogenetic dialogue between the organism and the environment. While such studies help us understand outcomes, they may not tell us much about the process of ``becoming''. For example, the longitudinal genetic study involving repeated measures of twins may be thought of as a multivariate genetic study in which the multiple occasions of measurement correspond to multiple traits in the conventional cross-sectional study. In the conventional multivariate study we ask such questions as ``How much do genes create the correlation between different variables?'', so in the longitudinal genetic study we ask ``How far do genes (or environment) account for the developmental consistency of behavior?'' and ``To what extent are there specific genetic and environmental effects expressed at each point in time?''. These are but two of a rich variety of questions which can be addressed with the methods we shall describe. One indication of the insight that can ensue from such an approach to longitudinal measures on twins comes from some of the data on cognitive growth obtained in the ground-breaking Louisville Twin Study. In a reanalysis by model fitting methods, Eaves et al. (1986) concluded that such data as had been published strongly suggested the involvement of a single common set of genes which were active from birth to adolescence and whose affects persisted and accumulated through time. By contrast, the shared environment kept changing during development. That is, parents who provided a better environment at one age did not necessarily do so at another, even though whatever they did had fairly persistent effects. The unique environment of the individual, however, was age-specific and very ephemeral in its effect. Such a model, based as it was on only that part of the data available in print, may not stand the test of more detailed scrutiny. Our aim here is not so much to defend a particular model for cognitive development as to indicate that a model fitting approach to longitudinal kinship data can lead to many important insights about the developmental process.