controlled, suggesting that the effect of education on fertility does not act through its economic consequences, but does so 'through less tangible, cognitive channels' ( Cleland and Rodriguez 1988: 438). In general, this second approach involves adding variables that either constitute controls for the conventional socio-economic effects, or more directly reflect effects of variables closely related to diffusion mechanisms. The nature and role of local or personal networks ( Goldberg 1976; Retherford and Palmore 1983; Watkins 1991) can also provide useful insights on how diffusion takes place.
The problems with these kinds of evidence, on the other hand, are not difficult to pinpoint: the residual type is only suggestive, and the diffusion 'effect' is hardly measurable in that context, unless one is willing to define diffusion as the within- group changes, a very indirect approximation at best. The more direct approach is problematic in that in general, there are very few variables that reflect diffusion effects exclusively, and most of them can be interpreted to reflect also some aspect of socio-economic development or modernization, broadly understood. Moreover, few studies have directly addressed interactions with socio-economic or policy 'filters' that facilitate or obstruct the diffusion process. Interactions between diffusion elements and socio-economic change appear to be relevant and their incorporation seems to be necessary to be able to explain (1) the persistence of substantial pre-decline differentials for extended periods of time, as well as (2) the timing of the onset of sustained overall fertility reductions. Different types and intensities of inter-group social interaction have been recently studied by Rosero and Casterline ( 1992) in the context of theoretical diffusion models; empirical analysis on the basis of such models could open the way for better understanding of the processes involved. Diffusion effects expressed as autoregressive, endogenous feedback fertility behaviour ( Montgomery 1992) appears to be another promising and more satisfactory statistical approach.
International and within-group dynamics of fertility trends in Latin America have been examined in a comparative perspective, in order to assess the relevance of broad theoretical approaches to fertility changes; the analyses of differentials at a point in time, useful as they may be, tell only a limited portion of the story of the fertility transitions in the region. The analysis suggests that interesting processes, though still poorly understood, seem to take place across countries and population subgroups over time.
This chapter has reviewed some of the principal aspects of a diffusionist approach based on an examination of the available evidence about the changes in fertility, which is used to delineate some general characteristics of the transition processes within the region. These may be summarized as follows: at the international level, mortality and development indicators--such as per capita income, literacy, and urbanization--correlate with fertility in the direction predicted by