Abstract

Social scientists regularly make use of multivariate models to describe complex social phenomena. It is argued that this approach is useful for modelling the variety of cognitive and social factors contributing to scientific change, and superior to the integrated models of scientific change currently available. It is also argued that care needs to be taken in drawing normative conclusions: cognitive factors are not instrinsically more "rational" than social factors, nor is it likely that social factors, by some "invisible hand of reason," generally work to produce scientific success. A multivariate model of the biasing factors within a scientific community at particular times is developed. This model, which is an example of work in social epistemology, yields normative conclusions.