Introduction

There are two underlying ideas in the empirical studies of most psychologists and social scientists that deserve critical attention. The first idea is that the phenomenon that is being explained is determined by certain factors, not directly but through the mediation of certain mechanisms. This means we must first decide which are the dependent variables that represent the phenomenon we want to explain and which are the independent variables from which we will try to find our explanation. We then go on to examine whether there are any correlations between those two categories of variables. If such correlations exist, then it is assumed that they reflect some sort of universal laws, psychological, social, or biological (Figure 7.1). These laws exert their influence through mediating mechanisms, which are represented by intervening variables.

Among the earliest techniques that were developed for the analysis of complicated relationships between variables were factor analysis, regression analysis, and Lazarsfeld's (1955) method of elaboration. Today it is possible, by means of covariance structure analysis, to combine into one analysis the different steps for which the psychologist of the 1950s needed factor and regression analyses (Schoenberg, 1989) or, by means of log-linear models, to correct the technical deficiencies of the elaboration procedure of the 1960s (Hagernaas, 1990, p. 23). In methodological terms, however, these new and elegant techniques introduced nothing new compared to the idea described in Figure 7.1.