Except perhaps for a few diehards, no sociologist today would think of conducting a large-scale empirical investigation of numerical data without a firm grasp of statistical methods – or at least a graduate assistant who has such a grasp. The old polemics about statistics in social research have died away and are now lost to view under a blizzard of statistical findings in journal articles and books. Asked to date the period when statistical analysis really ‘took over’ in social research, most sociologists would probably put the date around 1960, when large computers and efficient programs became widely available. But the date of widespread adoption is perhaps not as significant as the date on which the leaders in the field became convinced of the desirability of using powerful statistical methods in the analysis of standardised data.

To some extent the answers that one gives to this question depend on when and where one was trained. Those sociologists educated at Columbia in the last twenty-five years are likely to date the advent of powerful statistical procedures at the arrival of Paul F. Lazarsfeld in the United States, but this, I believe, is a mistake. Indeed, Lazarsfeld himself argued against such tools as multiple regression for some time after they became widely used and easily available on the computer, and he credits Stouffer with the same views (Lazarsfeld in Stouffer 1962: XVI).