Howard Wainer Educational Testing Service
David Thissen University of North Carolina at Chapel Hill
Psychology is full of an easy complexity. Physical sciences gain in truth when most quantitative, whereas psychology, when most quantitative has been of least scope, though the sweetness of elegance lingers on. Over the century since Wundt's laboratory started modern experimental psychology, its characteristic paradigms have emerged in fits and starts. In an attempt to deal with its frequently crude data, psychology has adopted, adapted, and polished a set of statistical procedures and models that are second in sophistication and complexity only to those employed by Carlyle's "dismal science." The often self- conscious attempt to make paradigms statistically rigorous when experimental controls lacked rigor was in keeping with comments of statisticians such as Wallis ( 1949), who argued that "So-called 'high-powered,' 'refined,' or 'elaborate' statistical techniques are generally called for when the data are crude and inadequate--exactly the opposite, if I may be permitted an obiter dictum, of what crude and inadequate statisticians usually think" p. 471.
Of course, there were many reasons other than self-conscious defensiveness for the development and use of high-powered statistical techniques. Often such techniques were desperately required. This need provided the impetus for the development of rigorous, formal statistical models, but de-emphasized the need for descriptive methods: Methods whose aims were exploratory rather than confirmatory.
The past 30 years have seen a vigorous development of methods, by the foremost of statisticians, that have as their very nature an informality that was unheard of in the social sciences for a half century. These techniques are pri-