THE last report on factor analysis in the REVIEW OF EDUCATIONAL RESEARCH (26) covered the period [1939][1940][1941][1942][1943][1944][1945][1946][1947][1948][1949][1950][1951]. Altho the period covered by the present survey is much shorter, the number of papers reviewed is nearly as great as that for the previous period. This accelerated research activity emphasizes the present interest in the state of the art. Examination of the literature reveals that most of the research is on the mechanics of fac tor analysis as a research technic or its application in some subject area. However, the new research does not overlook the origins of factor analysis, for it also includes some attempts at evaluating the validity of factor analysis as a mathematical model to describe observed human behavioral responses in terms of operationally meaningful human parameters.
Validity of Factor AnalysisFactor analysis, as we know it today, emerged as a research technic thru the creation of a mathematical model which attempted to portray the relationships between a response to a test of mental ability and the human factors which produced it. In view of recent emphasis on model building in the behavioral sciences it is interesting to note that the factor analysts were already at work in model building some 50 years ago. The model builders of today have the same goal as the original factor analysts; namely, the creation of a model which both reproduces and explains observed data. Altho there are many factor models which reproduce observed data (and some which do not), it remains an open question as to which models satisfy both conditions.
New ModelsThe term radex theory is used by Guttman (84) to describe a new model which attempts both to reproduce and to explain a correlation matrix developed from test scores of mental ability. Two distinct notions are involved in a radex. One is that of a difference in kind between tests, and the other is that of a difference in degree. Each of these notions gives rise to a simple order system; for instance, for all tests of the same kind, say numerical ability, there will be differences in the degree of their com plexity; such a set of test variables of the same kind is called a simplex. Correspondingly, all tests of the same degree of complexity differ among themselves only in the kind of ability they define. Since here a law of order Many articles on various aspects of methodology relate to (a) compari son of methods, (b) new twists on old methods, (c) efficiency of comput ing, and (d) statistical developments arising out of the complex sampling distribution theory one has come to expect in factor analysis.
ComparisonsFactorial methods have been compared by a number of investigators. Moursy (112) discussed three types of factorial organization: general and