The Development of Social Measurement

Abstract

Social measurement is the one which shows the most conspicuous development in the recent twenty years among research methods. The instruments of social measurement contains two types ; (1) scales and (2) indexes. Scales take the form of linear function of items and indexes do not. In this paper we are concerned with scales only. As we know, scales should have reliability and validity, and there are rather few obstacles about reliability. As for validity of scales, the situation differs among predicting scales and internal scales. Predicting scales, e. g. marital adjustment predicting scale, parole prediction scales, Warner's I. S. C., which have external criterion to be predicted, can use the correlation coefficient (or corresponding summarizing mesaures) between the external criterion and the scale scores, as a coefficient of validity. By internal scales I mean the ones which have not an external criterion, and are examplified in socio-economic status scales and attitude measurement scales in general. Internal scales, which have internal consistency, are considered at present time also to have validity. Therefore, scaling theories of internal scales have endevored to obtain internal consistency. The first scholar who gave attention to internal consistency and carefully constructed internal scales, was L. L. Thurstone. After his equal-appearing interval scale, there appeared Likert's scale which uses item analysis as a means of obtaining internal consistency. But item analysis could not secure complete internal consistency, as R. F. Sletto and L. Guttman indicated. Before the invention of Guttman's scale, factor analysis was considered the only means of securing internal consistency. Guttman's scale analysis was a splendid and epoch-making scaling theory, but the fact that his scalability is a special case of internal consistency, was discovered in the light of L, azarsfeld's latent structure theory. This latent structure theory, which introduced the concept of trace line and defined unidimensionality definitely by mathematical formula, offered a common theoretical foundation to various scaling techniques. But it is not yet completed as a new technique of scaling. We expect that it will be further developed in the future. Meanwhile, does internal consistency guarantee the validity of an internal scale? I cannot but say “no, ” because it is only a formal property. Still there remains the question concerning what the scale measures. Unfortunately at present, we have no way except to rely upon face-validity or validating the scale through long pragmatic processes. There is another important problem to be solved. Should all internal scales have an internal consistency? Does the situation exist where a multi-dimesional internal scale is necesitated? If this is the case, such a third scale would involve a future problem in my judgment.