Kelley, Truman L.

views updated

Kelley, Truman L.

Truman Lee Kelley (1884–1961) was highly influential in the introduction of statistical methods into psychological studies. His work also gave considerable impetus to the rise of psychometrics within the field of psychology.

Kelley was born in Whitehall, Muskegon County, Michigan. His early interests led him to study mathematics at the University of Illinois, where he received his A.B. in 1909. His first career in mathematics, although short-lived (he was an instructor of mathematics at the Georgia Institute of Technology, 1909–1910, and in a high school and junior college in Fresno, California, 1911–1912), became the keystone of his later career, which started when he turned to the new field of psychometrics. He received an A.M. in psychology from the University of Illinois in 1911 and his PH.D. from Columbia University in 1914. His mentors at Columbia were Edward L. Thorndike and Robert S. Wood worth. It was through the medium of psychometrics that Kelley carried forward into education and guidance the kind of work that Thorndike and Woodworth had done in experimental psychology. His later studies of the bearing of capability, accomplishment, and interest on the practical affairs of men took form while he was still a graduate student; it was in this period that he wrote Educational Guidance (1914).

The political and economic turmoil that coincided with Kelley’s years at the University of Texas, 1914–1917, and at Teachers College, Columbia University, 1917–1920, made certain government authorities more receptive to the application of statistics to practical affairs in education and work. Kelley pursued his interest in such application through service as psychological consultant to the Committee on Classification of Personnel, United States Army, and to the Surgeon General’s Office. In these appointments Kelley was able to use his ideas on the statistical treatment of data to predict, by means of psychological tests, the performance of men on one job or another.

Kelley moved to Stanford University in 1920 as assistant professor of education. He became professor of education and psychology in 1926. Lewis M. Terman, a colleague at Stanford, joined Kelley and Giles M. Ruch in preparing and publishing the Stanford Achievement Test Battery, a series that has remained in good repute throughout several revisions over forty years. This series of achievement tests was a step toward Kelley’s general goal of providing a practical system of assessing and describing the performance of men during the years of minimum required education.

Kelley’s Statistical Method (1923 a), an influential book, encapsulated much of his prior work in statistics for educational and psychological purposes. The text was well grounded in the statistics of correlation and regression. Kelley elaborated these basic ideas with his own emphases on the normal distribution, curve fitting, time series, standard errors, and multiple and partial correlation. At this time he was already very much concerned with enabling others to apply and compute statistics accurately, a concern that led eventually to his publication of The Kelley Statistical Tables (1938). All these themes were given further consideration in later publications; they were to be the major topics of his work.

Kelley usually considered statistics and applications simultaneously. The interaction led to his reputation in psychometrics; his still classic Interpretation of Educational Measurements (1927) clearly shows evidence of this interaction. The book contains Kelley’s ideas about the reliability and validity of test scores and offers evaluations of then existing tests. Other psychometricians have been quick to appreciate the importance of publishing professional judgments about mental and educational tests, and their subsequent work has superseded Kelley’s earlier evaluations of specific tests. Nevertheless, his theory of reliability endures today in much of its original outline.

Kelley developed his theory of the multidimensionality of intellect in interrelation with Spearman’s work on a single general factor of intelligence. He extended Spearman’s tetrad criteria for determining the number of factors needed to explain relations between variables to pentad criteria (1928). [SeeIntelligence and intelligence testing.] His belief in a multidimensional intellect also carried him forward to provide his solution of the principal components problem in factor analysis in Essential Traits of Mental Life (1935). This book was published during his tenure as professor of education in the Graduate School of Education at Harvard University, a tenure which began in 1931 and lasted until his retirement in 1950.

At about the same time that Kelley published a general solution for the principal axes problem, Harold Hotelling, with whom Kelley consulted, also published such a solution [seeFactor analysis]. Kelley, unlike Hotelling and others, insisted that judgments of the importance of mental functions should be incorporated in an analysis of principal axes when the analysis is originally designed. If a group of people is tested on a set of tests, a series of scores for every person who took the set of tests is obtained. Psychometricians are interested in the factor structure of the abilities, which gives rise to the observed correlations among the several tests contained in the set. Most psychologists, and particularly L. L. Thurstone, preferred to ascertain this structure from the set of observations alone. The Thurstone school claimed that this makes the structure empirical and that it can therefore be tempered at the user’s discretion with the knowledge of how he has modified and analyzed the information. Kelley, on the other hand, claimed that his procedure would focus attention on the importance of psychological processes as well as on the accuracy of statistical analysis and that Thurstone’s method might produce many accurate but inconsequential factors. Although Kelley insisted that one be aware of the problem of accuracy, he valued usefulness above accuracy.

Kelley insisted on completeness of the correlation matrix if work in education was to be solidly grounded. He favored a system of principal components; Thurstone, a system of primary mental abilities. Kelley analyzed the total variance of the matrix of test intercorrelations (as augmented even by the psychologist’s judgments); Thurstone analyzed only the communalities (or commonness) of test correlations. Whereas Thurstone advocated the rotation of what he called “simple structure,” Kelley advocated the rotation of test axes in orthogonal terms. Kelley’s incorporation of judgment in his conception of science is evident in his series of lectures published as Scientific Method (1929). These ideas endured in his efforts to define the components of mental functioning by simultaneous analysis of sets of criteria and sets of predictors. This technique, known as canonical correlation analysis, appeared in Talents and Tasks (1940). Hotelling again published a similar solution at about the same time, but in his solution he did not incorporate judgments of importance as did Kelley.

Kelley’s constant concern with making psychometric methods useful led him, in a series of papers on psychophysical scaling, to try to define a unit of measurement appropriate for everyday tasks in schools and personnel departments. When others were no longer interested in the concept Kelley continued to regard the “just noticeable difference” useful in this connection. In 1940 he contributed the idea of a ridge route norm, a general way of establishing a norm for a grade with control for age. These norms became a part of the Stanford Achievement Tests, as did the K (for Kelley) scores derived by his student and eventual coauthor, Eric F. Gardner.

Kelley extended his series of achievement measurements for use in education with Tests and Measurements in the Social Sciences (1934), of which he was coauthor. His other test, outlined in the Activity Preference Report (1954), grew from a project commissioned by the National Defense Research Committee during World War II, a project to which Kelley devoted his time while also serving as consultant to the secretary of war.

Kelley was active in professional organizations as well as in scholarship. He served as president of the Psychometric Society in 1938–1939 and also as vice-president of the American Statistical Association in 1926 and Section Q of the American Association for the Advancement of Science in 1928. He was a cofounder of the national honorary education society, Kappa Delta Pi.

David V. Tiedeman

[For the historical context of Kelley’s work, see the biographies ofThorndike; Thurstone; Woodworth. For discussion of the subsequent development of his ideas, seePsychometrics.]

WORKS BY KELLEY

1914 Educational Guidance: An Experimental Study in the Analysis and Prediction of Ability of High School Pupils. New York: Columbia University, Teachers College.

(1923 a) 1924 Statistical Method. New York: Macmillan.

1923 b Kelley, Truman L.; Ruch, G. M.; and Terman, L. M. Stanford Achievement Test: Test Forms A and B. Yonkers-on-Hudson, N.Y.: World Book.

1926 The Influence of Nurture Upon Native Differences. New York: Macmillan.

1927 Interpretation of Educational Measurements. Yonkers-on-Hudson, N.Y.: World Book.

1928 Crossroads in the Mind of Man: A Study of Differentiable Mental Abilities. Stanford (Calif.) Univ. Press.

(1929) 1932 Scientific Method: Its Function in Research and in Education. New York: Macmillan.

1934 Kelley, Truman L.; and Krey, August C. Tests and Measurements in the Social Sciences. New York: Scribner.

1935 Essential Traits of Mental Life. Cambridge, Mass.: Harvard Univ. Press.

(1938) 1948 The Kelley Statistical Tables. Rev. ed. Cambridge, Mass.: Harvard Univ. Press.

1940 Talents and Tasks: Their Conjunction in a Democracy for Wholesome Living and National Defense. Cambridge, Mass.: Harvard Univ. Press.

1947 Fundamentals of Statistics. Cambridge, Mass.: Harvard Univ. Press.

1954 Activity Preference Report. Cambridge, Mass.: Educational Research Corporation.