A Critical Study of the Criterion of Internal Consistency in Personality Scale Construction

Abstract

A PERSONALITY scale in its developmental stage usually contains some items that are not sufficiently discriminating to justify their retention in the final form of the scale. The criterion of internal consistency is frequently utilized to select the more discriminative items. It is essentially the use of total score on all items of a scale to evaluate each individual item. In the application of this criterion all items are scored according to uniform arbitrary scoring weights, and these total scores are placed in an array. Next, papers are segregated for persons whose total scores place them in the extremes of the array. The proportion included in these extreme segments has varied widely from one study to another; decides, quartiles, quintiles and other proportions have been employed. Whatever the proportion used, the segments drawn from each extreme of the array are equal in the number of persons they contain. Item responses are tabulated separately for the two extreme segments. The extent to which an item yields response differentiating between them is the measure of its discriminative value in the scale. The statistical significance of this discriminative value is usually determined by computing the critical ratio.' The criterion of internal consistency was recently employed to select items for a battery of six scales developed by E. A. Rundquist and R. F. Sletto.2 These scales were designed to measure morale, feelings of inferiority, family adjustment, attitude toward law, economic conservatism, and attitude toward education. Each scale contained 22 items, and five alternative responses were provided for each item. Typical items were as follows: