Another Look at Likert Scales

Abstract

Social scientists have produced an extensive literature on the nature and social correlates of public attitudes. The findings from some of these studies may both aid in social decision-making in a democratic society and contribute to our understanding of the sources of human behavior. Early scholars assumed attitudes were not an acceptable area of scientific inquiry since attitudes cannot be observed directly and therefore need to be inferred or deduced from individuals' actions. Thurstone (1928) challenged this position in his paper entitled Attitudes Can Be Measured.Although the methods suggested by Thurstone were cumbersome and are seldom used today (Thurstone and Chave 1929), his work was quickly followed by that of others. One method, developed in 1932 as a doctoral dissertation in psychology at Columbia University, has come to dominate current attitude measurement (Likert 1932). This approach consisted of asking subjects to indicate the extent to which they agreed or disagreed with each of a series of statements related to the focus of the desired attitude. The resulting answers were then scored and summed to yield a composite value used to index the respondents' attitudes toward the topic of interest. Although the developer of this simple, pragmatic method for measuring attitudes went on to have a distinguished academic career as a renowned survey statistician, cofounder and director of the Survey Research Center at the University of Michigan, president of the American Statistical Association, and leader in cutting-edge work on participative business management practices (Seashore 1982), he is most often identified today for his development of this attitude scaling procedure. His name was Rensis (lick-urt) and his methods are called Likert Scaling.CHARACTERISTICS OF LIKERT SCALESLikert consist of a series of related - statements concerning a specific referent, namely the focus of the attitude to be measured (Desselle 2005; 1932). A balance of both positive and negative items is generally recommended to reduce response-set bias. Subjects indicate their feelings concerning each item on a bipolar scale such as agree, agree, undecided, disagree, and strongly disagree. Responses for each subject are scored from one (1) to five (5), with negative items reverse-coded. The scores for the individual items are then summed to obtain a Summated Rating Score or Scale value for each respondent. Alternatively, the mean scores of the responses of each subject can be used so that the scale scores fall in the same 1 to 5 range as the individual items. Although these five category response alternatives are common, three, four, six, seven, and more have also been used. Factor analysis (Flora and Curran, 2004) and/or item analysis, including item-to-item, item-to-total correlations and/or reliability measures such as Cronbach's Alpha (Cronbach 1951), may be used to assess the extent to which the separate items are assessing a single attitude dimension.Although Likert-type items and Scales have been widely adopted throughout the social science research communities, the method is not without controversy. Through the years, the procedures involved in their derivation and use have been the subjects of debate by social science methodologists, psychometric scholars, and applied researchers concerning the number of necessary items, the number, and nature of the response categories, and the uses of the summated and item scores. The purpose of this paper is to explore the meanings and implications of these various issues and, by doing so, contribute to ongoing dialogue in this area. Specifically, the following three issues are addressed:1) How many Likert-type items are needed for a attitude scale?2) What and how many response categories should be presented?3) What are the meanings of the obtained responses? How can they be analyzed? …