Likert Scaling

Likert's Approach

In Likert's original research, which led to Likert scaling, Likert compared four ways of structuring attitude survey items believing that there was an alternative to the approach attributed to Louis Leon Thurstone. Although both approaches were based on equal-interval, ordinal stepped scale points, Likert considered Thurstone's methods to be a great deal of work that was not necessary. Setting up a Thurstone scale involved the use of judges to evaluate statements to be included in the survey. This included rank ordering the statements in terms of the expected degree of the attribute being assessed and then comparing and ordering each pair of item possibilities, which is an onerous task if there were many item possibilities. Originally, each item was scored as a dichotomy (agree/ disagree or +/—).

A Thurstone scale was scored in a similar manner as Likert's original method using sigma values, which were z scores weighted by the responses corresponding to the assumed equal interval categories. However, part of the problem with Thurstone's scoring method related to having a spread of judge-determined 1 to 11 scoring categories when scoring the extreme values of 0 or 1 proportions. These could not be adequately accounted for because they were considered as ± infinity z values in a sigma scoring approach and thus were dropped from the scoring. Likert felt [Page 715]there was another approach that did not rely so much on the use of judges and could include the scoring of items where everyone either did not or did select the extreme score category by using the ± 3.00 z values instead of ± ∞. Thus, Likert set out to use some of the features of a Thurstone scale but simplify the process and hope to achieve a similar level of reliability found with the Thurstone scale. His research met the goals he set out to meet.

A stem or statement was presented related to racial attitudes and then respondents were asked to respond to one of several response sets. One set used “yes / no” options, and another used narrative statements, and two of them used what we now know as a Likert item, using strongly approve—approve—undecided—disapprove— strongly disapprove as the response categories. The distinction between the last two types of items relates to the source of the questions as being developed specifically by Likert for assessing attitudes and the other as abbreviations of newspaper articles reflecting societal conflicts among race-based groups.

...