Likert Scale

The Likert (or summated rating) scale is a very popular device for measuring people’s attitudes, beliefs, emotions, feelings, perceptions, personality characteristics, and other psychological constructs. It allows people to indicate their position on items along a quantitative continuum. Three features characterize this sort of scale. First, there are multiple items that are combined into a total score for each respondent. Second, each item has a stem that is a statement concerning some object (person, place, or thing). This might concern an aspect of the person completing the SCALE or an aspect of another person or object. Third, the person completing the scale is asked to choose one of several response choices that vary along a continuum. They can be unipolar, ranging from low to high, or bipolar, ranging from extreme negative to extreme positive. These responses are quantified (e.g., from [Page 573]1 to 6) and summed across items, yielding an interpretable quantitative score.

The development of a Likert scale generally involves five steps (Spector, 1992). First, the construct of interest is carefully defined to clarify its nature and boundaries. Second, format decisions are made and the items are written. Third, the scale is pilot tested on a small sample to be sure the items are clear and make sense. Fourth, the scale is administered to a sample of 100 to 200 (or more) individuals. An item analysis is conducted to help choose those items that are intercorrelated with one another and form an internally consistent scale. Coefficient alpha is computed as a measure of internal consistency reliability with a target of at least .70 (Nunnally, 1978), although .80 to .90 is more desirable. Finally, research is conducted to provide data on the scale’s CONSTRUCT VALIDITY to be sure it is reasonable to interpret it as reflecting the construct defined in Step 1.

The response choices in Likert scales can differ in number of type. The most popular form is the agree-disagree, in which respondents indicate the extent to which they agree with each stem. Usually, there are five to seven response choices—for example, disagree (agree) very much, disagree (agree) somewhat, and disagree (agree) slightly. Other common alternatives are frequency (e.g., never, seldom, sometimes, often) and evaluation (e.g., poor, fair, good, outstanding).

There has been some debate with bipolar scales about whether there should be an odd number of response choices with a neutral response in the middle (e.g., neither agree nor disagree) or an even number without the neutral response. Those advocating an odd number argue that one should not force the ambivalent person to make a choice in one direction or the other. Those advocating an even number point out that the neutral response is often misused by respondents (e.g., to indicate that the item is not applicable) and that it may encourage people to be noncommittal. There is generally little practical difference in results using even or odd numbers of response choices.