Likert scaling, introduced Rensis Likert (1932, 1970), is the most widely used method of measuring personality, social, and psychological attitudes (Babbie, 1998; Nunnally, 1978). For example, prominent measures of self-esteem, depression, alienation, locus of control, ethnocentrism, racism, religiosity, spirituality, and homophobia have all used Likert scales to make operational the underlying latent construct (Hill & Hood, 1999; Raja & Stokes, 1998; Robinson, Shaver, & Wrightsman, 1991).
In addition to numerous established measures, a review of the social work literature reveals that researchers commonly use Likert scales in the development of new instruments that tap a broad array of constructs. Likert scales have been used to measure adolescent concerns that foster runaway behavior (Springer, 1998); appropriate practitioner responses to suicidal clients (Neimeyer & Bonnelle, 1997); homesickness and contentment among Asian immigrants (Shin & Abell, 1999); social worker empowerment (Frans, 1993); Spanish-speaking clients' perceptions of social work interventions in prenatal care programs (Julia, 1993); willingness to seek help (Cohen, 2000); attitudes toward illegal aliens (Ommundsen & Larsen, 1998); punishment (Chung & Bagozzi, 1997); and working with clients with HIV/AIDS (Riley & Greene, 1993).
The popularity of Likert scales can be traced to a number of factors, including ease of construction, intuitive appeal, adaptability, and usually good reliability (Babbie, 1998; Nunnally, 1978). Yet, despite these assets there are significant problems associated with Likert scales. This article delineates the problems, with a particular emphasis on multidimensionality and coarse response categories, and proposes a new measurement method called "phrase completions," which has been designed to circumvent the problems inherent in Likert scales. We also conducted an exploratory test, in which Likert items were adapted to phrase completions.
OVERVIEW OF LIKERT SCALES
In contemporary usage, Likert scales present individuals with positively or negatively stated propositions and solicit respondents' opinions about the statements through a set of response keys. Typically, participants are asked to indicate their level of agreement or disagreement with a proposition on a graded four- or five-point scale (for example, strongly disagree, disagree, agree, strongly agree). The fifth point, when used, allows for a neutral or undecided selection to be incorporated into the response key as a midpoint response option.
Agreement with a positively stated proposition is hypothesized to reveal the underlying construct. The responses are usually equated with integers (for example, strongly agree = 0, strongly disagree = 4). Negatively worded items are reverse scored. The items are summed, creating an index that indicates the degree to which the respondent exhibits the traits in question (Duncan & Stenbeck, 1987; Roberts, Laughlin, & Wedell, 1999).
Although the agree-disagree format is perhaps the most common form of Likert scale, other types of response keys also are widely used (for example, very unmotivated, moderately unmotivated, indifferent, moderately motivated, very motivated; below average, slightly below average, average, slightly above average, above average; and so forth). Instruments using these formats are sometimes referred to as Likert-type scales or more generically as rating scales. However, this basic approach, a positively or negatively stated proposition followed by a graduated response key using adverbs and verbs, is commonly understood as the distinguishing characteristic of Likert scales (Brody & Dietz, 1997; Duncan & Stenbeck, 1987).
PROBLEM OF MULTIPLE DIMENSIONS
One of the principle tenets in constructing instruments is that items be as clear and concise as possible. The more items are characterized as cognitively complex, the more likely respondents are to misunderstand the question and answer incorrectly. …