Should I label all scale points or just the end points for attitudinal questions?

Abstract

The decision about whether to label all response scale points or just the end points for attitudinal questions can be an important and vexing decision for question designers. This short article will first discuss theoretical and practical considerations that should help guide the decision about how to label response scale points. Next, I will discuss some of the important empirical findings on this question from the literature. I will also provide some advice about how to evaluate verbal labels. The amount of clarity that the labels add to the response scale is the most important consideration in the decision to label scale points. The approach to labeling should be the one that most clearly defines the response scale for respondents. One might argue that labeling of all scale points might offer an advantage in this regard. Several authors concede that it is probably more natural for a person to express his or her opinion using words (Fowler 1995; Krosnick and Fabrigar 1997). However, there is inherent ambiguity in the verbal labels that are frequently used with response scales. For example, people might have different interpretations of what it means to " somewhat favor " a public policy. Conversely, one might argue that even though numbers might be more abstract for most respondents, they might also be more accurate. For example, it has been noted that numbers in response scales at least convey the idea of equal intervals between points on a response scale (Krosnick and Fabrigar 1997). However, respondents can also vary in how they interpret numbers. Schwarz et al. (1991) present evidence that respondents interpret 10 point scales from –5 to +5 quite differently than 10 point scales ranging from 1 to 10. They found that negative numbers imply the opposite of something, whereas the low end of a scale with all positive numbers merely implies the absence of something. There are many other aspects of the research design that might influence the decision to label scale points. For example, the length of the scale will determine the feasibility of labeling all points. It will be much easier to create labels for 5 point scales than it will be for 11 point scales. As Fowler (1995) writes, " it is difficult to think up adjectives for more than 5 or 6 points along most continua. " The mode of the interview also influences whether or not all scale points can be labeled. Generally, the use of telephone interviewing encourages shorter scales so that the respondent does not have to listen to a long list of response options before answering a question. Furthermore, when longer scales are used in telephone surveys it is more common to label only the endpoints of the scale (Dillman et al. 2008). Data collection methodologies that utilize visual modes of communication can more easily incorporate a full set of labels for response scales.