Self-Reported Metrics - Measuring the User Experience: Collecting, Analyzing, and Presenting Usability Metrics - page 127

situations a neutral reaction is a perfectly valid reaction and should be
allowed on a rating scale. So in most cases we use rating scales with an
odd number of points. However, there's some indication that not includ-
ing a midpoint may minimize the effect of the social desirability bias in
a face-to-face administration of rating scales (e.g., Garland, 1991).

• Total number of points. The other issue, of course, is the actual number

of points to use on the rating scales. Some people seem to believe “more
is always better,” but we don't really agree with that. The survey literature
suggests that any more than nine points rarely provides useful additional
information (e.g., Cox, 1980; Friedman & Friedman, 1986). In practice,
we almost always use five or seven points.

IS FIVE POINTS ENOUGH FOR A RATING SCALE?

Kraig Finstad (2010) did an interesting study comparing five- and seven-point versions
of the same set of rating scales [the System Usability Scale (SUS), discussed later in this
chapter]. The ratings were administered orally. He counted the number of times that the
participant answered with an “interpolation,” such as 3.5, 3½, or “between 3 and 4.”
In other words, the participant wanted to pick a value between two of the values given
on the scale. He found that participants using the five-point version of the scale were
significantly more likely to use interpolations than those using the seven-point version.
In fact, about 3% of the individual ratings on the five-point scale were interpolations,
while none of the ratings on the seven-point scale were. This would suggest that verbal
(and perhaps paper-based) rating scales, where the participant could be tempted to use
interpolations, might yield more accurate results with seven-point scales.

SHOULD YOU NUMBER SCALE VALUES?

One of the issues that comes up in designing rating scales is whether to show the user a
numeric value for each scale position. Our sense is that with scales of no more than five
or seven values, adding numbers for each position is not necessary. But as you increase
the number of scale values, numbers might become more useful in helping the user keep
track of where she or he is on the scale. But don't use something like −3, −2, −1, 0, +1,
+2, +3. Studies have shown that people tend to avoid using zero or negative values (e.g.,
Sangster et al., 2001; Schwartz et al., 1991).

6.2.7 Analyzing Rating-Scale Data
The most common technique for analyzing data from rating scales is to assign
a numeric value to each of the scale positions and then compute the aver-
ages. For example, in the case of a five-point Likert scale, you might assign a
value of 1 to the “Strongly Disagree” end of the scale and a value of “5” to
the “Strongly Agree” end. These averages can then be compared across different