Database Reference
In-Depth Information
SIDEBAR: ODD OR EVEN NUMBER OF OPTIONS?—cont'd
Those who believe it is a good thing argue that being able to focus on a speciic midvalue makes
it easier for the responder and allows for more accurate assessments.
The proponents of having an even number argue that an odd number of points provides a
responder with the ability to not have to commit himself/herself either way, while an even number
of choices forces the responder to commit himself/herself to one side of the midpoint or the other—
and argue that this is a good thing.
Another thing to consider is this: there is a cognitive bias in human behavior called
“anchoring and adjustment” (e.g.,
Tversky and Kahneman, 1974
). This has been demonstrated
over and over again and, in essence, says that when people focus on one value, it becomes an
“anchor” (whether voluntarily or not, and a speciic midpoint, such as a “3” on a 1-5 scale,
is an anchor), which retards movement. Thus, a responder does not, on average, move away
from the anchor as much as might be chosen were there no anchors. Thus, the extremes are not
chosen as much as they should be because responders stay closer to the anchor more than they
should.
However, “anchoring and adjustment” applies more when there are a larger number of points
among which to choose (e.g., a 7- or 9-point scale). With only 5 points, the “anchoring and adjust-
ment” bias is less prominent, and perhaps, immaterial.
We always use an odd number of points in our scales because we feel that a neutral response
is always possible, and denying that kind of neutral choice is akin to “leading the witness”—
something we never try to do.
A typical set of end labels is “Strongly Disagree” and “Strongly Agree,” like the
following:
StronglyDisagree
12345
StronglyAgree
The respondent simply circles (or checks) the number that indicates his/her level
of agreement with the statement. Sometimes, an open text ield is provided for the
respondent to add some comments regarding the score.
In practice, Likert scales (and other types of rating scales, like the System Usabil-
ity Scale) are a vital part of your UX researcher tool kit. They provide you with
quantiiable data about the user's perception of the system, design, ease of task, or
anything else, as opposed to what you (or anyone else on your team) think those
perceptions are.
In addition, Likert scales can provide vital complementary data that can either
corroborate or negate other indings. For example, a post-test ease-of-use rating for
a particular task may be 4.6 (on a 5-point scale), and this would complement a task-
completion rate of 90%. With these two values, you are probably fairly conident that
the task is not problematic for your target audience. Conversely, a post-test ease-of-
use rating for a particular task may be 1.2 (on a 5-point scale), and this complements
a task-completion rate of (only) 20%. With these two values, you are probably fairly
conident you've got a problem on your hands.
Search WWH ::
Custom Search