Jeff sent me the following off-list, with permission to send it to the list
if I wanted... and since he made a good point, I thought I would...
At 11:35 AM 10/25/2000 +0000, Jeffrey Nagelbush wrote:
>Marc said (among other things):
>>
>Are the two groups using the scale
>>in the same way? My feeling is that when a participant approaches a scale
>>like this they form an idea in their mind that represents the mid-point.
>>They then use this imaginary mid-point to determine how they respond. Not
>>only could there be differences in interprtation between groups, there
>>could be lots of variation within a group... and hence lots of noise and
>>error in our measurements.
>>
>Call me naive about these scales (which I am) but if differences WITHIN
>groups may be due to different ideas as to what the labels mean, then how
>can we interpret these scales at all?
For me this is a question that I've struggled with for a while now. If you
do not know if people interpret the scale in the same way how can you draw
any conclusions from it? Basically, we seem to make an assumption that
people within a group interpret the scale in the same way. (Borrowing Ken's
example, all males use the scale in the same way to judge the level of
violence in a video game.) And if we are going to make this assumption,
then we might as well make the assumption that everyone interprets the
scale in the same way because it all depends on how you choose to define
the "group". Is it based on gender (males and females interpret the scale
differently)? Or could it be based on, say, major (that psychology majors
interpret the scale differently than fine arts majors)? Well, now we have a
case where males and females interpret the scale in the same way because
they are in the same major (group), but males differ from other males and
females differ from other females (variation across the same groups
thought to be consistent in the first situation).
And this is just one of many unresolved issues about Likert scaling.
Experts can't agree on how many categories you should have (5-10 seems to
be the most common though). Nor can they decide whether or not you should
have a neutral category and/or a "does not apply" option. And then you get
to the issue of how many points on the scale should you label (does every
point get a label, or just the end points, or perhaps the end points and
the middle category... if you use an odd number.) This is then topped off
by the rather heated debates about whether the data can be treated as
interval, or if you are really only dealing with ordinal information.
We have general rules of thumb to provide answers for all of these
questions in practice, but from a technical standpoint there is still no
clear cut answer that I've encountered. Perhaps someone who has spent more
time dealing with these issues can let us know more and perhaps provide
some more information.
- Marc
PS- Despite my uncertainty about what these scales tell us, I do still use
them on a regular basis. But, I am always careful in my interpretation of
the findings, keeping in mind that we don't know for sure whether or not
people use the scales in the same way.
G. Marc Turner, MEd
Lecturer & Head of Computer Operations
Department of Psychology
Southwest Texas State University
San Marcos, TX 78666
phone: (512)245-2526
email: [EMAIL PROTECTED]
