Web site Editor's Note:
This e-mail was compiled by Professor Rob Foels from a discussion on the SPSP List Serve, dealing with the issue of whether one should use unipolar or bipolar response scales when constructing a measure. The original message was dated Thursday, 25 Aug 2005. It is posted with his permission (8/26/05).
Unipolar vs. Bipolar Response Scales
Compiled in an e-mail by Rob Foels
A couple months ago I asked for cognitive explanations for differences in unipolar and bipolar scale responses. Here is my belated summary, thanks to all who responded. I would suggest anyone who uses rating scales (i.e., all of us) read Schwarz (1991) and Gannon & Ostrom (1996), these provide enlightening insights into the importance of scale construction (both numerical and endpoint labels). We should probably be putting a lot more thought and pilot work into our measurements.
-- Original message --
I've got an intriguing finding regarding ratings of others and the type of measurement scale used, results are basically opposite in unipolar and bipolar measurement. I've got a reference with a similar result (Locke, 2002), but I'm looking for a cognitive explanation of why this might occur, e.g. simultaneous (bipolar) vs. sequential (unipolar) processing. Does anyone know of cognitive discussions related to unipolar and bipolar measurement? I'll summarize any responses.
Responses from the list:
_______
Schwarz, N. (1991). Rating scales: Numeric values may change the meaning of scale labels. Public Opinion Quarterly, 55(4), 570-582.
ABSTRACT: Three experiments with 1,155 Ss examined whether the numeric values provided as part of a rating scale may influence respondents' interpretation of the endpoint labels. In Exp 1, Ss rated their success in life on an 11-point scale. When the scale ranged from 0 to 10, 34% of the Ss endorsed values between 0 and 5. However, only 13% endorsed values between -5 and 0 when the scale ranged from -5 to +5. Exp 2 provided a replication of this finding, and Exp 3 demonstrated that recipients of a respondent's report draw different inferences from formally equivalent but numerically different values. Findings indicate that respondents use the numeric values to disambiguate the meaning of the scale labels. (PsycINFO Database Record (c) 2004 APA, all rights reserved)
_______
I suggest that you email Ulrich Schimmack at U Toronto Mississauga about this issue, as he has relevant findings.
_______
You might want to take a look at Norbert Schwarz's work on bipolar vs. unipolar scales (see reference below). The basic notion in this research is that unipolar scales induce participants to think of the presence vs. absence of a particular attribute (e.g., presence vs. absence of positive characteristics), whereas bipolar scales induce participants to think of the relative proportion of two opposite attributes (e.g., relative proportion of positive versus negative attributes). Similar to findings in research on framing effects, participants may selectively consider positive attributes in the unipolar case, thus leading to more favorable evaluations of the first in comparison to the second object. However, they may consider both positive and negative attributes in the bipolar case, leading to more favorable evaluations of the second in comparison to the first object.
Another useful reference may be Solomon (1978), who discussed bipolar vs. unipolar assessment of dispositional vs. situational attributions (see reference below).
Schwarz et al. (1991). Rating scales: Numeric values may change the meaning of scale labels. Public Opinion Quarterly, 55, 570-582.
Solomon (1978). Measuring dispositional and situational attributions. Personality and Social Psychology Bulletin, 4, 598-594.
_______
Re your question, two things come to mind:
(1) the entire literature on ambivalence (esp. attitude ambivalence) makes clear that any mid-range rating on a bi-polar scale may be the result of two oppositing evalutions that a respondent may have. Amb is defined as the simultaneous occurence of oppositing evaluative cognitions. Thus, depending on what you are tapping into with your unipolar rating may neglect the opposing cognition that is involved in the bipolar rating as well. Check that literature, you'll find lots of hits in PsycInfo, or whatever you are using. Names that come to mind are Priester, Petty, Zanna.
(2) bipolar and unipolar ratings may indeed pan out quite differently. Gerald Sande in the late 80s (JPSP)demonstrated as much for self-related evaluations/cognitions. It's primarly a matter of complexity.
______
This is an old topic in personality research. It is often subsumed under the question of bipolar and unipolar factors. It is a problem that has roots in the extensive and unquestioned use of Osgood's SD, which is bipolar of course. Your assumption that the effect is due to cognitive factors is an unfortunate reflection of the current cognitive bias in social and personality. A better explanation might lie in the fact that experimenter constructed scales are misleading in themselves. And bipolar scales are even more biased by the experimenter by the arbitrary choice of the poles which might not reflect the participant's perception and certainly force a different construction of meaning. A repetition using subject generated scales might be enlightening. One reference that comes to mind (and I don't remember if it deals with bipolar response formats or not) is...
Watson, D., & Tellegen, A. (1999). Issues in the dimensional structure of affect: Effects of descriptors, measurement error, and response formats: Comment on Russell and Carroll (1999). Psychological Bulletin, 125(5), 601610.
_______
you might want to check:
Sande, G. N., Goethals, G. R. & Radloff, C.E. (1988). Perceiving one's own traits and others': The multifaceted self. Journal of Personality and Social Psychology, 54, 13-20.
Green, R.F. & Goldfried, M.R. (1965). On the bipolarity of semantic space. Psychological Monographs, 79, whole no. 599.
______
I think Dovidio found racism in unipolar scales but none in bipolar. He attributed to it to the unipolar scales appealing to aversive racism.
Here's the reference, but other Dovidio articles describe the study if you can't dig this out: Dovidio, J. F., Gaertner, S. L. (1991). Changes in the expression and assessment of racial prejudice. In H. J. Knopke, J. Harry, & R. J. Norrell (Eds.), Opening perspectives on race relations in contemporary America. (pp. 119-148). Alabama: University of Alabama Press.
______
And my own research turned up the following:
Locke, K. D. (2002). Are descriptions of the self more complex than descriptions of others? Personality and Social Psychology Bulletin, 28(8), 1094-1105.
Dean Peabody's work in the late 60s and early 70s.
Bruner, S. B., Shapiro, D., & Tagiuri, R. (1958). The meaning of traits in isolation and combination. In R. Tagiuri & L. Petrullo (Eds.), Person Perception and Interpersonal Behavior. Stanford, CA: Stanford University Press.
Gannon, K. M., & Ostrom, T. M. (1996). How meaning is given to rating scales: The effects of response language on category activation. Journal of Experimental Social Psychology, 32, 337-360.
Rob Foels
Department of Psychology
Westfield State College