Constructs and attitudes

CSU ( csu@brain.wph.uq.oz.au )
31 Jul 1996 09:10:01 -0500

Rainer wrote,

>Ute Klein, a student of mine, is just analysing the data of a
>'zero aquaintance' study. She had about 70 students to complete
>a standard repgrid and a standard personality questionnaire
>(NEO-FFI). The repgrids were typed and given to a group of
>raters (advanced psychology students) who completed the peer
>rating version of the same questionnaire. It turned out, that
>these raters who had nothing but the repgrid where able to
>predict the self-report data. The accuracy of these predictions
>is about as high as that for raters who (in other zero
>aquaintance studies) had observed a behavioral episode of that
>person. In my view, this study demonstrates that repgrids
>contain valid information about the persons' behaviors and that
>there is a relation between behavior and constructs (see
>Brunswik's (1956) lens model). Comments on this kind of study
>are welcome.

I would be interested to clarify three points. Firstly, were the
raters given a number of rep grids and asked to fill out
questionnaires about the persons who completed the rep grids.
If I have this right, how did you determine the accuracy of the
predictions (e.g., correlation, some measure of agreement such
as kappa or trying to predict the ratings of students as a group
by raters as a group via regression)? Finally, what is a zero-
acquaintance study, excuse my ignorance on this last one?

Your comment about Brunswik's lens model is of interest,
particularly if constructs are seen as cues in this model. My
understanding is that to use this model idiographically, that the
person whose judgment is being modelled has to make a number of
judgments (e.g., 50 or so). Someone who has used this approach
and who was sympathetic to grids/PCP, suggested to me that he saw
the value of grids as being the identification of cues/constructs
which may be meaningful/included in a Lens analysis.

A criticism and in my view a reason for the failure of various
predictive models, is the failure to adequately understand
factors which are important to judges/raters. Rather a
researcher will select constructs/cues or clump together a group
of individuals and develop an aggregate model.

Looking forward to any comment you care to make,

Regards,

Bob Green

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%