a) apologising for causing a flamewar
to which all I can say is gosh! The most I've seen is some useful and
friendly flickers, (like my Zippo when it's short on fuel): you should see
what _real_ flamewars are like on some of the other Newsgroups!!!
and
b) by saying
>The problem which has arisen has to do with two facts: (1) Many Ss
>produce the same constructs/contrasts under different labels; such
>constructs/contrasts should be organized into a single construct/contrast
>and (2) Many Ss organize two constructs/contrasts as direct opposites of
>each other -- as an hypothetical example, one construct may display
>GOOD/BAD while the other may display BAD/GOOD. These two
>constructs/contrasts may be identical, excepting only that the S wrote
>them in reverse scores. This is the situation in which reflection of a
>construct is necessary.
Well, a good package like Mildred Shaw's REPGRID will pick up the
degree of their similarity or difference during analysis, regardless of
orientation; this alone isn't sufficient to classify them as the same
construct, however, and some verbal interaction with the respondent over
his/her intended meaning is de rigeur.
For example, (assuming we're working with a 5-point rating scale), no
computational algorithm is going to pick up that a set of 5 elements
consisting of foods which I might order off a menu, construed in terms of
(i) Good for my figure - Disaster for my figure
and
(ii) Tastes sour - Tastes sweet
given that the ratings are
1, 3, 5, 4, 4
and
1, 3, 5, 4, 4
respectively
(or with Reid's second example in mind,
(ii) Tastes sweet - Tastes sour
with ratings of
5, 3, 1, 2, 2)
are actually _different_ constructs conveying _different_ meanings about
the elements. Identity of ratings, in other words, isn't any guarantee of
identity of meaning, and some verbal interaction ("learn the words!" as
Terry Pratchett says of mime artists) is usually required.
Conversely, of course, during the course of elicitation, both the
researcher and myself are perfectly capable of recognising identity, and of
agreeing to drop the repetition, when we agree that the verbal labels
convey the same meaning, and that all the elements are receiving identical
ratings on the repetition and on the original. Thus, we might agree that
(iii) Will make me fat - Won't make me fat
with ratings
5, 3, 1, 2, 2
on the same elements is in fact the same meaning as construct (i), albeit
the ratings are reversed, and we'd most probably agree to drop it as
genuinely redundant.
Hope that's useful.
Kindest regards,
Devi Jankowicz
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%