re: Introduction and request for advice
Thu, 8 Feb 1996 22:34:04 +0000

Carolin Kreber writes:

>In theory this project sounds quite feasible, doesn't it? However, I
>have been getting some doubts recently. It
>seems to me that the process of completing the grid is very time-consuming

Indeed it is. Kelly did say that "if you want to know what someone thinks,
why not ask, they might just tell you" or words to similar effect. But what
repeats itself in his theory, his work, and indeed in the experience of
everyone who tries to know what others think, is that if you want a
non-trivial account of what someone thinks, you have to be prepared to work
a little at it.

>I am no longer sure whether faculty will actually take the time and
>complete the grid for me (and I will need at least thirty faculty in each
>group). Does anybody have some experience with administering
>the grid "by mail" and could give me some advice in terms of how the
>process could be made easier, quicker, etc??

The repertory grid works well in a one-to-one setting, where the
respondent's meaning can be negotiated with the investigator, rather than
his/her "responses obtained". It works less well in a group setting, where
the investigator works with a group of respondents issuing grid
instructions en masse. The point being that getting people to cooperate
with what is a fairly unconventional research procedure to most people is a
matter of a close, careful, and personally-attentive skill. You can't do a
grid by mail!

(Crumbs, that's true even of the most automaton-like assessment technique,
the psychometric test!)

I shudder to think what trivia you will obtain by trying to do a repertory
grid by post!

You say you have to have at least 30 respondents in each of (I understand)
2 groups. Okay, but who says so? Why does that person say so? Have you the
freedom to ponder this question (e.g. no higher degree supervisor breathing
down your neck)?

Next. Eventual analysis.

You can do a perfectly respectable content analysis of the constructs of
just 20 people, if each person gives you 8 - 10 constructs. That's roughly
20 hours of face-to-face contact. I've just completed a consultancy which
involved obtaining repertory and resistance to change grids from a sample
of 36 people, each interview taking around 1.5 hours and resulting in 36
repertory grids each of around 6 - 9 constructs, _and_ 35 construct
hierarchies of 4 - 7 pyramided constructs each, the latter in a priority
ordering. Yep, over 50 hours.

The 300-odd constructs were just the right sort of number for a construct
analysis; indeed, I've found that such analyses become fairly redundant
with reasonably homogenous respondent samples once one has more than 350
constructs with which to work.

But you have to put in the time!

(Or, of course, adopt a research design in which you only do rep grids on 4
or 5 people, treating them as exemplars, case representatives, as it were,
of different positions along some key variable which governs your research
design: (Yin, 1995, is a good reference, call me if you want to know more).

But then I guess you'd need to triangulate the descriptions you obtained
with the grid technique against some other technique; and why no,t if it
helps answer your research questions?)

Kindest regards,

Devi Jankowicz.