(Fwd) Re: Can Anybody Help?

Chris Evans (chris@psyctc.org)
Sat, 27 Mar 1999 16:44:16 -0000

I suspect John meant to send this to the list, not to Malcolm and I
(if you send want to send technical problems or other queries to
us, send them to:
I'm away much of the time until Monday week so Malcolm's your
main man!

Things to the list itself need to go to:

Best wishes all,


------- Forwarded Message Follows -------
Date sent: Fri, 26 Mar 1999 14:46:04 +1200
To: pcp-request@mailbase.ac.uk
From: John Mayes <christopher.mayes@actrix.gen.nz>
Subject: Re: Can Anybody Help?

Melanie writes in part:

>The reason I chose rep grids was to try and get underneath the concept of
>trust in this domain as no-one else has looked at it before to my
>knowledge. However being a novice and having already collected the data (I
>hope correctly) I am having great dificulty in analysing it.
>I have read Valerie Stewarts book on Business Applications of Repertory
>Grids. However what I don't understand is how to reduce the constructs of
>20 different people (albeit same job same status) over 24 elements in order
>to get the group constructs for the three different factors.

I referred the request to Dr Valerie Stewart who has provided a response as

I do not think that there is an easy answer to your question, if indeed
there is one at all. May I ask the following questions?

1. Research Design:

What is/are the hypotheses you set out to test? You phrase your first
question rather like a hypothesis (If people in a highly volatile domain
(e.g. HSC) construe trust differently (from other cohorts/ domains - e.g.
literature on diff orgs and dyadic/professionalrelationship) but a passing
reference to the literature on different organisations etc. is not enough
to allow you to formulate a hypothesis. Apart from anything else, it
assumes that the information on other cohorts/domains is accessible in a
format which allows Grid methodology to use it as a comparator, and you say
later that nobody else has, to your knowledge, tested it. Your second
question ( How important trust is intra-team inter-team and in interaction
with technology) isn't phrased as a testable hypothesis. Your third may or
may not be - it's hard to tell. But it you cannot formulate your research
as a testable hypothesis then no methodology, Grid or otherwise, will help
when you collect the data.

2. Grid methodology.

You say that you use as elements people (either intra-team, inter-team) or
technology (systems they interact with). This is a non-homogeneous element
set. It therefore won't work, and this is one of the first points made in
all the Grid literature. What is surprising is that you collected so much
data without trialling one or two analysis methods first.

You have 24 elements. This is a large number to begin with, although a Grid
can grow during the process of conversation and feedback so that it
includes that number of elements. But - where did the elements come from?
Did you suggest them? Did you derive them in cooperation with the
interviewees? What superordinate purpose for the interviews did you have
(it should have arrived naturally from your research hypothesis) and what
qualifiers did you use (they should also have suggested themselves a
result of your hypothesis)?

You say you hope you have collected the data correctly. The absolute
minimum test for 'correctness' is the response you got from your
interviewees when you fed it back to them. Grid is a conversation in which
feedback and joint exploration of the topic between the interviewer and the
interviewee is an absolute essential. (See, for example, almost anything
written by Mildred Shaw or Maureen Pope over the last 25 years. They have
always stressed this).

Why do you want to 'reduce' people's constructs? From what to what, and
what will you learn at the end?

By the way, a sample of 20 people seems rather small for testing so many

3. Analysis.

If you haven't formulated your initial hypothesis then no method of
analysis is, I'm afraid, going to help you because you won't know what you
want the analysis to do for you. You have a collection of words and
numbers. If you give that to a statistician and ask them to analyse it the
first question they will ask you is 'what do you want it to tell you?'
Your best plan will be to see whether, in all the data you have collected,
and bearing in mind the 'good practice' guidelines for the use of Grid
which are generally available, you can discover any serendipitous natural
experiments which your data are robust enough to support.

I am sorry to have to give you such a negative answer but I think you will
waste your time looking for an analysis package which will answer the
questions which should have been more clearly formulated at the beginning.
Laurie Thomas, one of the finest teachers of Grid to walk this earth,
always used to say that Grid is a 'powerful empty procedure' and that the
greatest danger was to start a project without asking some very clear
questions about what the project was, and was not, intended to reveal.
Otherwise, he said, you would finish up with great bundles of data which
you had no idea how to analyse. I think you need to seek much clearer
counsel, step-by-step, to go back over the detail of your data and
protocols and see what can be rescued, rather than look for an analysis
package which will only have you asking the same questions in a few months'

Valerie Stewart.

John Mayes

PSYCTC: Psychotherapy, Psychology, Psychiatry, Counselling
and Therapeutic Communities; practice, research,
teaching and consultancy.
Chris Evans & Jo-anne Carlyle
http://psyctc.org/ Email: chris@psyctc.org