I'm conducting a Masters dissertation on the differences in decision
strategies and knowledge representations for chess players of
differing skill levels and I'd like to run my proposed methods by the
group - comments extremely welcome!

To elicit knowledge representations I have chosen personal construct
elicitation. Each element will be a chess position, characterised by
a board + pieces, a move history and indication of whether it is
wehite or black to play. I intend to present these as triads and ask
participants to identify ways in which two positions are similar and
thereby different from the third in the context of their
consideration of the next move. I am aiming for about 10-15 elements
and approximately the same number of triads (i.e. each element
appears in 3 triads)

I am aiming for approxiamtely 20 players of differing abilities. I
shall run through the above method and use laddering after each
construct to explore construct systems. I shall continue through all
triads or until the participant and I believe that there are no new
constructs to be elicited.

I intend to compare and classify constructs across all participants
using content analysis. I shall then make comparisons between the
nature of construct systems (i.e. number of constructs, overlapping
of constructs etc.) across all participants, probably by considering
them as two groups (experts and novices) based on their chess ratings.

I have a few issues that I'd like to discuss with PCP experts like
yourselves:

1. ELEMENTS/ TRIADS. I believe I require more triads than number of
expected constructs - am I correct? I am considering a structured
sample of elements for each triad to ensure that (a) each element
appears in three distinct triads (b) no pair of elements appear in
more than one triad - is this sound?
2. CONTENT ANALYSIS. I have applied the content analysis method in
another study and understand the subjectivity I introduce through
this procedure, but see no other way of making comparisons between
linguistic constructs. My hope (rather than belief at this stage) is
that if I use laddering, I stand a better chance of finding common
constructs between participants. By laddering down to obtain ever
more concrete constructs, I hope to see convergence here. I then
intend to compare construct systems from these common 'roots'
upwards. I do have one expectations about this differences between
experts and novices, namely that experts will possess more abstract
constructs than novices. However, in terms of the NUMBER of
constructs elicited, I am not sure whehter experts will possess more
(because they have a deeper understanding of their domain) or less
(because they reationalise their construct systems). I also do not
know to what degree expert and novice construct systems will overlap
and also understand that this could only really credibly be done for
a single player in a longtudinal study as he/ she became more expert.
3. REP GRIDS. I may ask participants to complete rep grids but I
cannot see much value in this since my aims are to compare knowledge
representations ACROSS the group and I would therefore require rep
grids against the SAME constructs to do this. Comments?

I look forward to reading any replies and would be happy to clarify
any points.

Thanks,

Paddy Turner.