A few brief comments.
My approach to a similar problem was to perform individual
differences multidimensional scaling using SPSS (you would need
technical advice regarding the suitablity of the shape of your
data) and content analysis. I chose not to use cluster analysis,
a better approach for you might be to perform principal
components analysis on the individual grids using a package which
computes standardised loadings (you could use the factor analysis
option in G-Pack if you know how to make the necessary
computations). This approach allows you to set some criteria for
loadings and you can compare the types of constructs used and the
constituents of highly loading components.
Regards,
Bob Green
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%