Peter is quite right. I was mistaken in saying graph shape and category shape are mathematically equivalent. Category shape for diagrams has graph shape plus the commutativities, which graph shape on its own does not capture. His justification for considering category shape is much more profound than the ones I mentioned. Nonetheless, when verifying the conditions for cones/cocones, or for limits/colimits, the commutativities make no difference and it suffices to check on a generating graph. Thus it is worth bearing in mind that in such contexts one might correctly and more simply just use graph shape, as Gillian and I did in "Presheaves as configured specifications". (For those interested it's on the web via my home page at http://mcs.open.ac.uk/sjv22 .) Steve Vickers. 26-Sep-2001 13:27:29 -0300,2418;000000000000-00000026