I use spatial and visual centres, I think, as I tend to view problems in an almost mechanistic fashion, as components in a multidimensional space, which interact, and have causal results. Of course, if it's a non-abstract problem, then I visualise it directly - or abstract it to then visualise a general class of problems it sits under. Most things fit into a finite number of patterns.

It's really hard to explain, but it makes for easy and rapid evaluation of possibilities, as you can just "see" what fits, and what doesn't.