TragicMonkey
Poisoned Waffles
Instead of speculating over the validity of indirect measurements with an assumed causal link and correlation, not to mention the fallibility and unreliability of the underlying data, why not simply interview a cross-section of rapists and seek out their motives and, more importantly, prompts. That might sound simple and fundamentally flawed (in some way), but if you cannot do that, then how can you possibly hope that indirect measures with all their fundamental weaknesses, as described above, can be helpful, as the two methods of measurement are inextricably linked, with the indirect measure, by definition, being far less reliable.
Incidentally, I should not be surprised at all to learn that certain 'styles' of pornography have indeed led directly to, or at least contributed directly to, some instances of rape, but I suspect the incidence is low in comparison to all instances of rape, even assuming a 'more violent' definition of rape.
I think you'd run into the same sampling problem as when people try to link violent video games to crime. They go into the shooter's house and find "Grand Theft Auto" and scream "Aha!" But they do not mention the millions of people who have that game who have not murdered anybody. If you start with the sample and then study all their stuff to see what qualified them to be in your sample to begin with, you're working backwards.
The proper way to do a study would be to give the stimulus to a randomly selected sample, then wait to see how many exhibit the behavior you suspect they will..and compare to all of those that don't to see if it's significant.