polls gone wild

News I'd like to see Ryan's take on:

I have a hard time accepting this poll as having anything like a real handle on the population it claims to poll. Not that I think the results are wrong; it seems to me the results could be arrived at by anyone who had read much of anything about the situation in Afganistan. But if the stories on the lack of basic infrastructure are anything close to accurate, how can you do a "poll" of Afgans at all?

This also ties into a story I heard on New Orleans radio today: a telephone poll to find out how many people had suffered a stroke. Doesn't it seem the results would be slightly skewed? I think there is probably a statistically significant number of stroke victims who are either reluctant to answer to phone or phsycially incapable.