Just out, a report by American Association for Public Opinion Research (full disclosure: I'm a member of AAPOR, but I had no role in the report) on why four 2008 election primary polls had some odd results, especially for Hillary Clinton.
I've only started reading the report, but the authors especially attempt to explain the weird stuff from the New Hampshire polls (at least the part I'm reading now). Some problems have to do with sampling, some with shifting dates by the states themselves, some with question wording or style of questions, and some problems with the magic pollsters use to adjust numbers to fit the general population. The lay person will struggle with some of this. Hell, I'm not "lay" and I struggle with some of this, but it makes for interesting reading if you're into polls and politics and how the two interact.
This may find its way into my summer Public Opinion class.