Eugene Volokh comments on some weird results he's finding in perusing some Los Angeles Times poll data.

Contemporary social scientists would argue that a lot of what Eugene is seeing is due to a problem that political scientist/sociologist Philip Converse first identified around three decades ago, something he called “non-attitudes”. His basic point is that responses to survey questions often don't reflect respondents' true attitudes about things; if they don't have a real attitude, often they just make something up on the fly to stand in for it. Others have argued that “public opinion” is merely constructed; there is no such thing as public opinion until you start asking questions.

Political scientist John Zaller (who I've mentioned before in this category, and is probably the leading authority in public opinion research today) doesn't necessarily agree that there are non-attitudes; rather, people in responding to survey questions sample from their relevant “considerations” (or underlying attitudes) based in large part on question wording.

Getting to Eugene's quandry: the concept of "weapons of mass destruction" may activate particular considerations in women that it doesn't in men — women may have a more visceral reaction to the possibility of Saddam nuking innocent civilians than men do, for example. The second question uses "George W. Bush" as part of its wording; since women are less supportive of Bush than men, the gender disparity may be due to the “Bush considerations” being more heavily weighed here. One need not believe that “many voters' views are ill-formed” (which I think would be Converse's argument) to accept these results as valid.