Another causal factor beyond respondents' awareness was the order in which the response alternatives were read to them. Data from the NEP exit poll show that when "moral values" was presented to respondents near the beginning of the list, over a fourth (26 percent) said it was the one issue that mattered most in their vote decision. But when it was presented near the end of the list it was less likely to be selected (22 percent)-a good example of what survey researchers call a " primacy effect." Notice, too, that "health care" was nearly twice as likely to be selected (11 percent) when it appeared at the beginning, rather than at the end, of the list (6 percent).

An even more dramatic example of the causal influence of the order of the response alternatives comes from the national exit poll conducted for the 1992 Bush-Clinton-Perot election, in which respondents were asked, "Which 1 Or 2 Issues Mattered Most In Deciding How You Voted?" In this case the list of issues included something called "family values," a vaguely worded forerunner of the "moral values" option that would surface twelve years later in the 2004 exit polls, but which was very much a part of the culture war theme that emerged in Patrick Buchanan's address to the 1992 Republican National Convention. Not surprisingly, the order in which the issues were presented to respondents in the exit poll questionnaire made a sizable difference in how important they thought "family values" was in deciding how to vote.

Twice as many respondents (20 percent) chose "family values" as the critical issue in deciding their vote when it was presented as the first item in the list than when it was listed last (10 percent). Their perception of the influence of "family values" on their vote decision was essentially an illusion, one that was generated, in significant part, by the order in which the choices were presented to them-a causal influence of which they could not possibly have been aware.

Surely, no respondent in the 1992 election exit poll would be likely to have volunteered that one of the causal reasons they chose "family values" was because of the order in which the response alternatives was presented to him or her. Surely, no respondent in the 2004 election exit poll would be likely to have known that among the causal influences on their selection of "moral values" was the form in which the question was asked and the order in which the response alternatives were listed.

So much, the cognitive-psychological determinists might say, for what respondents believe are the causal influences on their behavior. So much, they might say, too, for the rational voter, and other rationalist illusions of people making conscious choices and presuming to tell us why they do what they do. The apparent influence of "moral values" on the 2004 presidential election, they would argue, represents no more than another rationalist illusion. A grim conclusion perhaps, but it's not out of line with contemporary thinking in cognitive neuroscience.