A third approach, one that cuts to the chase and relies on the respondent's own self-assessment, is to ask people whether or not, or to what degree, religion is important in their everyday lives. Gallup, Virginia Commonwealth University (VCU), and some of our own Harvard School of Public Health surveys ask this question, in differing forms. About 10-15 percent of Americans say that religion is "not (or not very or not too) important." The advantage of this approach is that a single-question measure saves space on the survey, and the question asks the respondent's own self-assessment on an issue where he or she might be expected to be the best judge. The disadvantage is that it may not accurately reflect religious behavior or heritage that may be related to the respondent's attitudes, even if the person does not think religion plays an important role in his or her life.

A more refined method uses two or more questions to define a group as "secular" or to create a scale of religious commitment, or "religiosity." The Pew Forum on Religion and Public Life and the HARI survey both define "seculars" (10-11 percent) using religious preference and attendance at religious services. Again, this is a large enough group for analysis, but not large enough to break down into subgroups of the nonreligious.

More complex scales, such as Pew's "religious commitment" and Washington Post/Kaiser Family Foundation/Harvard University's "religiosity" measures, have the advantage of allowing researchers to divide the scale in different places to vary the number of less-religious people. Normally such scales are divided in thirds. This makes it possible to look at less-religious men or women, Protestants or Catholics, college grads or non-college grads. The downside is that the scales are harder to explain to readers or listeners, and that the multiple questions used to define them take up space on a poll.

We have looked at the advantages and disadvantages of each approach to defining the non- and less-religious. But here's the interesting part: no matter which of these definitions one uses in order to look at social attitudes, they generally yield surprisingly similar results—not in every case, of course, and one would have to be sure that the idiosyncrasies of a particular method were unlikely to affect responses to a particular type of attitudinal question. But a look at Figure 2 demonstrates the similarities.

By any measure, differences in attitude between non- or less-religious versus more highly religious Americans have been most striking on social (including "life") issues and moral values. Using two or more of the definitions we have discussed, the less religious differed significantly from other Americans in their views about:

• same-sex sexual relations and same-sex marriage, where the less religious were 26-38 percentage points more likely than other Americans not to think same-sex sexual relations are wrong and more likely to favor allowing same-sex marriage.
• premarital sex, where the less religious were 26-29 points less likely than others to think it is wrong.
• the role of women, where the less religious were 20-21 points less likely to believe it is better if the man works and the woman tends to the home.
• abortion, where the less religious were 23-44 points more likely to think a woman should be able to get an abortion for any reason and that abortion should be legal under all or most circumstances, and less likely to feel that abortion is unacceptable.
• physician-assisted suicide, which the less religious were 23-38 points more likely than others to favor.
• stem cell research, which the less religious were 18-23 points more likely than others to favor.

The one obvious exception to the pattern was capital punishment, where competing values joined in such a way that the views of the less religious and more religious did not differ significantly. But, again, this was true regardless of which way one defined the less religious.