I always take public opinions polls with a grain of salt. Being involved in politics, I know how easy it is to get a poll to pretty much say whatever you want depending on how you ask the questions. Here is a great example from FRC's daily email of how opinion polls can vary widely from actual voting on an issue. This is also another example of how far the Senate is out of step on this issue.
Be leery of public opinion polls on the issue of marriage. Sen. Rick Santorum (R-PA) has pointed out how very different the polls are from the actual votes. Here are some typical cases showing what the polls said about protecting marriage in various states prior to the votes and how the citizens of those states actually voted. Arkansas: 64.8% (75%); Georgia: 69% (76%); Kansas: 56% (79%); Kentucky: 71.6% (75%); Louisiana: 62% (78%); Michigan: 52% (59%); Montana: 61% (67%); Nevada: 43% (67%); North Dakota: 53% (73%); Ohio: 48% (62%); Oklahoma: 59% (76%); and Oregon: 50% (57%). Why such a big difference between pre-vote polls and the actual votes? It's probably a combination of two major factors. Having been in politics for a while I've seen how polling questions can be constructed to obtain the desired outcome. The slanted poll questions no doubt account for some of the difference. Secondly, and probably more prominent, is the intimidation factor. Unscrupulous politicians like Sen. Ted Kennedy (D-MA) rant that protecting marriage is "bigotry, pure and simple." It's no such thing, but Kennedy and company want to intimidate people. Some people are fearful of being labeled a homophobe or bigot so they keep their opinions to themselves until they are in the privacy of the voting booth and then on average over 70% defend marriage as the union of one man and one woman.