The survey in question was conducted by Annenberg, which has an OK reputation. It's necessary to know who did the survey, who paid for it, and to know a little about the target group to assess a poll's validity. For example, I doubt that many of the troups in Iraq get the news from Reuters. They probably use the sources the brass make most convenient, thus slanting everything from the very begining. Polls that allow 'self selection' are a waste of ink because the undecided and the 'don't cares' are excluded. On line polls are too easy to sabotage. Etc.