Just as a reminder of how numbers can change over time, and from poll to poll:
A Sept. 18, 2008 Field pollreported 55%-38% "no" (i.e., against overturning the California Supreme Court decision that mandated the recognition of same-sex marriage) among likely voters (+/-3.5%).
A Sept. 26, 2008 Survey U.S.A. poll reported 49%-44% "no," apparently also among likely voters.
Is the reason for the shift "a television ad campaign that features footage of San Francisco Mayor Gavin Newsom proclaiming same-sex marriage is here to stay 'whether you like it or not'"? Or is there no real shift, given the margin of error and given the broader possibility of variation based on factors such as who tended to be at home when the pollsters called? In any case, this is just a reminder not to put too much stock in pre-election polls.