Legalization could reduce small-time, pot-related arrests and give supporters a chance to show whether decriminalization is a viable strategy in the war on drugs.
The effort raised more than $6 million in contributions, with more than $2 million of that coming from Progressive Insurance Co. founder Peter Lewis, who used marijuana to treat pain from a leg amputation.
Some people in the marijuana reform community also objected to the DUI standard, which they called arbitrarily strict.
The campaign had little organized opposition but raised objections by law enforcement officials and some substance abuse experts who said increasing access to pot was a bad idea.
"Legalizing is going to increase marijuana use among kids and really create a mess with the federal government," said Derek Franklin, president of the Washington Association for Substance Abuse and Violence Prevention. "It's a bit of a tragedy for the state."
For many voters, it came down to the notion that decades of marijuana prohibition have done more harm than good.
George Cannon, 43, of Seattle said it was an issue of personal freedom: "I'm not into getting into other people's business."
Initiative 502 found strong support among liberals and moderates, Democrats and those with more than a high school degree. Independents and women were split on the issue, as were suburbanites.
I-502 fared well in King County and the Puget Sound area, but not in Eastern Washington, Southwest Washington or on the Olympic Peninsula.
Opposition came from voters 65 and older, conservatives, Republicans and those with a high school degree or less. Weekly churchgoers rejected the measure, while those who said they never attend religious services or considered themselves occasional churchgoers favored legalizing pot.
The survey of Washington state voters was conducted for The Associated Press and television networks by Edison Research. It includes preliminary results from a survey of 1,493 voters who cast ballots early or submitted absentee ballots, and were interviewed by landline or cellular telephone from Oct. 29 through Nov. 4.
Results for the full sample were subject to sampling error of plus or minus 3 percentage points. The figure was higher for subgroups.