Science journal retracts gay-marriage study after evidence of fraud

NEW YORK (Reuters) - Amid evidence of fraud in a high-profile study on how canvassers can convince people to back same-sex marriage, the journal Science, which published the study, retracted it on Thursday.

The senior author agreed to the retraction, Science editor-in-chief Marcia McNutt said in a statement on the journal's website. The graduate student accused of making up data, lying about funding and other violations has not.

The study, which received widespread coverage when it was published, tested a longstanding theory in social science that contact with targets of prejudice can reduce intolerance.

Specifically, the study examined whether door-to-door canvassers who identify as gay can convince people to support same-sex marriage and do so more effectively than heterosexual canvassers.

The authors, UCLA graduate student Michael LaCour and Columbia University political scientist Donald Green, concluded that canvassers who tell voters that they are gay can indeed change minds on same-sex marriage. That result, the authors said, was based on a survey of some 9,500 people in California.

When other scientists were unable to replicate key aspects of the study, they began digging into it, including contacting the polling company that reportedly did the 9,500-person survey.

The firm, Qualtrics, claimed it "had no familiarity with the project" and "denied having the capabilities to perform many aspects" of the survey LaCour and Green described, graduate students David Brockman and Joshua Kalla of the University of California, Berkeley, wrote in a timeline they posted online.

In her retraction notice, McNutt said that the Green and LaCour paper had lied about paying people to participate in the supposed survey, lied about being funded by prominent groups including the Ford Foundation and failed to produce the original data it was based on.

Last week Green asked that the paper be retracted, telling Science in a letter that he was "deeply embarrassed by this turn of events and apologize to the editors, reviewers, and readers" of the journal.

LaCour has not agreed to have the paper retracted, McNutt said. He has said he stands by the reported findings and told the Retraction Watch blog last week that he "will supply a definitive response" to the allegations and is "gather(ing) evidence and relevant information."

(Reporting by Sharon Begley; Editing by Cynthia Osterman)