Radical. But also too good to be true, now that the study - widely circulated in the US - has been retracted by one of its authors after his co-author allegedly admitted to fabricating data.

In a memo to the respected academic journal Science, which published the paper, Columbia University professor Donald Green said he was "deeply embarrassed" at holes in the research found by peers at other US colleges.

The outcome has disappointed marriage equality campaigners in Australia and abroad.

Australian Marriage Equality director Rodney Croome said supporters of same-sex marriage should not be discouraged from telling their stories and trying to persuade opponents.

The bulk of the work behind the study - When contact changes minds - was undertaken by a PhD student at the University of California Los Angeles, Michael LaCour. Its findings were striking: that 20-minute conversations with voters opposed to gay marriage could turn them into enthusiastic supporters of marriage equality.

The research also contended that, if the person conducting the interview was gay or lesbian, the voter was likely to remain convinced about same-sex marriage when interviewed again nine months later - rather than reverting to their original views.

"Contact with minorities coupled with discussion of issues pertinent to them is capable of producing a cascade of opinion change," Mr LaCour and Professor Green concluded.

The study was reported by news agencies including The New York Times, The Wall Street Journal and radio program This American Life. Its implications were enormous, bucking widely accepted wisdom that most people do not really change their minds, or that, if they do, the process is slow or beyond the sphere of political influence.

But following an attempt to replicate the results, academics at Stanford, Yale and the University of California Berkeley found:

• The survey firm alleged to have performed one of the studies cited by the paper claimed it was not familiar with the project.
• The employee Mr LaCour claimed to have dealt with did not exist.
• The firm denied having the capability to perform some of the activities described in the paper.
• Mr LaCour was unable to provide the original data sets and said he had lost it.

Furthermore, Professor Green told Yale assistant professor Peter Aronow that Mr LaCour "has been confronted and has confessed to falsely describing at least some of the details of the data collection".

In a subsequent interview with the host of This American Life, Ira Glass, Professor Green said he was astounded at the level of falsification that appeared to have taken place.

"There was an incredible mountain of fabrications with the most baroque and ornate ornamentation," he said.

"There were stories, there were anecdotes, my Dropbox is filled with graphs and charts, you'd think no one would do this except to explore a very real data set.

"[With] all that effort that went in to confecting the data, you could have gotten the data."

Mr Croome said he was disappointed but undeterred by the retraction of the study. Having spent his entire life trying to convince people to support same-sex marriage, he was surprised by the paper's initial findings, which he described as "overblown".

"Real change takes time. It's rare that a single conversation will make the difference," he said.

"It's an accumulation of conversations and then there will come a point where ... the penny will drop."

But the principles of sharing personal stories and engaging emotionally were still vitally important to the marriage equality campaign, Mr Croome said.

AME encourages supporters to have conversations with their family, friends, neighbours and MPs about the issue.

"Personal interaction on marriage equality is still the path forward. The path may not be quite as short and well-paved as that study suggested," Mr Croome said.

Science journal editor-in-chief Marcia McNutt said in a statement that the journal was working towards correcting the record while following due process.

Mr LaCour has generally refrained from comment, telling The New York Times he had received legal advice not to speak but adding: "I'll be fine."

On his Twitter account the PhD student said he was "gathering evidence" and would provide a comprehensive response.

(1/2) I read "Irregularities in LaCour (2014)," posted at http://t.co/SJ3jE89XeD on May 19, 2015 by Broockman et al. I’m gathering evidence— Michael J. LaCour (@UCLaCour) May 20, 2015

(2/2) and relevant information so I can provide a single comprehensive response. I will do so at my earliest opportunity.— Michael J. LaCour (@UCLaCour) May 20, 2015

About 9500 voters from Los Angeles County were involved in the study. The research was conducted in 2013 during the month leading up to a US Supreme Court decision that effectively overturned Proposition 8, a measure to ban same-sex marriage in California.

The study had also reported that the 20-minute doorstep conversations had a measurable "spillover effect", in which some residents who did not speak to the gay canvasser also formed a positive opinion of gay marriage.

David Fleischer, project director of the Los Angeles LGBT Centre Leadership LAB, said the group was "shocked and disheartened" by the turn of events.

"How it feels is like a punch in the gut," he said.

"It really hurts when you trust somebody to be doing an honest assessment of your work and then it turns out that they did not. We have a wealth of qualitative data that has made us feel like we're having an impact, but we are well aware that that's not a substitute for independent information."

- with McClatchy Newspapers and The New York Times