“I could summarize it as follows: I wrote a program that gave a wrong answer, and then I fiddled and fudged until I finally got the output I wanted, and then I stopped. This is not a protocol to be recommended.”

http://www.americanscientist.org/template/AssetDetail/assetid/42368?&print=yes

He presents a model of rumor passing and gives some empirical results. The article isn’t an exact match for your discussion, but if you are thinking about rumor passing I think you’ll enjoy Brian’s article.