Cringely has an article about an experiment with AdWords.
pbs.org/cringely/pulpit/pulpit ...

Someone had an existing web site with an AdWords campaign that produced 15000 clicks per day at $.10/click.

He cloned his own site, bought the same words with the same ads, but varied the price per click on the clone while he kept the original site the same.

When he paid $1.00/click on the clone it did better than the original that maintained its 15000 clicks/day.

When he changed the clone to $.40/click he dropped to 1200 clicks/day (less than one tenth). I think that the clone site was the one that dropped, but I am not certain. Check the article yourself. Perhaps we should ask Cringely to clarify this.

Some Slashdot comments speculate on the cause.
tinyurl.com/akkp2

Upon reading more Slashdot comments I found out about some more Cringely wrote about this.

pbs.org/cringely/pulpit/pulpit ...
pbs.org/cringely/pulpit/pulpit ...