It was on HN yesterday.
"Optimizing Wikipedia's Fundraising Banner Ads" (http://blog.wikimedia.org/2009/12/11/annual-fundraiser-checking-banner-results) asked readers to post their thoughts on the difference of average gifts donated by wikipedia users, when shown different banners.
I wrote a blog post about the "anchoring heuristic", which explains the phenomenon.
From a very quick scim of the original blog Wikimedia blog post it looks like the total for #18 was higher (with the same number of impressions). What this suggests to me is that the lower amount shown on the banner made people feel there donation would count even if it wasn't very much. If more of these people donated one would expect the average to go down and the total to go up, and that appears to be what happened.
The total amount (not the average amount) is what counts to wikimedia. From the original blog post (less suggested -> more donated):<p><pre><code> #17 suggests $30, totals $22,000
#18 suggests $1.95, totals $11,000 (same views)</code></pre>