Here's the article (couldn't find it earlier because I was mis-spelling Hannover). Unfortunately it's not free:
http://www.newscientist.com/article.ns?id=mg13918843.300
The abstract claims "only one in 40" could distinguish CD from vinyl, but the data, as I recall, supported an even stronger conclusion than that, as 1 in 40 was what one would expect from chance given the experimental protocol described.
I'm surprised this isn't more widely known - expected to find a slurry of audiophile denial about it on the web, but it seems to have been just ignored.
John