Have you been able to find the Thomas et al paper that supports this claim? I have been unable to find it, and others seem to have this problem as well:

http://www.gdb.me/computing/citations-greg-wilson-cusec.html

the slides from the talk are here: http://www.slideshare.net/gvwilson/bits-of-evidence-2338367 (see slide 20)

I realized today I got it from somewhere else.. (Facts and fallacies of Software engineering) Looking up the whole reference in that book gave me the title, and googling that gave me the the paper

http://drum.lib.umd.edu/bitstream/1903/703/2/CS-TR-3424.pdf