In any human endeavor there's an execution component that scientists tend to overlook. You can't just form study groups and expect everyone to excel. The details add up and you end up with a successful program.

The original article states that Treisman's grant ran out, but all of Berkeley was eager to get a part of the program, it changed hands and then went downhill from there, so I guess the guys who inherited the program overlooked the finer details that made it successful in the first place.

Probably mostly the Hawthorne effect; making "any" change tends to lead to temporary improvements.

http://en.wikipedia.org/wiki/Hawthorne_effect