I've posted a new article (http://seanolive.blogspot.com/2010/03/method-for-training-listeners-and.html) summarizing some previous research I did on a method for training listeners and selecting program for listening tests.
The training task focussed on having listeners match the sound of different spectral distortions added to music to the frequency response of the filter that generated the distortion, using 20 different program selections. Listener performance in the task based on percentage of correct responses was significantly influenced by the program and its spectrum.
Other factors that influenced listener performance were the EQ-filter, the combination of EQ's presented to the listener (context effect) and an interaction effect between EQ and program.
Audio Musings (http://www.seanolive.blogspot.com)