The test was that of speech (obvious interest of AT&T in that era as the telephone company). The setup of the experiment is on the left. Speakers were placed at the 9 locations and then placed behind curtain. Listeners then voted as to the location of the voices relative to the plane of the curtain:
View attachment 9436
Start at the bottom with the first box I have highlighted in red. That is what the direct sound was voted as. We see distortion even there (indicated by O vs X). Then look up at the middle diagram for 2-channel which I have highlighted. Configurations 4, 5 and 6 lack depth. The top configuration is for 3 discrete channels and that depth is restored. But error is introduced in the left and right configuration relative to stereo.
Other configurations explore mixing of the discrete channels. This is key as obviously 3-channel mixes can sound bad if all the vocals are dumped in the center channel alone for example. You would get clean sound there but would have shrunk the soundstage. Hope is that some is also put in left and right channels and that the room has side reflections to help with that.
The paper goes into much depth trying to explore what is a pretty complex situation.