The study did not list confusion as to the definition of running a read light (i.e. assuming permissive yellow in a non-permissive state) as a contributing factor. I suppose for a Texas study that's understandable while it would be significant for a non-permissive case study. (pp 2.2-2.3)
While that tamu study did say too long a yellow cycle will cause a small incidence of increased red-light running (RLR) p. 2.9, it also showed a substantial improvement in compliance with a one second increase (53% fewer RLR) p.5-20. The fifth summary finding on page 6-8 says a nominal increase of .5s to 1.5s of yellow still yields a 50% decrease in RLR's even considering the learned misbehavior. The College Sta. observations on page 5-29 bear that out.
My state is the permissive yellow, too.
One of the studies I included is frequently quoted around here by people who say that lengthening the yellow increases safety. However, they only quote what they want to hear from it. The guy who wrote it is pressing for the maximum ITE yellow time to go from 5.0 seconds to 5.5 seconds. From my understanding, the safety curve is essentially flat between 5.0 and 5.5 seconds. The establishment wants to leave it at 5.0 seconds. So he is arguing with the establishment over 1/2 of one second.
To read some of the people on this website, 7 to 10 seconds would be better. It would not. No peer reviewed article I know of supports that. Even the guy mentioned above agrees that anything more than 5.5 seconds is dangerous.
We include a one-second, all-red indication on all collector and primary roads. That is pretty well supported in the literature and in our experience as being better than longer yellows. That seems to help things around here, but the suggestions I hear about going to 3 seconds scare me. You only have 60 seconds in a minute to move traffic. Anything you do to reduce the number of vehicles that can pass through an intersection will make it more dangerous since people are impatient enough as it is.