Unfortunately, there are some disreputable experts who have overstepped their bounds and testified (over simply) that "people haven't been injured in those crash tests therefore no one can be" which is just as wrong as saying "low speed crashes always injure people."
We can further debate the reliability of methods Szabo, Welcher and others have taken to create the "unaware" situation and I can tell you from my own first hand experience that they are not only effective but I have had more than one non-researcher human subject come up to me later and say "wow, that was really a surprise!"
On the other hand, if we adopt the position that these people are "cognitively braced" and THAT's why they don't get hurt, then we must also adopt the notion that no braced or aware-the-crash-is-about-to-happen person can be hurt in these low speed collisions. next patient walks in and says, "I saw him coming, I knew he was about to hit me and I braced for the impact," can't be injured, fair enough?
Based on what I've seen in hundreds of tests, repeated tests with men and women in a variety of ages, I would stand by the conclusion that "a reasonably healthy human occupant can withstand a rear impact with a change in velocity of 8 km/h [5 mph] without sustaining significant injury, assuming the presence of a head restraint and a reasonably "normal" initial seating position." It just doesn't happen in MODERN cars with (the conditions outlined in the statement)but I will allow for the possibility when other conditions exist which would provide a potential for injury (lack of headrest, substantially out of position and others).
The epidemiological studies are far more flawed than the human subject crash tests when applied to real crashes. Compare the two:
(a) The HUMAN SUBJECT CRASH TEST: human subject crash tests are done with known delta-V's and instrumented male and female human subjects of varying ages and physical conditions. The delta-V for the subject's car is recorded objectively by instrumentation as is occupant head acceleration (and the acceleration of other body parts where the researchers are capable of such data collection).
(b) The EPIDEMIOLOGICAL STUDIES: studies of a database of medical cases from medical or claims reports of injury descriptions. It is the review of a subjective complaint of pain the diagnosis of which can't universally be agreed upon even within the same branch of the medical community. Nonetheless, enough such reports are collected to satisfy the statistical review or an extrapolation to the larger population (an "acceptable" "n" is obtained) and then these subjective injuries are related to collision severity based on grossly subjective and wholly unreliable (un relatable) accounts of the severity of the collision like "he must have been going at least 40 (mph) when he hit me!"
These studies don't - can't - objectively consider things like collision duration, the ACTUAL delta-V for the collision, structural differences between the involved cars, and restrain use (of course, EVERYONE was wearing their belts, right?). The larger problem with these analysis is that it neglects the unknown "n" for that number collisions where nothing was reported to the insurance company or to a doctor and how that might effect the statistical analysis of those cases reviewed. (Are the 2000 cases reviewed part of 200,000 similar cases, 180,000 of which resulted in no claim of injury or could the number be 2,000,000 such cases?)
To the logical, thinking mind, which research is (a) research and (b) which is more meaningful, useful, RELIABLE to an objective understanding of injury potential? If we want to adopt the epidemiological approach, let's also adopt the