Testing has shown that head loads are not a linear function of impact, but it's close enough that proportionality is used frequently as a rule of thumb for estimating absolute levels. For comparative purposes it's a non-issue, i.e. if a product has a certain head load reduction at 50Gs, say, it's reasonable to see the same general level of performance at different impacts.
[/b]
I'll be Earnhart's belts would have worked fine at 20G, but failed at a higher level. If the SkullSaver Supreme slips out of the harness at a high G level, isn't that a major non-linearity which a low-G crash wouldn't demonstrate?