BBC NEWS | Health | Doubt over shaken baby diagnosis: "Scientists have cast doubt over the theory that certain eye injuries are a sign a baby has been violently shaken.
This whole business is a much greater scandal than physician over-enthusiasm for estrogen replacement.
In the 80s we thought certain vaginal findings were proof of abuse. Until recently in the UK a second case of SIDS in a family was considered proof of murder. Some have thought that retinal bleeds were a sure sign of shaken baby syndrome. Before we understood von-willebrand's disorder I'm sure bruises in affected children led to abuse cases. At one time certain warts were felt to guarantee child abuse (that may still appear in textbooks, but I wonder how good the science is).
Crappy science, crappy logic. There's something about the horror of infant abuse that drives people to nutty conclusions, and drives thoughtful researchers far from the domain. There is no good, reliable, test for child abuse short of the most extreme findings. I doubt we'll ever have one -- unless we research in advanced "lie detection" actually works out. We'll either err towards removing children from loving and trustworthy homes, or err towards leaving them to be hurt and abused. There are no perfect solutions, which is why child-protective services is such a terribly hard business.