The Chicago Tribune recently reported that unpublished government crash tests showed 31 of 66 infant car seats separated from their bases or exceeded injury limits.

The government ran the tests, but never published the data—or even informed many car seat makers of the concerns.

Why did so many infant car seats fail? The government’s tests were more severe than the usual tests that car seat makers use. Infant car seats are tested on a bench seat in a 30 mph head on crash. But these government tests used actual vehicles in 35 mph crashes, both frontal and side impact.

Three specific seats were cited in the report: The Combi Centre/Shuttle, which was recalled and redesigned after failing the government test. Also the Britax Companion “exceeded injury limits”—that seat is now discontinued but still for sale on some sites. The Graco SafeSeat also was cited in the report as flying off its base during a crash—

Graco disputes that report and says the government didn’t properly secure the seat.

The Evenflo Discovery also failed the tests—and this led to a recall of 1 million Discovery infant seats last year.

Of course, the big question here is: why didn’t government make public these crash tests? And are the current bench seat tests accurate in assessing car seat safety?