Chicago ? In a government crash-test video, the infant car seat flies off its base, smashing the baby dummy — still strapped into the carrier — upside down and face-first into the back of the driver’s seat.

Think what could happen in a real crash.

This seat was one of 31 that either flew off their bases or exceeded injury limits in a series of frontal crashes conducted by federal researchers using 2008 model year vehicles, a Chicago Tribune investigation found. The test results were never publicized, and even some infant-seat makers were unaware of their existence.

The Tribune found the results buried in thousands of pages of test reports from the National Highway Traffic Safety Administration. These tests are used to rate the safety of cars, not the child restraints in them.

What the newspaper unearthed calls into question the rigor of the current safety standards for such seats. The investigation also highlights how little information parents are armed with as they make one of the most important safety decisions for their babies.

Flaws in safety standard

You can compare safety ratings for cars, but not for the safety of car seats. Parents often have no way of knowing which seat fits best in their car and whether conventional wisdom is accurate. For instance, the Tribune found that two of the most expensive seats tested had some of the poorest results, and that some small cars protected car seats better than larger ones.

“What you’ve uncovered totally reveals the flaws in the current safety standard and also NHTSA’s negligence in not reporting this to the public,” said Joan Claybrook, a former NHTSA administrator and president emeritus of the advocacy group Public Citizen.

Responding to the Tribune investigation, newly installed Transportation Secretary Ray LaHood said in a written statement Friday that he ordered a “complete top to bottom review of child safety seat regulations” and directed the NHTSA to make the crash-test results “more available” to consumers.

Of the 66 infant seats tested in frontal crashes, nearly half of the seats either separated from their bases or exceeded injury limits.

The government describes the tests as research. But the results for two seats were so troubling that NHTSA recalled those seat models and one manufacturer completely overhauled how it evaluated its seats.

The infant restraint in the crash video, the Graco SafeSeat, flew off its base, but the NHTSA said it didn’t seek a recall because the seat remained connected in five other vehicle tests and, thus, it was “not a repeatable event.”

In a statement, Graco’s parent company dismissed as “anomalous” the one SafeSeat test where the seat separated. Dale Matschullat, general counsel for Newell Rubbermaid, wrote in a letter to the Tribune that all the crash tests were “purely experimental” and are “worthless for purposes of evaluating and comparing infant restraint system performance.”

The NHTSA said it is analyzing all of the test results and doesn’t yet know what they mean. Car seats save lives. If infant seats performed as poorly on American roads as they did in these crash tests, said Ron Medford, NHTSA’s acting deputy administrator, “We would expect to see higher numbers of fatalities or serious injuries than we’re aware of.”

In 2007, 63 babies were killed and about 7,000 were injured in crashes where they were strapped into infant restraints.

Before being sold, seats must pass a test that simulates a head-on crash at 30 mph on a sled bench. In the tests analyzed by the Tribune, regulators crashed actual vehicles into a wall at 35 mph.

Claybrook said the crash tests suggest something that is common sense: the effectiveness of car seats can be more thoroughly judged when evaluated inside a real car as it is crashed.

The Tribune found, for instance, that many of the higher injury ratings occurred when the baby dummy’s head hit the rear of the vehicles’ front seats. The sled tests wouldn’t pick up this potential injury because there’s no front seat to hit, so regulators are considering adding one.

In addition, manufacturers do not have to test their seats in side-impact scenarios. However, when the NHTSA crashed 3,000-pound barriers into the sides of real cars, Evenflo Discovery infant seats repeatedly flew off their bases. This led to the recall of 1 million of the seats.

The NHTSA says it is studying how to improve the sled tests. And in January 2008, it began deploying teams of investigators to reconstruct certain accidents involving young children in car seats.

Seats properly installed?

Other car-seat manufacturers reacted not by questioning the sled tests but by questioning how the crash tests were performed.

In disputing the results, Dave Galambos, a Graco Children’s Products manager who was in charge of testing the company’s car seats for years, said the NHTSA contractor must have improperly installed the SafeSeat in the test in which it flew off the base. Galambos points to the slow-motion videos of the test as evidence that the infant carrier wasn’t securely snapped into its base.

NHTSA officials reviewed the same data and said they believe the seat was properly snapped into its base.

The allegation that the infant seats were improperly installed during NHTSA’s crash tests was common among the car-seat makers. But federal safety officials stressed that all car seats for the tests were installed by highly trained technicians.