Consumer Reports said it is withdrawing the report, issued Jan. 4, because some of its test crashes were conducted at speeds higher than it had claimed.
The original report said most of the seats tested "failed disastrously" in crashes at speeds as low as 35 mph. In one test, it said, a dummy child was hurled 30 feet.
But the National Highway Traffic Safety Administration said some of the crash tests were conducted under conditions that would represent being struck at more than 70 mph.
"Consumer Reports was right to withdraw its infant car seat test report, and I appreciate that they have taken this corrective action," NHTSA administrator Nicole Nason said. "I was troubled by the report because it frightened parents and could have discouraged them from using car seats."
In an interview, Ms. Nason said more than 100 worried parents had called the agency's hot line on the evening the original report was released.
Phil Haseltine, the executive director of the National Safety Council's Air Bag & Seat Belt Safety Campaign, said the report had raised doubts among many parents about their car seats despite the "very rigorous standard at NHTSA."
"I think it's going to take a substantial educational effort to undo that damage," said Mr. Haseltine, whose organization was created through a partnership of automakers, insurance companies and safety groups.
Consumer Reports said it would review its study, retest the car seats and publish a new article as soon as possible.
In a statement Thursday, the magazine said it had received information from the NHTSA "concerning the speed at which our side-impact tests were conducted" - supposedly 38 mph. Consumer Reports spokesman Ken Weine said new information from the federal agency showed that the speeds were higher.
The Yonkers-based magazine tested the type of infant car seat that faces the rear and snaps in and out of a base. It found only two of the 12 seats worth recommending, and it urged a federal recall of one seat, the Evenflo Discovery. Evenflo had immediately disputed the tests' validity.
Mr. Weine said a recall is still being urged for the Discovery and for another seat judged unacceptable because it did not fit well in several cars. Evenflo Co. said Thursday that it had run 17 tests on randomly purchased Discovery seats in the past week and the seat passed federal standards each time.
The original report found that all the car seats except the Discovery performed adequately in 30 mph frontal crashes, which is the standard for seats sold in the U.S. But it noted that cars are tested by federal regulators at higher speeds - 35 mph for frontal crashes and 38 mph for side crashes - so the magazine said it tested the seats at those speeds.
"When NHTSA tested the same child seats in conditions representing the 38.5 mph conditions claimed by Consumer Reports, the seats stayed in their bases as they should, instead of failing dramatically," Ms. Nason said.
Consumer Reports' Don Mays, a product safety director, said at the time, "It's unconscionable that infant seats, which are designed to protect the most vulnerable children, aren't routinely tested the same as new cars."
In the 35 mph frontal test, seats separated from their bases, rotated too far or would have inflicted grave injuries, Consumer Reports said. At 38 mph, four seats flew out of their bases after side impact, it said.
Mr. Weine said Thursday that there is no information casting doubt on the 35 mph crashes. He said an internal investigation is under way and he could notsay how the test may have gone wrong.
Consumer Reports, published by Consumers Union, has a reputation for objectivity that it backs by refusing any advertising and by refusing to permit use of its reviews in others' advertising.
It does occasionally get challenged by manufacturers. In 2004, as part of a settlement of an eight-year lawsuit, Consumer Reports said its finding about the Suzuki Samurai SUV - that it "easily rolls over in turns" - applied only to severe swerving turns on the test track.
WHAT TO DO
Consumer Reports asked consumers "to remember that use of any child seat is safer than no child seat, but to suspend judgment on the merits of individual products until the new testing has been completed and the report republished."