The Transportation Department will continue its controversial crash test comparisons of new cars despite auto industry opposition, but it will no longer issue "pass" or "fail" ratings, Raymond Peck, the administration's chief auto safety official says.

Peck said Friday that, pending further evaluation of the program, the DOT's National Highway Traffic Safety Administration would publish the crash test results and let the public draw its own conclusions.

The crash test comparisons, in which cars equipped with electronically monitored dummies are rammed into walls, have raised a dilemma for Peck, head of the NHTSA.

The latest test results, obtained by The Washington Post, show a dramatic improvement in the safety performance of a Honda Civic, one of the Japanese subcompacts that NHTSA publicly criticized a year ago for poor passenger protection. The publicity given the crash tests apparently contributed to the safety improvements, Peck said, a plus for the program.

But the results also include a very poor performance by a Ford Motor Co. EXP, one of the new front-wheel-drive subcompacts important to Ford's success. In the EXP test, the head of the passenger dummy struck against the dash with a enough force to kill or seriously injure a human, according to NHTSA's analysis.

Peck challenged the validity of the EXP test results after watching replays of a movie of the crash in his office Friday and then consulting his agency's experts. Peck said that the EXP would almost certainly be retested but that he would wait until today to disclose the apparent defect in the test that led to his decision.

Peck said he was concerned that the EXP situation might damage the program's credibility--already under attack from the auto industry. The crash ratings, begun two years ago, were a key part of NHTSA's consumer information program during the Carter administration. The auto industry and consumer groups have been waiting to see what Peck and the Reagan administration would do with the program.

In the crash tests, the cars are propelled into a wall at 35 miles per hour, either head-on or at an angle. Dummies are seated in the driver and front passenger seats, restrained by standard seat belts and shoulder straps. Electronic sensors monitor the crash impact on the dummies, recording particularly high scores when a dummy's head hits the dashboard or windshield, or when the steering column spears the driver dummy.

Low scores indicate a high likelihood that humans would have survived the crash. High scores indicate a probabilility of serious injury or death.

During the Carter administration, NHTSA picked specific threshold numbers for evalulating the tests results. Cars with scores above these numbers were listed as having failed the test and vice versa for scores below the mark. These pass-fail ratings were publicized at press conferences and in NHTSA's "Car Book" of consumer information.

The improvement in the Honda was precisely what NHTSA was hoping for when it began the crash ratings.

After its 1980 Civic performed so badly, Honda changed the material used in the safety belt and relocated the belt anchors. The steering column was redesigned to lessen the danger to the driver. Then Honda asked NHTSA to test a 1981 Civic with the modifications. It passed with flying colors, indicating that "relatively minor improvements can significantly effect the test results," Peck said, at least for the cars with the worst performance.

"Honda has reduced the potential for injury very significantly apparently without major cost," said James Hackney, an NHTSA safety expert.

The question of the accuracy of the EXP rating, however, illustrates the auto industry's chief complaint about the comparisons--that an entire car line should not be branded because of the results of one crash test.

Roger Maugh, Ford Motor Co. safety director, said the crash tests are useful in guiding the design of safety features, but are worthless for comparing the safety of various cars.

Ford has crashed more than 20 cars of the same model and found differences of up to 40 percent in the numerical crash force readings, he said, calling the DOT "pass-fail" ratings an "insult" to scientists.

Hackney countered that high crash numbers usually indicate defects in passenger restraint systems or design common throughout a car line.

Where Peck will side in this dispute remains to be seen.

The latest crash data were obtained through a Freedom of Information Act request. A half-dozen cars remain to be tested in the current series, and NHTSA said it planned to release all the results at one time. The results available now cover thirteen 1981 models.

Among the light subcompacts tested, only the two-door Honda Civic and a two-door American Motors Corp. Spirit met all of the NHTSA safety criteria. A large two-door Chrysler Imperial also met the criteria.

The Ford EXP two-door hatchback, a four-door Honda Civic, and a four-door Toyota Cressida met all but one safety standard, NHTSA said. In the Cressida, the hub of the steering wheel was pushed up into the "driver" by the crash, NHTSA said. Inadequate seat belt protection was apparently a factor in the crash results with the four-door Honda, the agency said.

Cars failing to meet two or more of the crash standards were a Toyota Starlet, an Audi 5000, a Renault 18i and a Volkswagen Jetta. The Toyota Starlet's steering column "performed poorly in the crash," NHTSA said.

DOT also released reports on 35 mph tests of five other models that were crashed at an angle rather than head-on. A Plymouth Reliant, a Ford Escort, a Mazda GLC and a four-door VW Jetta all passed, and a Renault 18i was barely over the limit on the passenger side. These results are not comparable with those in the head-on crashes, which generate far greater impact forces.

DOT did not test 1981 cars essentially unchanged from 1980 models whose crash test results were previously reported.

A key test result records crash forces measured by sensors in the "heads" of the dummies in the driver and front passenger seats. NHTSA considers that with scores above 1,000 there is at least a 50-50 chance that a human occupant would have suffered serious injury or been killed.

Results for 1981 models in 35 mph crashes, driver and passenger sides, respectively, are as follows.

Head-on crashes:

AMC Spirit (2-door), 702 and 652. Audi 5000 (4-door), 1,288 and 1,583. Chrysler Imperial (2-door), 976 and 590. Ford EXP (2-door), 745 and 2,668. Honda Civic (2-door), 607 and 492. (The results for this model a year ago were 2,626 and 1,506). Honda Civic (4-door), 985 and 1,391.

Renault 18i (4-door) 1,150 and 1,659. Toyota Cressida (4-door, automatic belt system) 1,980 and 771. Toyota Starlet (2-door) 1,836 and 1,351. Volkswagen Jetta(4-door, automatic belt system) 1,210 and 1,272.

Angled front crash,(not comparable with head-on crashes):

Contact point, left front fender:

Mazda GLC (4-door), 555 and 516. VW Jetta, 161 and 517.

Impact on passenger side:

Ford Escort, 243 and 603. Plymouth Reliant, 355 and 566. Renault 18i, 731 and 1,001.