The electrical fire in Barbara Ware's home was bad enough; it happened on a Sunday morning, when most of her family was asleep. But most distressing was the failure of her two-week-old smoke detectors to sound an alarm as her Beaumont, Tex., house filled up with smoke.

Ware complained to the manufacturer, BRK Brands, which asked her to return the detectors. After testing them, BRK wrote Ware and said they failed because the fire involved "cold smoke"--a term to describe what happens in a smoldering fire that's not hot enough to drive the smoke up toward the ceiling and the detector.

"Had this been a serious fire, this detector would have alarmed," BRK said, noting that when her detector was retested, it passed the smoke-sensitivity safety tests required by Underwriters Laboratories.

That assurance did little to appease Ware. "There's no such thing as a fire that's not serious," Ware testified in a 1998 court case against BRK, which sells its detectors under the First Alert and Family Guard brands.

Nor did passing the UL test satisfy Iowa District Judge David E. Schoenthaler, who heard the case.

Last year, he upheld a jury's decision to assess $12.5 million in punitive damages against BRK for a detector that failed to sound an early warning in a fire that killed a 3-year-old and severely burned an 18-month old after a faulty baby monitor (also approved by UL) caught on fire.

Schoenthaler's opinion said trial testimony showed that BRK received about one complaint a week about ionization smoke detectors failing to sound an alarm when there was smoke. But when retested by BRK, these detectors passed UL's safety standard. Schoenthaler's conclusion: The UL standard "is not an adequate predictor of the [detector's] performance in real world fire events."

The case is now being appealed to the Iowa Supreme Court, but Schoenthaler's opinion is part of a chorus of criticism concerning UL's testing process for the ionization smoke detector.

Ionization models account for 90 percent of the smoke detectors found in U.S. homes, largely because they are the least expensive detectors on the market. They're good at responding quickly to flaming fires but don't respond as fast to smoldering fires, such as when a sofa or mattress is ignited by a cigarette.

BRK President Michael Paxton said the company receives very few complaints about faulty detectors. The ionization detector has saved "thousands and thousands of lives," he said. "In just the last 10 years, there's been a 35 percent drop in fire deaths." Paxton said smoldering fires account for only 6 percent of the nation's fires.

But to Joseph Fleming, Boston's fire marshal and deputy fire chief, it is the smoldering fires that, by their very nature, demand accurate smoke detectors because they give people "plenty of time to fall asleep before the hazard develops."

Fleming has spent the past 10 years researching smoke detectors and has concluded that the ionization detector does not provide sufficient protection in smoldering fires.

The reason: UL's smoldering-fire test was written more than 20 years ago and has not been updated to reflect the different synthetic materials now used in upholstered furniture and mattresses.

So while an alarm may sound in UL labs, it may not go off at residences because the smoke particles that are released during the fire may be radically different today than they were 20 years ago.

Fleming points out that UL puts smoke detectors through four different tests to measure their response to fast-flaming fires, but only one test to measure response to smoldering fires.

For at least a decade, local fire officials and federal safety experts have urged UL to reconsider this test. In November 1994, for example, the staff of the Consumer Product Safety Commission told UL it was concerned that the smoldering-fire test "does not represent the smoke produced in residential smoldering fires."

UL engineers have repeatedly said they believe their test replicates a true smoldering fire. According to UL minutes of a March 1997 meeting of industry executives and local and federal fire officials, UL engineers said they were confident that the company's test was "a consistent, representative test for smoldering smoke."

The engineers added that "to develop a test with the most 'accurate' smoldering particles would be extremely difficult, if not impractical based on all the variables."

Only recently has UL indicated a willingness to review its smoldering-fire test.

"We're working with the [CPSC] right now" to see if other smoldering-smoke profiles are warranted, said Paul Patty, the UL associate managing engineer who oversees tests of smoke detectors.

Underwriters Laboratories

Headquarters: Northbrook, Ill.

Founded: 1894 by William H. Merrill, an electrical investigator

What it does: Sets more than 700 safety standards to test 17,000 kinds of products. In 1998, it conducted 89,630 product eval-uations and 14.7 billion products were made bearing the UL mark.

Facilities: 13 laboratories worldwide

Employees: 5,258

Assets: $512 million

Annual revenue: $407 million ($376.3 million comes from companies paying for UL's testing services)

SOURCES: UL documents