In early November, a Massachusetts family of four was hospitalized for acute carbon monoxide poisoning. The first sign of trouble came when the young daughter woke her parents, suffering from nausea and a possible seizure. Not a peep was heard from the carbon monoxide detector--even though the rescue department recorded such a high level of the odorless but lethal gas that the alarm should have sounded.

It's still unclear why the alarm failed. But since 1992, when Underwriters Laboratories Inc. issued a safety standard to make carbon monoxide alarms more widely available for residential use, they have been a constant source of complaints.

First there were a high number of false alarms. Chicago became one of the first major cities to require carbon monoxide alarms in homes, and what happened there has become an infamous tale in the product-safety community.

Chicago's law went into effect on Oct. 1, 1994. Almost immediately, the city's fire department was besieged by calls from homeowners whose detectors had sounded. In one day, more than 1,800 calls were logged by 911.

Almost all were false alarms, triggered by an unusual weather event when the ambient carbon monoxide levels were unusually high. By the end of 1994, only three months after the rule went into effect, more than 8,600 calls had been made to city fire officials--and only a small proportion were due to harmful levels of carbon monoxide.

After a series of emergency meetings with government, industry and UL officials, the UL standard, called UL 2034, was revised to help reduce false alarms.

Now, a number of new studies indicate the alarms either sound too late or fail to go off at all. In 1997, for example, the Gas Research Institute, a private laboratory funded by the gas industry, tested 96 detectors. After three months of operating, nearly half failed to go off when they should have; after six months, one-third failed.

The GRI's conclusion: "The results of these tests raise concerns about whether or not alarms certified to UL 2034 provide consumers with acceptable performance in homes."

Earlier this year, the GRI released the results of a more recent survey: Twenty of 80 detectors, or 25 percent, were defective. Twelve of the 20 defective detectors were "nonfunctional at the time of purchase" and one-third of these defects would not be readily apparent to the consumer.

Debra Rade, UL's chief legal officer and senior vice president of administrative operations, questioned the study's findings, saying the GRI is "funded by the fossil fuel industry," whose products--furnaces and stoves--could trigger the alarms if they are faulty.

Earlier this year, however, the Consumer Product Safety Commission announced a voluntary recall of 1 million alarms because they sounded either too late or not at all.

At its own expense, UL is now conducting field tests on the alarms, buying detectors from stores, placing them in employee homes and testing them periodically to track their performance over time. In one of these tests, 25.3 of the 75 samples failed at least one of the UL-required response tests in homes.

Paul Patty, UL's associate managing engineer in charge of the standard for carbon monoxide alarms, said that, overall, the failures were minor, sounding either a little too early or a little too late. In less than 3 percent of the tests, the alarms failed to go off as they should have before a very high level of carbon monoxide was reached, he said.

Safety officials remain concerned. The CPSC, for example, is gearing up to run its own reliability tests in the next few months.

Of particular concern to the CPSC and the GRI is that the current UL 2034 does not require long-term reliability testing for the gas sensor.

There is a "test" button on the alarm for consumers to push to see if the detectors are still working. But industry officials say all this button really does is let consumers know if the circuitry still works, not whether the detector can still accurately measure carbon monoxide. So while many detectors are being sold with five- or six-year warranties, there's no way for consumers to know if the sensors remain accurate for that long.

Patty said that's because "there really isn't a way of doing [a consumer test] without hitting the alarm with carbon monoxide."

Patty added that warranties are marketing claims that do not fall within UL's domain. "A warranty says that a manufacturer will fix a product if something breaks before the warranty ends," he said. "It doesn't necessarily mean that a product will last that long."

But, Patty added, UL recently has proposed that alarm manufacturers every three months set aside hundreds of alarms and then test these alarms every three months for the life of those products to see how well they age and how long the sensors remain effective.