"In the early days, it was not unusual to see people in first-stage anesthesia -- fairly drunk, staggering -- from solvent exposure," says LaDou, sitting in the sunroom of his Woodside home, a cylindrical structure built from the recycled redwood of an old water tower. "We treated literally dozens of hydrofluoric acid burns every day. The safety and health provisions in these companies were primitive at best."

While such provisions improved in the subsequent decades -- first with precautions like gloves, splash guards, face shields and safety glasses, followed by automatic loading techniques and more sophisticated air-monitoring systems -- risks of workplace exposure by inhalation or skin absorption have by no means been eradicated, says LaDou.

Recall that the percentage of work-loss injuries and illnesses involving "exposures to caustic, noxious and allergenic substances" is three to four times higher in the semiconductor industry than in manufacturing industries as a whole. Even so, LaDou feels that the safety statistics are fundamentally flawed, and reflect a far lower rate of illness than is actually taking place.

	Print story E-mail story

The reporting system for workers' compensation, he points out, dates back to a time when no occupational illnesses, not even lead poisoning, were recognized. It is thus geared toward tracking injured workers rather than sick workers. Because no one is losing fingers making integrated circuits, at first glance the industry can seem relatively safe. Employees who are fighting "psycho-organic" symptoms like headaches and nausea or who are losing weight because of solvent fumes are more likely to end up in the records of a personal physician than in filings with the Department of Labor.

Even when illnesses are properly reported, the industrial codes for workers' compensation offer no way to distinguish between semiconductor employees who actually work in the "clean rooms" -- who constitute approximately 25 percent of the workforce in a typical chip-making company -- and the remaining 75 percent who've never donned a bunny suit in their career. Compare this with the automobile industry, where approximately 90 percent of workers coded as automobile manufacturers are actually working on the auto production line.

"It makes epidemiologic research impossible, unless the company will tell you the magic information on who works where," LaDou says. "Otherwise you're studying people who work in the canteen, or work in offices, or sell product out of their cars." One can understand why LaDou turns ashen with frustration when discussing workers' compensation data. "They're such gross understatements of what's actually taking place," he insists. "We've wanted to look at the industrial hygiene data from inside the fabrication plants, and it's never been published, and it's never been made available to any experts."

Exasperated with the absence of reliable data, in 1998 LaDou and members of a working group developed a preliminary study plan with the Environmental Protection Agency's Common Sense Initiative to measure cancer and birth defects among California semiconductor workers by cross-linking the state's cancer and birth defect registries with an industry-provided database of semiconductor employees, broken down as to which employees held which job and in what occupational setting. With a few additions -- like tracking birth defects and collecting detailed job descriptions -- it was the same idea as a 1983 Swedish study that found an elevated risk of cancer among electronics workers as a whole, and that concluded with a call for further study "focusing on particular features of the work environment."

The EPA put forward $100,000, and California's Department of Health Services, which had been chosen to conduct the study, promised "an umbrella of confidentiality" to protect the privacy of both workers and specific companies.

At the last minute, the semiconductor industry pulled out, led by representatives from IBM and Intel. In a widely reported statement -- leaked to the press in violation of confidentiality rules -- Intel spokesman Tim Mohin declared: "To participate in a project like this would be like giving [legal] discovery to plaintiffs. I might as well take a gun and shoot myself."

The Free Software Project
Read Andrew Leonard's book-in-progress on Linux and open source -- and post your comments.