that does not pose substantial hazards to workers” (American Dental Association v. Martin, 1993), but no cases since the Benzene case have squarely faced the question of how low a risk can be and still be substantial. Instead, the cases have dealt with two basic kinds of problems: the quality of OSHA’s risk evidence, and the reliability of its statistical analysis of that evidence. The ultimate question for reviewing courts is whether OSHA’s findings of “significant risk” are supported by substantial evidence. The standard of review colors court decisions regarding both of these problems.

Texas Independent Ginners v. Marshall is an example of a case turning on the quality of OSHA’s evidence. OSHA had found that American cotton gin workers were at a “significant risk” of developing byssinosis due to cotton dust exposure. OSHA had relied on studies of foreign cotton ginners and American cotton manufacturing employees other than ginners to establish that the prevalence of “acute respiratory problems” among American ginners was likely to lead to byssinosis. There was no dispute that U.S. manufacturing and foreign gin workers experienced more prolonged and higher concentrations of exposure to cotton dust than did American ginners. Since OSHA had failed to offer an explanation for how these differences in working conditions would affect the ultimate occurrence of byssinosis among U.S. ginners, the reviewing court held that the studies would not suffice to constitute substantial evidence of “significant risk.”

Most of the “significant risk” cases deal with OSHA’s choices among statistical methods or scientific data. For example, in ASARCO v. OSHA, members of the smelting industry challenged OSHA’s standard governing exposure to airborne arsenic because it relied on studies treating duration as the key factor in determining the risks of exposure. The industry, backed by its own studies, claimed that it was the level of dosage that determined the level of risk. The court rejected the challenge, reasoning that a dispute in the scientific record did not justify overruling OSHA where OSHA articulated clear reasons for choosing its study over the one in question: “where an agency presents scientifically respectable evidence which the petitioner can continually dispute with rival, and we will assume, equally respectable evidence, the court must not second guess the particular way the agency chooses to weigh conflicting evidence or resolve the dispute” (ASARCO v. OSHA, 1984).

So long as OSHA presents reasonable explanations for accepting one version of scientific analysis over another a court will not engage in an evaluation of OSHA’s decision. However, if OSHA fails to explain why it chooses one method of risk calculation over another, then the court may remand the standard for OSHA to provide these explanations (UAW v. Pendergrass, 1989). Throughout OSHA’s tuberculosis risk assessment OSHA provides reasons for rejecting scientific studies critical of its analy-