It's ironic that the author complains about a contempt for science on the part of "environmentalists," but then goes on to recommend the following rule for sett

It's ironic that the author complains about a contempt for science on the part of "environmentalists," but then goes on to recommend the following rule for setting safe levels of some toxic substance: Wait for a major kill-off of a marker species, such as fish, in response to some level of that substance, then set the limit "an order of magnitude or so" lower. The author correctly points out that it's unethical to perform controlled experiments with suspected toxins. However, his response seems to be to give up any attempt to gather evidence (short of watching for massive die-offs of fish). For instance, he complains about the impossibility of conducting experiments to learn about arsenic poisoning, and concludes that one therefore cannot "obtain valid data" about the effects of arsenic. To make this argument, he mis-characterizes the Bayesian position as, effectively, "perform controlled experiments." He evidently never heard of epidemiology, which attempts to make sense, statistically, of all the millions of experiments that nature and chance play on people all the time (http://en.wikipedia.org/wiki/Epidemiology http://en.wikipedia.org/wiki/Epidemiology). While epidemiological evidence does, in fact, require careful analysis, ultimately one can back out valid inferences, given enough data. Sadly in the case of arsenic there are no end of data, provided by people drinking from contaminated wells.