This session examines four cases of errors in biology, how biologists found them and then recovered from them. In the same way that lack-of-function studies in organisms can lead to understanding of biological structure and function, such historical analyses can lead to deeper understanding of the methods for developing reliable knowledge in science. In extending the work of William Wimsatt, Lindley Darden and Deborah Mayo, we hope to illustrate the value of investigating the epistemology of error and the reasoning by which scientists use anomalies to develop scientific concepts.

Organized by: Douglas Allchin

Kevin Elliott, University of Notre Dame
"Error as Means to Discovery"

The work of several philosophers, including William Wimsatt, Lindley Darden, Deborah Mayo, and Douglas Allchin, suggests that the identification and confirmation of scientific error plays an important role in scientific discovery. William Wimsatt, in "False Models as Means to Truer Theories" (1987), outlines a process for progressively improving the fit between scientific models and empirical data. He shows how one can use even false models as templates in order to discover "residual," anomalous effects that lead to the identification of errors and the development of more complex models. The first section of this paper summarizes previous work by Wimsatt, Darden, and Allchin and argues that further investigation of the reasoning by which scientists use anomalies to identify and confirm scientific errors will contribute to a deeper understanding of both the productive role of error in scientific discovery and the epistemology of scientific error. Section II then presents a case study that describes anomalous data concerning the low-dose biological effects of toxic and carcinogenic chemicals. These anomalous data may indicate that the toxicological models used for developing United States regulatory policy regarding exposure to carcinogenic chemicals are erroneous. Section III draws four insights from this case study. First, it argues for two weaknesses in Darden's previous accounts of the use of anomalies to identify error. Second, it suggests alternative strategies for using anomalous data to identify and confirm the presence of error in a scientific model or theory. Third, depending on the sorts of errors that may be highlighted by the anomalous data, the case study offers possible techniques for changing false models in response to anomalous data as part of a gradual discovery process. Finally, it argues that, to the extent that policy-related science develops via the gradual identification and confirmation of error described in this paper, this case study suggests several ramifications for science policy and education.

Nancy Hall, University of Maryland
"Randomize? Why Bother? Because Your Experiment May Be Disaster Otherwise! (Fisher, Lanarkshire, and Darwin)"

Sir Ronald A. Fisher, statistician and geneticist, was an advocate for randomization in experimental design, beginning in 1926. Then in 1930 in Scotland the Lanarkshire Milk Experiment was conducted, by the Department of Health, to investigate the advantage of giving extra milk, raw or pasteurized, to school children. But the experiment, involving 20,000 children, was ineptly carried out. After the supposed results were announced, Fisher, his good friend "Student" ( of Student's t-test) and several others discussed possible ways to salvage the situation. In 1935 Fisher published The Design of Experiments; in it he critiqued Charles Darwin's 1870's experiment with fifteen pairs of cross- and self-fertilized plants. Darwin took great care to try to choose plants of identical age and ancestry, and to try to equalize the growing conditions. Further, he asked Francis Galton to analyze his results statistically. In Fisher's opinion, Darwin should have flipped a coin, and Galton's statistics were in error. [ Full Paper ]

Rivers Singleton, University of Delaware
"Biochemical Error: Seeing Old Data with New Glasses"

In the early 1930s, Hans Krebs began elucidating the biochemical pathway whereby glucose is oxidized to CO2 and water. Within a decade he put together a cyclic sequence of reactions, which Krebs "baptized" "... the citric acid cycle after its characteristic intermediate." Carbon isotopes (11C and 14C, which are radioactive, and 13C, which is non-radioactive) then became important tools to trace carbon flow in metabolic processes. In a series of experiments using 13C as a tracer, Harland Wood demonstrated that citrate did not appear to be an intermediate in glucose metabolism. (At about the same time, Earl Evans presented similar data using 14C as tracer.) Today we know that Krebs' original formulation of the citric acid cycle was essentially correct despite these powerful experimental objections. In this paper I will explore the nature of Wood's objections to citrate involvement and show how they arose from a fundamental misunderstanding of carbon chemistry.

Douglas Allchin, Minnesota Center for the Philosophy of Science
"To Err and Win a Nobel Prize II: Peter Mitchell's Mistakes in Developing Chemiosmotic Theory"

[tentative]