International Society for History, Philosophy, and Social Studies of Biology

twitter 2015
     facebook 2015


TUESDAY, JULY 7  /  15:30 - 17:00  /  DS-M340
Organized session / standard talks
Experimental design in the life sciences around 1900

Jutta Schickore (Indiana University, United States); Garland Allen (Washington University, United States)

One of the major topics in the historiography of the life sciences has been scientific experimentation. Historians of experimentation have examined various aspects of experimental practice – research materials, laboratory infrastructure and instrumentation, techniques of visualization, and so forth. In this session, we shift the perspective from materials, practices and instruments to an aspect of experimentation that has received less scholarly attention, namely the methodological concepts and strategies that experimenters were using to support experimental findings and conclusions. We concentrate on the decades around 1900. In this period, new methods of assessing the credibility and significance of experimental research were developed, debated, and – gradually – implemented in a number of different fields, ranging from agriculture to embryology. The papers in this session examine aspects of this process, focusing on key methodological concepts such as randomization, experimental error, and control.

The experimental error of field trials: From conceptualization to computation

Giuditta Parolini (Technische Universität Berlin, Germany)

Since the beginning of the twentieth century the determination, control and reduction of experimental errors have become an essential part of field and laboratory research in the life sciences. Experimental errors have been advocated as a mark of scientific credibility and strategies for their determination and reduction have been devised based on the design of field and laboratory trials and the type of analysis performed on the resulting data. In this process expertise in statistics and computing has proved crucial. Life scientists have forged alliances with statistical and computing consultants or have turned to tools able to facilitate statistical analysis, such as statistical cookbooks or present day software packages. Therefore, the conceptualization and computation of experimental errors are a crucial element to understand how mathematics and number crunching have shaped practical and theoretical developments of the life sciences. The paper will address this issue using as a case study the experimental error of field trials in agricultural science. It will examine the involvement of statistical consultants in the extensive field experimentation performed by a major British agricultural institution, Rothamsted Experimental Station. Concentrating on the first half of the twentieth century, I will discuss how the determination of reliable experimental errors changed field experimentation. In so doing I will examine the statistical tools deployed to tame and control experimental errors, the mutual relationship between accuracy and experimental design, the computing work involved in the process, the presentation of experimental errors in the official publications of the agricultural institution, the relevance of experimental errors for advising farmers on cultivation practices. Despite addressing a specific case study, the paper remains of general interest for all the life sciences due to the success that the statistical methods developed at Rothamsted Experimental Station had in conceptualizing and computing experimental errors across these disciplines.

Randomization? Fisher, "Student", and the Society for Psychical Research

Nancy Hall (University of Delaware, United States)

In his pivotal 1925 book Statistical Methods for Research Workers, Ronald Fisher first published the requirement of randomization in experimental design, to eliminate bias derived from the experimental material or from the researcher, and to make possible a valid test of significance. No where in Fisher’s known correspondence, notes or publications have I found anything directly discussing his development of the concept. But circumstantial evidence suggests two influences: a major one in Fisher’s relations and correspondence with “Student” and a minor one in his advisement to the Society for Psychological Research. In both of these, the concern was mathematics for small samples. “Student” (William Sealy Gosset), a master brewer for Guinness, often had to deal with samples as small as four; the statistics then current were unreliable for samples smaller than about thirty. By 1924, when Fisher was writing the book, he and “Student” had been corresponding for nine years. The Society for Psychical Research, founded in 1882, had the same problem as in many other fields of research: when are results significant and when are they merely due to chance? Fisher did not believe in either telepathy or clairvoyance, but he published three papers discussing the guessing of playing cards, another case of small samples. He analyzed the various probabilities and set up a simple method of scoring. Correspondence makes it clear that Fisher was working on the playing card question in 1923.

"Control(led) experiments" in late 19th-century biomedicine

Jutta Schickore (Indiana University, United States)

This paper offers an interpretation of experimenters’ views on control(led) experiments in mid- to late 19th century biomedical sciences. There are at least two strands in the discussion about controls, one concerning the comparison of populations, the other the comparison of individual experiments. For methodological thought in 19th-century biomedicine, the second tradition is the more relevant. Focusing on concepts of control in late 19th-century bacteriology, immunology, and experimental embryology, particularly in the works of William Welch, Paul Ehrlich, and Jacques Loeb, I show that in these contexts, the concept of control could mean three things: a practice that “controls for” the impact of specific factors on experimental outcomes, a practice that corrects for unknown variables, and the design of new forms of organic life. The first two practices were by no means novel – but only in the late 19th century, they were called “controls”. I argue that the introduction of the term “control” signals a loss of trust in the practical applicability of John Stuart Mill’s methodology of experimentation. Various experimenters maintained that Mill’s methodology expressed an unattainable ideal and that it could not address the most pressing problems of scientific experimentation in the life sciences – the complexity of living things. Control(led) experiments were seen as a realistic alternative to Mill’s ideal.