Claude Bernard was one of the great scientists of the 19th century, and arguably of all time. By most accounts (including his own), he played a pivotal role in reshaping physiology into an experimental science, based on physics and chemistry, which sought to uncover the mechanisms behind normal and pathological processes in organisms by deliberately and systematically manipulating them in precisely reproducible ways. This ideal continues to dominate large parts of biology, including very much molecular biology, and even influenced psychology (via behaviorism) and the arts (via Zola, whose grasp of what it meant to conduct an experiment was at best tenuous). Two key ideas we normally associate with the 20th century — falsificationism and homeostasis — were explicitly central to Bernard's thought, but I've written about that elsewhere and won't repeat myself.
Bernard did, however, have a scientific weakness, which was to disdain quantitative studies in biology, and more especially statistics. He was very much of the school which held that the need for statistical analysis was never anything more than a sign of bad experimental design. (He put it much more elegantly.) He also looked down on contemporary quantitative studies. Kelvin is supposed to have said that "When you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot express it in numbers, your knowledge is of a meager and unsatisfactory kind; it may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of science." Bernard (as it were) pre-emptively countered this argument, as follows:
The application of mathematics to natural phenomena is the aim of all sicnece, because phenomenal law should always be mathematically expressed. To this end, data used in calculations should be results of well-analyzed facts, so that we may be sure that we fully known the conditions of the phenomena between which we wish to establish an equation. Now, I think that efforts of this kind are premature in most vital phenomena, precisely because these phenomena are so complex that we must not only assume, but are in fact certain that, beside the few among their conditions which we know, there are numberless others which are still totally unknown. I believe that the most useful path for physiology and medicine to follow now is to seek to discover new facts instead of trying to reduce to equations the facts which science already possesses. This does not mean that I condemn the application of mathematics to biological phenomena, because the science will latter be established by this alone; only I am convinced that, since a complete equation is impossible for the moment, qualitative must necessarily precede quantitative study of phenomena. [An Introduction to the Study of Experimental Medicine, p. 129]In other words, a meager and unsatisfying understanding can very well find its expression in numbers, if prior scientific knowledge hasn't marked out what you should be measuring, how you should be measuring it, and what other measures you should be relating it to. The difficulties of physiological measurement are compounded by the fact that organisms have inner environments, physio-chemical variables internal to the organism, whose homeostatic maintenance makes organisms more or less autonomous, and these too need to be identified and measured. Though I am myself a quantitative scientist and a statistician — in fact, co-author on a just-submitted paper on the statistical analysis of neurophysiological data — I must say that Bernard had a point. If I look at successful ventures into quantitative biology (e.g.), there is a long history of qualitative science behind them, getting us to the point where do grasp at least some of the "conditions of the phenomena".
But I promised cats. Bernard made all these points in, among other places, his classic Introduction to the Study of Experimental Medicine, giving over a fairly long section (part II, chapter II, section IX; pp. 129--140 of the English translation) to explaining why he did not look kindly on "calculations in the study of living beings", permitting himself what I can only call snark. Drawing a veil over the episode in which an enterprising physiologist claimed to be collecting samples from a railway station men's room (pp. 134--135; perhaps a precursor of Sen. Craig?), let us consider cats (p. 132):
In the part of their investigation devoted to nutrition, Bidder and Schmidt described a very notable experiment, perhaps one of the most laborious ever performed. From the point of view of elementary analysis [i.e., tracking the amounts of different chemical elements], they kept a balance sheet of everything taken in and given out by a cat during eight days' nourishment and nineteen days' fasting. But this cat was in a physiological condition of which they were unaware; she was pregnant, and had her kittens on the seventeenth day of the experiment. In these circumstances, our authors considered the kittens as excretions, and calculated them with other eliminated materials as a simple loss of weight. I believe that these interpretations should be rectified when trying to define such complex phenomena.The fate of the cat and her kittens is not recorded, but I like to imagine them slinking around Bidder and Schmidt's laboratories in Dorpat like a pride of domestic sphinxes.
Manual trackback: Siris
Posted by crshalizi at December 28, 2007 18:06 | permanent link