REDUCTIONISM AND VARIABILITY IN DATA: A META-ANALYSIS ROBERT SAPOLSKY and STEVEN BALT* I A cornerstone of Western science is an acceptance of reductionism, the belief that an effective way to understand a complex system is to understand its component parts. Such a reductionist framework has been, of course, enormously successful, providing a means for understanding , analyzing, and, if necessary, repairing objects of all sorts: natural phenomena, technological creations, and human bodies. A consequence of the triumph of reductionism has been an asymmetric and often uneasy relationship among scientists, when considering disciplines that differ as to how integrative or how reductive their basic techniques are. There is a tendency for an adherent of a particularly reductive approach to view the more integrative disciplines in a manner approaching hegemony. Consider the pronouncement of Francis Crick that "The ultimate aim of the modern movement in biology is in fact to explain all biology in terms of physics and chemistry" [his emphasis; quoted in I]. Such imperialist fervor need not be restricted to the more reductive disciplines, but can be heard by any scientist viewing more integrative approaches. For example, consider the prediction of E. O. Wilson, much disparaged in behavioral science and social science circles, that disciplines such as ethology and comparative psychology ultimately "are destined to be cannibalized" by neurosciences and sociobiology [2, P- 6]. For the scientist peering from an integrative stance across the divide towards the more reductive disciplines, there is the frustration and invidiousness of "physics envy," the belief that true insight and clarity Support for this study was provided by a Bing Award; manuscript assistance was provided by Russ Fernald. *Department of Biological Sciences, Stanford University, Stanford, California 94305. Correspondence: Robert Sapolsky. FAX: 415 725 5356.© 1995 by The University of Chicago. All rights reserved. 0031-5982/95/3901-0933$01.00 Perspectives in Biology andMedicine, 39, 2 ¦ Winter 1996 | 193 will emerge only with increasing reductionism, leading inevitably to the physicists who proverbially only talk to God. Intrinsic to reductionism is a view about the nature of variability in data. Some variability is deemed legitimate and interesting, as it reflects as-yet-unrecognized factors in the workings of the system under study. For example, data regarding oxidative damage to DNA became less variable once it was known that mitochondrial and nuclear DNA have markedly different vulnerabilities to oxidative attack, and scientists design and interpret their experiments accordingly. From the reductive viewpoint, the other source of variability is little more than an irritant, a problem of measurement instruments—or the humans who use them—not being sufficiently precise; i.e., the variability is simply "noise" that will decrease with improved instruments. As a critical corollary, more reductive levels in science should provide instruments of measurement that are more precise; thus variability decreases, revealing more certain patterns. Most scientists and lay people probably adhere to this fairly intuitive view: it seems plausible that we can learn more about the etiology of schizophrenia, for example, by studying mutations in D2 dopamine receptors than by studying the patterns in Rorschach inkblots seen by schizophrenics. The quantitative tools of successful reductive science involve linear mathematical approaches to problems, and assumptions of additivity of component parts. A challenge to these approaches has become fashionable . This approach, often termed "chaos theory" or "nonlinear systems analysis," relies heavily on the notion that complex systems (which are widespread, as they can be a society, an organism, or a cell) can contain non-additivities and non-linearities, such that a miniscule change in one component part can have vastly amplified and unpredictable consequences (the so-called "butterfly effect"). A major consequence of this is, of course, a grave limitation on the usefulness of reductive study of the component parts of such a system. Intrinsic in the nonlinear approach is a very different view of variability . The emphasis in this new science of complexity and chaos is on the emergent properties of complex systems. In such cases, variability is not mere noise, but is intrinsic to the component parts of the system; moreover , it is independent of the scale of observation. Fractal geometry captures these features in a potent and elegant manner; as one examines a fractal...
Read full abstract