How Science Works

Statistical literacy is becoming an issue of growing concern for scientists and philosophers alike, the underlying drive being the social impact it has on a variety of realms. Sociologists (e.g. Beck) have made us aware of the relevance of risk assessment in our everyday life, while psychologists (e.g. Gigerenzer) revealed that our statistical estimations are usually biased, albeit for good evolutionary reasons. To trust in the statistical expertise of scientists might then seem a wise choice when their empirical findings are invoked to support such decisions as, say, the promotion of vitamin C for our health improvement or a conviction on the basis of DNA identification. The immediate question is to what extent this public trust is justified.

The improvement of our understanding of statistical evidence in biology and medicine is precisely the goal of How Science Works and the argumentative strategy of the biologist Stephen H. Jenkins could not have been more effective. Each of the nine chapters of his book provide an introductory presentation of the evidence available concerning an equal number of fascinating topics: besides the two mentioned above, Jenkins discusses environmental and ethological issues (such as the decline of certain amphibian populations or the cognitive mechanisms that control the storage of food in certain species), medical subjects (the causes of cancer and aging, the effects of coffee), or a combination of both (the influence of climate change on the spread of malaria). Intended as brief reports on the work in progress in all these areas, each one of them contributes to the making of a single methodological point: what the statistical analysis proves in these nine cases is that there are as yet no ultimate answers.

What Jenkins argues is that this inconclusiveness arises from the very complexity of biological and medical research, even after being tamed by the most sophisticated statistical techniques available. There are many sources of evidence and various possibilities to incorporate them into our theories. On the one hand, Jenkins presents a number of standard bio-statistical methods that allow us to design experiments and analyze data with various degrees of reliability. The art of the biomedical scientist would lie in the achievement of a certain consilience between all these sources, together with more traditional ones, such as fieldwork. Jenkins shows that this is not an easy task, as the evidence might prove unavoidably conflicting. Several philosophical dilemmas come to complicate even more data analysis. To name just three, let us think of the typical disjunction (sometimes inclusive, often exclusive) between case studies and all-encompassing theoretical approaches, the difficulties of accounting for the simultaneous interaction of various levels of causality and the pros and cons of purely statistical vs. mechanism-based modeling.

Though it might seem impossible to compress such a range of topics into about 200 very readable pages, Jenkins succeeds in providing a survey that might be of equal interest both for the curious reader in search for a popular state of the art discussion of the nine mentioned topics and for the methodologically-minded scholar looking for an introduction to the beauty and perplexities of biomedical research. Footnotes are kept to a minimum and several suggested readings complement each chapter. Jenkins's approach might well be deemed "ecological": though he clearly favors certain approaches, his main concern throughout the book is to show us how many of them proliferate now in his field and what sort of results they yield. Only by taking all this into account will we be able to qualify our trust in the alleged scientific grounds of many policies currently under discussion. In other words, the idea of informed consent should be generalized as to encompass (and legitimate) every public decision on environmental and medical matters.

Though Jenkins' case is articulated and convincing, it might be also a bit puzzling. We may well be aware of the difficulties of attaining conclusive results in the biomedical sciences, but to whom are we rely on to assess them? Jenkins's implicit assumption seems to be that there will always be non-partisan scientific experts to provide a reliable assessment. Yet, it might happen that controversy among scientists could prevent general audiences from discerning whom to trust. This is a classical dilemma: quis custodet ipsos custodes? Be it what it may, Jenkins' proposal will probably be part of the solution to this problem.

© 2004 David Teira

David Teira, Ph.D., Department of Philosophy, University of Salamanca (Spain)