The Hedgehog, the Fox, and the Magister's Pox

by Stephen Jay Gould

  Thus, for example, Wilson is entirely right (although he calls the process “consilience” in this passage rather than by its rightful name of “reductionism”) that the primary reason for the great success versus comparative failure between two disciplines for study of equally complex systems—the medical and the social sciences—lies in the ability of medicine, and the failure of social science, to achieve reduction to sciences of more-basic constituents, better understood and more easily manipulable. Wilson writes (Consilience, 1998, page 198):The crucial difference between the two domains is consilience: The medical sciences have it and the social sciences do not. Medical scientists build upon a coherent foundation of molecular and cell biology. They pursue elements of health and illness all the way down to the level of biophysical chemistry. The success of their individual projects depends on the fidelity of their experimental design to fundamental principles, which the researchers endeavor to make consistent across all levels of biological organization from the whole organism down, step by step, to the molecule.

  Wilson’s dream, and his avowed purpose in writing Consilience, lies in his firm belief, frankly proclaimed as a metaphysical assumption and not as a proven scientific reality, that the chain of reductionism heretofore so successful in stretching from particle physics well into the reaches of biological complexity, now (and for the first time) stands poised to make its boldest move upward—starting with (and fundamentally encouraged by) our startling initial successes in beginning to understand the workings of the human brain, and then moving through the social sciences and eventually, and ultimately, into the traditional humanities of arts, ethics, and even parts of religion. He writes (page 9):The belief in the possibility of consilience beyond science and across the great branches of learning is not yet science. It is a metaphysical world view, and a minority one at that, shared by only a few scientists and philosophers. . . . Its best support is no more than an extrapolation of the consistent past success of the natural sciences. Its surest test will be its effectiveness in the social sciences and humanities.

  Wilson’s own testimony provides a best sense of the verve and purpose behind his grand vision of full unification along a single consilient chain of reduction, directed, at least in large part, by his feeling for the elegance and beauty, not to mention the explanatory power and potential emotional satisfaction, that would attend the success of such boldness. And yet, albeit by unintentional use of language (I assume), Wilson also exposes an undiminished belief in the superiority of science, and a devaluing based on misunderstanding the aims and definitions pursued by other forms of knowledge and inquiry—an assumption that cannot forge the kind of allegiances he presumably hopes to establish with scholars in the humanities. For example, his explicit definition, in the following statement, of philosophy as “the contemplation of the unknown,” combined with his desire to convert much of this discipline into science (the fruitful study of the knowable and known), will, I am confident, either annoy or at least amuse most professional philosophers. For these scholars, if I understand their enterprise aright, do not define their task as a mere license to speculate or pontificate about things yet unknown, but hold instead that such non-empirical inquiries as rigorous analysis of rules of logic, the structure and classification of argument, and careful examination of the verbal and ideological bases for how people justify and coordinate their beliefs, all constitute valid subjects of study, capable of growth and insight, but not rooted in the scientist’s admittedly powerful procedure of validation by the different criterion of coincidence with the structure of material reality, or explanation by general principles that regulate objects and forces in the physical world. (Of course, philosophers will want to know and study everything that the neurosciences can learn—quite a bit already in fact—about our predispositions not to reason logically in many circumstances, or even generally. But the full analysis of the logic and rhetoric of arguments lies largely outside the compass of material factuality, and therefore represents another form of intellectual inquiry, fully compatible with, and offering much insight to, the work of science, without simply becoming a topmost branch on the single consilient tree of science.)There has never been a better time for collaboration between scientists and philosophers, especially where they meet in the borderlands between biology, the social sciences, and the humanities. We are approaching a new age of synthesis, when the testing of consilience is the greatest of all intellectual challenges. Philosophy, the contemplation of the unknown, is a shrinking dominion. We have the common goal of turning as much philosophy as possible into science. If the world really works in a way so as to encourage the consilience of knowledge, I believe the enterprises of culture will eventually fall into science, by which I mean the natural sciences, and the humanities, particularly the creative arts. These domains will be the two greatest branches of learning in the twenty-first century. The social sciences will continue to split each of its disciplines, a process rancorously begun, with one part folding into or becoming continuous with biology, the other fusing with the humanities. Its disciplines will continue to exist but in radically altered form. In the process the humanities, ranging from philosophy and history to moral reasoning, comparative religion, and interpretation of the arts, will draw closer to the sciences and partly fuse with them.

  Other Wilsonian statements underscore his conviction that the bitterness of past failures and the giddy excitement of present possibilities reside in recent scientific advances, primarily in social theory based on evolutionary biology, and more-conventional reductionistic success in understanding the workings of the human brain. For these reasons, and only now, science can finally resume its reductionistic march by breaching the previous wall against a siege that lasted for centuries: our former inability to reach beyond the mechanical functioning and evolutionary history of complex biological forms, especially as expressed in our frustrating failure to penetrate the workings of the brain—what Darwin had called “the citadel itself,” and what Descartes, while admitting a material substrate subject to scientific understanding, also regarded as the seat of the soul (located perhaps in the pineal gland), the nonscientific “better” half of his great duality between mind and matter. Wilson (page 66) specifically locates past failures in a conjunction of traditional habits abetted by our previous inability to identify the physical basis and character of mind:No intellectual vision is more important and daunting than that of objective truth based on scientific understanding. Or more venerable. Argued at length in Greek philosophy, it took modern form in the eighteenth-century Enlightenment hope that science would find the laws governing all physical existence. Thus empowered, the savants believed, we could clear away the debris of millennia, including all the myths and false cosmologies that encumber humanity’s self-image. The Enlightenment dream faded before the allure of Romanticism; but, even more important, science could not deliver in the domain most crucial to its promise, the physical basis of mind. The two failings worked together in a devastating combination: People are innate romantics, they desperately need myth and dogma, and scientists could not explain why people have this need.

  I have already outlined, on pages 200–203, why I doubt that the pure reductionistic program, Wilson’s full chain of “consilience” if you will, can work either in fact or in principle—and therefore represents the wrong pathway toward such a worthy goal of integration between the sciences and humanities. I do not believe that past failures resided only in a temporary inability to breach a particularly hard barrier (the human mind) in our upward surge to capture new and ever more complex material for inclusion within the reductionistic program. I do revel in the stunning success of the neurosciences. (Speaking personally and emotionally, I could not possibly be more grateful for what we have learned about the genetic components of serious mental disabilities, including the autism of my older son—both for what this knowledge suggests as aid in practical terms, but even more for the emotional and moral liberation thus provided to loving parents
previously blamed by formerly canonical psychobabble for bringing on a condition of such pervasive seriousness by some slight, albeit unintentional, suboptimality in parenting.) To such knowledge and liberation, I can only say: give me more and more, quicker and quicker.

  Reductionism has enjoyed centuries of triumph, and will continue to fill encyclopedias of additional success. God bless. But just as a sequence in size from mite to Ultrasaurus does not imply infinite (or even very distant) further extrapolation, reductionism may not, despite its triumphs in a large domain of appropriate places, be universally extendable as an optimal path to complete scientific understanding. I argued before (see pages 201–203) that two properties of complex systems may deny any dominant status to reductionism, even within the scientific subjects of its evident potential validity; and that, in any case, a third property precludes the incorporation of the humanities into a single consilient chain, whatever the putative success of reductionism in explaining scientific complexity. I shall treat the first, or scientific, reasons here, and save my argument about the humanities for the next section, where my case wins support from Whewell’s own negative views about the potential extension of his concept of consilience beyond the natural sciences.

  I have already praised Wilson for abjuring the first traditional injury of reductionism—the “inbred” tendency of trained scientists to read its hierarchy of subsumption as a statement about relative worth or “maturity” of the various disciplines—in particular, to praise the charms and colors of particle physics while damning the dismal science of economics. Wilson, as an evolutionary biologist (like me) who works near the disregarded end of this chain, recognizes the chief fallacy of this argument—mere rhetoric for silly personal advantage, usable in either direction (and therefore best avoided at both ends). After all, I might praise particle physics as best because all the other sciences derive from more-complex arrangements of the same particles subject to invariant laws discovered within this most basic of all domains. Yet I might, on the other hand, choose to praise evolutionary biology as the highest science because its levels of complexity, including history and interaction, must apply all the explanatory principles learned from all “lower” sciences on the reductionist chain—so this profession must rank as highest for encompassing most. But then, maybe I should just shut up because such silly posturing reminds me of too many arguments I had at ages nine to eleven on the schoolyard of P.S. 26, Queens. And I’ll be damned if a single one of those altercations ever resulted in anything more positive than my opponent’s (very occasional, for I was short, and a less-than-ninety-seven-pound weakling) bloody nose.

  I don’t wish to belabor one of the most thoroughly adjudicated and widely discussed issues of intellectual life. I will therefore, more as a placeholder than as a claim for adequate coverage (and also, in large part, because the next and rather different argument against incorporating the humanities holds more importance in this general brief), restrict myself to reiterating the two crucial claims generally advanced against the full efficacy of reductionism within science. I shall then apply these arguments to a single case, the best recent example in public consciousness, the “deciphering” of the human genome.

  1. Emergence. This debate has become freighted with all the usual academic impedimenta: confusion in prose and, especially, enormous differences in weights and definitions, ranging from the purely and narrowly technical (my usage here, as I shall explain) to the very broadly religious, with emergence extended and misapplied to unresolvable debates about God’s being, meaning, and existence. The basic logical or philosophical question, however, can be posed quite simply. Reductionism works by breaking down complex structures and processes into component parts, and then ultimately explaining the complexity as a consequence of properties and laws regulating the parts.

  Now, and obviously, just knowing the properties of each part as a separate entity (and all the laws regulating its form and action as well) won’t give you a full explanation of the higher level in terms of these lower-level parts because, in constructing the higher-level item, these parts combine and interact. Thus one must also include these interactions as essential aspects of an adequate higher-level explanation. How, then, can reductionism work if interactions among lower-level parts must figure prominently in any higher-level explanation?

  In such cases (effectively including almost any higher-level phenomenon), reductionism still suffices if the interactions can be fully understood and predicted from the parts considered separately. That is, if A and B make C, but if C’s distinctive properties arise by predictable necessity from properties inherent in A and B considered separately, then the reduction still works. That is, we still only need to know the components and principles of A and B in order to predict the form and properties of C. The interactions of A and B may impart distinctive properties to C (as we do not taste salt in either pure sodium or pure chlorine). But so long as we can predict these distinctive properties from knowing A and B alone (as we can infer the production of table salt from two components of such different appearance), then reductionism applies.

  In technical parlance, interactions predictable from the constituent parts alone are called “additive” or “linear.” And so long as interactions remain additive, we can achieve full reduction because nothing must be known or observed exclusively and explicitly at the higher level, in and for itself. That is, we can formulate an explanation, and make correct predictions, simply from our knowledge of the components and their linear interactions.

  But suppose that the interactions among constituent parts do not simply cumulate to build the higher-level result by addition. Suppose, to choose an abstract example (representing a pervasive phenomenon in complex systems, I would argue), we wish to study the ecological interaction between species A and B. Suppose we can predict, from the properties of A and B considered separately, that A will always win under a definite set of circumstances. Suppose, just to be sure, that we also approach the issue experimentally, put A and B together on the same field, under the same simple conditions with no other species present, and A indeed wins every time. These results look good for reductionism. But now suppose that when we add species C to the mix, A beats B only half of the time, with B prevailing just as frequently. Suppose also that we carry the example further and discover that the relative frequency of victory for A or B depends upon hundreds of different environmental factors that we can vary at will, obtaining complex but distinctive outcomes each time, and for each set of factors. Maybe A usually wins when D is also present, but always loses when E occupies the field as well.

  Maybe, to come to the crucial point, all these complex outcomes fall into an interesting order, even leading to fairly precise predictability about A’s or B’s eventual prevalence. And maybe, after studying the system for years, we recognize that this clear and complex order cannot be inferred from the components considered separately. That is, we can only achieve our repeatable results by mixing the components together and observing their interactions at their own level of totality. Suppose, finally, that we can even formulate general principles about the nature of these interactions, but only at the level of their direct occurrence. Such kinds of interactions among components are called “nonadditive” or “nonlinear”—and many scientists, myself included, believe that complex systems may well be dominated by such nonadditivity, thereby precluding reductionistic explanation in principle.

  For if the domination of nonadditive effects requires that we comprehend the regularity of a system by studying its components as they interact all together, and not by isolating each component, and learning more and more until all the interactions can be predicted from the parts, then reductionism fails in principle. Of course, the reductionist will reply that we have taken the easy and unproven way out. The interactions are surely nonadditive, but perhaps we will be able to know and predict the form of this nonadditivity, once we understand the individual components well enough in our complex but basically determinis
tic world. In principle, the reductionist claim could be right; and thus the general debate persists among scientists. But I strongly suspect that irreducibly nonadditive interactions pervade natural systems, and that the number, strength, and determinitive power of these interactions increase as systems become more complex—hence the common feeling that, whereas molecular physics may explain the properties of simple chemical compounds in classically reductionist terms, the physiology of individual neurons may not generate an adequate theory of memory.

  In any case, and to close with a technical (or definitional) point, properties that make their first appearance in a complex system as a consequence of nonadditive interactions among components of the system are called emergent—for the obvious reason that they do not appear at any lower level (and have therefore “emerged” or shown their face for the first time at the new level of complexity). In the strongest form of the argument, we may be able to claim the irreducibility of such emergent features “in principle.” For if these emergent properties simply do not exist at the lower level, and can’t be inferred, as a consequence of their nonadditive character, from knowledge of lower-level components or their interactions at their own level, then these properties have “emerged” at the higher level, and have no standing within any reduced science on the consilient chain. And, finally, if these emergent properties (as they so often do) become central principles of explanation at the higher level, then reductionism has failed, and the higher level must be studied in its own totality if we hope to achieve satisfactory scientific explanation.