When making things simple does not always give the right answer

Reductionism holds that even the most complicated processes can be explained in simple scientific terms. Steven Rose begs to differ

Reductionism is a portmanteau word; its multiple meanings generate confusion and raise hackles. Reductionism's advocates regard it as the one certain path to scientific knowledge, while its opponents see it as providing an impoverished, sometimes misleading understanding of the world.

The world is a complicated place, full of multiple simultaneous events, processes, causes and effects. Science needs to simplify, to design experiments in which nature is caged, parameters held constant and variables changed one at a time. In studying an enzyme reaction for instance, one might hold the acidity of the solution constant and change the temperature - and derive simple equations describing the consequences. But if both acidity and temperature change simultaneously, as indeed may happen in "'real life", the equations will not work and we lose our capacity to predict.

The triumph of the reductionist methodology of experimental science since its birth in the 17th century has been to extract simplicity from complexity. It has worked brilliantly in dealing with the problems set by physics and by chemistry, even biochemistry, the chemistry of living systems. However, many of the phenomena science wishes to explain, from the dynamics of weather and ecosystems to the history of life itself, from the orchestrated metabolism of a single cell to development of the fertilised egg into the fully formed adult and the workings of the human brain, seem irreducibly complex. Isolating single variables can only confuse. New methods are required.

This is even more the case when we consider reductionism not as methodology, but as philosophy. The classic goal of science has been to find all-embracing explanations - preferably couched mathematically - for seemingly disparate phenomena. Reducing the fall of an apple to gravitational forces is one example, while the classic case is the recognition that the evening star and the morning star are both manifestations of the one planet, Venus. But this does not work too well in biology. Despite claims to the contrary, "a gene" is not straightforwardly identical to "an array of nucleotides within a strand of DNA", and the more we learn of molecular biological mechanisms the less can such a simple equation be seen to hold.

But reductionist philosophy's claims go further. It claims there is a hierarchy of sciences, from, say, the psychological to the physical, and the further "down" that hierarchy we descend the truer are the explanations that we can offer, until we arrive at the so-called "Theories of Everything" beloved of cosmologists and physicists. So the task of science becomes that of replacing psychological descriptions with physiological, physiological with biochemical, biochemical with chemical and physical. As James Watson put it: "There is only one science - physics; everything else is social work."

Consider a frog sitting at a pool's edge. Suddenly it jumps into the water. What is the cause of this phenomenon? A physiologist might say that images on the frog's retina result in transmissions down the optic nerve to the brain, then down motor nerves to the frog's leg muscles, which then contract to create the jump. An animal behaviourist might point to a snake curled round the tree and argue that the frog has jumped to escape the predator. A biochemist would analyse the frog muscle, and describe the sliding filaments of the proteins actin and myosin which compose it, the role of the "energy-providing molecule ATP..." - molecular processes that are understood in exquisite detail.

Each of these orthogonal explanations is a meaningful but partial way of understanding the phenomenon of the jumping frog. Neither of the first two can be reduced to the third, yet for reductionists only this last - and its conversion into physicochemical equations and eventually quantum states - is to be accepted as scientifically "satisfying".

I think this says more about the mindset of scientists than it does about the nature of the phenomena we study. I believe we do live in a world which is ontologically unitary, but to understand it we need epistemological diversity. Certain phenomena are only manifest, and only interpretable, at a particular level of organisation of the world. I study memory, and can recount in great detail the molecular and cellular cascade of processes occurring in the brains of my experimental animals when they learn and remember a particular task. But however complete my biochemical analysis, I do not expect to find brain memory embedded in the molecules. It is a property of the system of interconnected nerve cells, and the organism in which those cells are located, to which the biochemical processes pertain.

Reductionism claims simplicity. While I do not doubt that neuroscience will eventually be able to specify those cells and brain processes that become engaged when a person is angry or in love, I cannot conceive of a time when we would wish to replace the meaningful statements "I am angry" or "I am in love" with a statement about the position and firing state of the hundred billion neurons in my brain and the hormones and neuromodulators circulating between them. The relationship between neurobiological statements about the brain and those about mental states is not one of causation, in the sense that the first is primary, the second merely an epiphenomenon, but of identity: two different and equally legitimate languages in which to describe the same phenomenon. Which language we decide to use should be decided by context, not philosophy.

Which brings me to reductionism as ideology:the seemingly irresistible tendency to seek to explain and indeed "treat" complex social phenomena in terms of disordered brains or genes. The list of such putative explanations is long: genes are supposed to be responsible for violence, "antisocial and criminal behaviour", "impulsivity", non-heterosexual orientation, alcoholism, "female intuition", male adultery and infanticide, even "compulsive shopping". This ascription of unilinear causal power to genes has several peculiar features.

First, the selective choice of behaviours: I am not aware of research on genes for bank fraud or homophobia. Second, the evangelical zeal with which such claims are made by researchers and reported in the press. Third, the neat way that such genetic claims to locate the causes of social problems within individual biology fits the needs of societies, such as those of Britain and the United States.

It no longer becomes appropriate to seek the causes of violence on the streets of the US in terms of poverty, racism or the ubiquity of handguns. Instead research programmes are directed towards locating abnormal biochemistry, itself presumed to be genetically caused, in the brains of inner-city infants that may "predict" an individual destined to violence. Similarly the presence of vodka-sodden drunks on the streets of Moscow prompts a major Russian research effort into the molecular biology of alcoholism. If Karadzic and Mladic ever arrive in The Hague they will probably plead abnormal monoamine oxidase activity in their brains as the cause of Bosnian genocide.

All of which might be dismissable if the genetic claims could be shown to be "good science". However, despite publication in journals of the highest repute, it rarely is. Tentative findings are trumpeted, retractions and failures to confirm received in silence. Claims to have discovered a gene "for aggression" are based on the study of eight Dutch men over three generations, whose crimes varied from violent temper to arson, exhibitionism, rape and murder, all of whom carried a genetic "marker" for one enzyme of neurotransmitter metabolism. If only I could get away with publishing a paper based on the study of only eight chicks!

But beyond such critiques there is embedded within the ideological reductionist paradigm a cascade of conceptual errors. These begin with arbitrary agglomeration - lumping together many different phenomena as if they are all exemplars of the same underlying process. Next comes reification - turning complex social processes involving several participating individuals into "things" or "tendencies" located in the biology of individual participants. Inappropriate quantification assumes that such "lumps in the head" can be measured, so that it becomes possible to say of one individual that they are twice as aggressive as another, or to rank the entire population along a linear scale of "intelligence" measured by IQ tests. Such IQ tests, along with many of the numerological manipulations of psychometry and behaviour genetics, reveal another conceptual error: that of mistaking a statistical artefact for a biological reality.

Finally there comes what I describe as evolutionary fantasies, in which metaphors and analogies are mistaken for homologies. "Aggression" provides a good example. A standard technique in quantifying such behaviour is to find an "animal model", in this case rats, which when placed in a confined tank and presented with mice, will often kill them. How fast a rat kills the mouse, and the hormonal and brain processes that trigger such mouse-killing are taken as surrogates in explaining the causes of drive-by shootings among Los Angeles gangs.

The dangers of such reductionism are manifest: a Gresham's law in which bad science drives out good, a misapplication of scarce scientific resources, the search for technical fixes to complex social problems, the belief that the world of living systems is simple and not complex, and that, ultimately, we are not in charge of our destiny but merely "lumbering robots" serving the needs of our genetic puppet-masters.

Steven Rose is professor of biology at the Open University. His book Lifelines will be published by Penguin this month.