The Century of the Gene is, in fact, a jihad against our notion of the gene. Keller insists that the gene is neither the stable, self-replicating entity we thought it was, nor a repository of information about development. To Keller, 'gene' is simply an outmoded term, a semantic straitjacket signifying something that can't be defined. Were she less constrained by publishing convention, I suspect her book would have been called The Century of that Nebulous, Ill-Defined Entity Formerly Known as 'The Gene'.

Keller, a philosopher and historian of science, is best known for A Feeling for the Organism (W. H. Freeman, 1983), her biography of the geneticist Barbara McClintock, which was written for a general audience. Given the high technical level of discussion, The Century of the Gene is, however, clearly aimed at professional biologists.

Unfortunately, the book is long on complaint and short on substance, and ultimately fails to make its case against the primacy of the gene. Despite her repeated claims that the recent history of genetics is replete with "major reversals", "serious provocations" and "radical modifications", the gene emerges unscathed. Many of the alleged problems highlighted by Keller turn out to be semantic issues likely to be of little interest to either working biologists or serious philosophers of science. Moreover, the level of analysis is disturbingly superficial: Keller seems more interested in forcing genetics into the Procrustean bed of her thesis than in presenting a balanced argument.

She claims, for example, that the idea of the gene as a unit of structure or function is outmoded because some bits of DNA do not produce proteins, but instead regulate genes, because some genes can be spliced or read in alternative ways, and because the products of some genes perform several functions. Although it is true that genes are often complex, the word gene is still a perfectly good working term for biologists, especially when defined as a piece of DNA that is translated into messenger RNA. Farmers are still called farmers even though their job is far more complex than that of their predecessors.

Keller asserts that DNA is not a 'self-replicating' molecule because enzymes are needed for replication. She also claims that genes do not direct development because gene activation depends on many different factors (such as chromatin structure, egg cytoplasm and local differences in the cellular environment which turn on different genes in different tissues). Again, these are pseudo-problems: replication enzymes and many inducers of development are themselves products of genes. One might as well argue that political candidates are not self-promoting because they hire others to do that job for them. Certainly, non-genetic factors influence development, but ultimately we differ from chimpanzees because of our genes, not our environments.

The supposed non-autonomy and complexity of genes lead Keller to suggest that we should replace a reductionist approach to genetics with a more holistic programme that incorporates trendy concepts such as developmental networks and self-organization. But she does not specify how this approach would work. In fact, history shows clearly that the greatest triumphs of genetics have been born of reductionism: progress nearly always comes by first studying single genes and then examining their interactions with others. The remarkable advances in understanding the developmental genetics of Drosophila, for example, confirm the value of reductionism in molecular biology.

An example of Keller's one-sided treatment of more substantive issues is her discussion of 'evolvability'. A recent buzzword in evolutionary genetics, evolvability refers to the idea that, in some species, natural selection may favour traits that increase the likelihood of future evolution. There is considerable controversy about whether and how this could occur, but Keller ignores these disputes. Instead, she promotes a particular form of evolvability that, she claims, is both ubiquitous and a radical challenge to modern Darwinian theory. She is wrong on both counts.

Keller argues that species have evolved ways of increasing their mutation rates to generate genetic variation — the raw material for further evolution — and that this evolution undermines the idea that genes are stable. Her evidence for 'adaptive mutability' is the observation that, in some microorganisms, various forms of environmental stress (such as starvation, ultraviolet light or extreme temperature) appear to activate genetic systems that increase the mutation rate. Although most mutations are harmful, some may be useful, and genetic linkage between 'mutator genes' and their adaptive products may drive mutators to high frequencies. Permanently increasing the output of new variants could accelerate future evolution. There are, however, serious problems with this argument.

Keller's prime example of adaptive mutability is the SOS repair system, a mechanism for DNA repair best characterized in the bacterium Escherichia coli. When pervasive, stress-induced damage overwhelms normal repair mechanisms, the SOS system comes into play. This system reverses many mutations, but in so doing introduces a few others. Keller suggests, as do some microbiologists, that the SOS system is an adaptation for increasing the mutation rate under stress. But as this system acts to repair mutations, a more parsimonious explanation is that it evolved simply as a second line of defence against DNA damage and, like many adaptations, is imperfect.

Unfortunately, Keller mentions neither this alternative explanation nor the continuing debate about the nature and meaning of stress-induced mutability. Moreover, she fails to note that selection for higher mutation rates via linkage does not work in sexually reproducing organisms. In such cases mutator genes will be separated from their adaptive products by recombination and then eliminated by natural selection.

Finally, such inducible mutator systems can yield an adaptive response only to factors that impinge directly on DNA molecules. In multicellular organisms with separate germ cells, most forms of selection do not work this way. The presence of lions on the savanna does not increase the mutation rates in gazelles.

Some individual genes, including vertebrate antibodies, have apparently evolved new ways of generating variation as an adaptive response to constantly changing selection. But this, as well as any selection for inducible mutation in bacteria, can be completely explained by evolutionary genetics. By unwarranted extrapolation from bacteria to all organisms, Keller grossly exaggerates the challenge of evolvability to both Darwinism and genetics.

Keller concludes that "gene talk", the argot of geneticists, is passé because of "accumulating inadequacies of an existing lexicon in the face of new experimental findings". Gene talk persists, she says, because it is an easy way for biologists to communicate, and because it helps geneticists get grants and biotechnology companies make profits. Her remedy is to call for a new vocabulary that incorporates concepts from engineering and computer science. Sadly, she fails to suggest what words or concepts we need. Although my enthusiasm for neologisms is limited, they can be useful, as in physicists' distinction between 'mass' and 'weight'. But the notion that geneticists are semantically challenged is simply silly. There is not the slightest evidence that future advances in genetics will be stalled by an outmoded lexicon. What we need is more work, not more words.

The physicist Richard Feynman, famous for his one-liners, supposedly said that the philosophy of science is as useful to scientists as ornithology is to birds. His criticism is overstated, because philosophy can give scientists intellectual perspective on their work. The Century of the Gene, however, ranks as opinionated and poorly informed ornithology. The gene is no albatross.