First, let me say that this article is deservedly famous (cited 1873 times in ISI Web of Science, which is 10 times more than even quite famous and influential scientific papers of similar vintage). It reads best if read as an intentionally-provocative conceptual critique, and there's absolutely value in both intentional provocation and conceptual critique. Every student of evolution should (and probably does) read it.
Having said that, I think Gould and Lewontin deserve heaps of criticism, both of their ideas and of how they present those ideas. Put bluntly, I think they're full of ****. Below are my criticisms, none of which are unique to me, although perhaps the blunt language is. Sorry, but parts of this article really make me angry.
1. The spandrel analogy doesn't work, nor does the analogy of the gaps in fan vaulting, for reasons Robert Mark lays out. Maybe a better example of what Gould & Lewontin mean by a "spandrel" exists, but if so I can't think of one. The deeper problem here is that I simply don't know what is meant by "architectural by-product", and I'm not sure any example could make it clear. The notion of an architectural by-product seems to me to be very fuzzy, and a very poor analogy for any sort of constraint to which functioning organisms might be subject. Maybe this is a case where we'd be better served by actually thinking about functional biology, rather than analogies to functional biology. There are certainly times when a good metaphor or analogy clarifies an argument, but there are also times when a superficially-plausible metaphor or analogy serves only to hide unclear thinking.
2. As we discussed in class, complaining in a generic way about "atomizing" organisms into traits is useless. We can't do without atomization; the phenotype of an organism is too complex to summarize in one number. Fortunately, there are objective strategies available to us to identify which (arbitrarily-individuated) "traits" are correlated with which others, and to what degree. But Gould and Lewontin aren't in the business of talking in a specific way about the practical research approaches real scientists use or could use in order to address precisely this or any other of the issues the article raises. Nor do Gould & Lewontin talk in a specific way about how one would go about actually studying adaptation of integrated whole organisms. I have more to say below about Gould & Lewontin's refusal to come down from the rhetorical clouds and actually engage with how to do science.
Nor can natural selection due without atomization (this is what justifies individuation of traits by scientists). As Lewontin himself once pointed out in print, an organism that wasn't atomized into independent traits couldn't evolve by natural selection, because beneficial mutations would be all but impossible (any mutation would almost certainly just lead to a non-viable organism). Of course, this doesn't mean that natural selection "sees" individual traits. It is indeed true that natural selection only "sees" the whole organism. Specifically, all that natural selection "sees" is (relative) fitness, which is an integrated measure of how well the whole organism is adapted to its current environment. Natural selection only acts indirectly on every other trait, however those traits are individuated. And you know who emphasized this point most keenly? Architect of the Modern Synthesis and famous arch-selectionist R. A. Fisher. Welcome to the selectionist camp, Gould & Lewontin! Nice of you to join us!
3. We shouldn't tell stories about how every feature of every organism is adaptive, say Gould & Lewontin. Fine (obvious, but fine). So presumably that means we should be better adaptationists, and you will describe specific research approaches by which we can achieve that goal? No? You say all you plan to do is tell stories like how the chin isn't an "atomic" trait (despite the fact that one could very well perform an artificial selection experiment to increase or decrease chin size)? Stories like how the small forelimbs of Tyrannosaurus must be an example of correlated growth, without explaining how one would go about testing this hypothesis? I see. Can I make a suggestion, gentlemen? The same storytelling approach that delights readers of popular science columns in Natural History magazine may not go down so well with Ph.D. scientists who understandably will be looking for a little more rigor. Especially when your goal is to argue against the use of story-telling in science!
Gould & Lewontin's lack of willingness to actually say anything about the practical conduct of science is why the last paragraph on p. 151 pisses me off so much. Here, Gould & Lewontin go beyond intentionally-provocative into borderline offensive. Gould & Lewontin complain about how adaptationists admit genetic drift, allometry, etc. in principle, but dismiss their importance in practice and then congratulate themselves for being such undogmatic and ecumenical chaps. To which the only possible response is: How dare you? How dare you tar every single working evolutionary biologist with the same broad brush, without having the guts to actually quote even a single example of anyone dismissing alternative hypotheses to natural selection without evidence? Your claim here is far stronger than the weak, uninteresting claim that some scientists have sometimes drawn mistaken or overhasty conclusions. Your claim here is that the entire field of evolutionary biology is systematically and willfully biased. Which is complete and utter crap. Science is hard, scientists mostly do their best, and their best is mostly pretty good (or do you think we haven't learned anything about how the world works since the 17th century?) How dare you criticize others for how they choose to pursue their science, without saying one word about how to pursue the alternative research program you suggest? Could it be that we do in fact routinely test for and reject some of the hypotheses you claim are ignored (no population geneticist can publish a paper claiming selection on a trait without ruling out drift)? More broadly, could it be that history has shown that a research program focused on testing adaptationist hypotheses is actually a productive way to make progress? And by the way, I love that passage on p. 157 where you point out in passing that correlations of growth (allometric patterns) are themselves subject to selection, only to immediately go back to focusing on non-adaptive hypotheses. I love how you set such a good example for the rest of us by taking alternative hypotheses so seriously.
4. Gould's idea that there's an important distinction between adaptations and exaptations (i.e. traits co-opted by selection to perform a new function) is worthless. As is patently obvious, all adaptations are exaptations if one looks back far enough in the history of life. I'm suspicious that Gould was only able to push this idea (which he did in in several publications) because of how famous he was.
5. Gould & Lewontin claim that drift, developmental constraints, correlations of growth, etc. may well be collectively so important that their effects should be regarded as the "main story" of evolution, with adaptation by natural selection becoming a relatively unimportant detail. What they never bother to explain is how the heck we ever got complex, functioning organisms that could experience genetic drift and exhibit developmental constraints and correlations of growth in the first place? Seriously, how in the name of Darwin was that supposed to happen, except via natural selection?
6. Gould & Lewontin go on and on about how important forces besides natural selection are, but they completely miss all kinds of interesting questions that can be asked about the interplay between multiple evolutionary "forces". They talk as if these other forces are alternatives to natural selection, when in fact it's typically the case that all these forces co-occur and really surprising phenomena arise from their interplay (it's not always that the "strongest" force just swamps all the others). For instance, developmental systems themselves vary and variation in developmental outcomes has fitness consequences, so developmental systems are subject to selection. So how do developmental constraints themselves evolve? Can organisms evolve to become more "evolvable" (e.g., to have more atomized traits)?
Unfortunately for Gould (though not for Lewontin, who was trained as a mathematical population geneticist), developing sensible hypotheses about these kinds of feedbacks between development and evolution pretty much requires mathematical models. Gunter Wagner's work is a good recent example of this kind of modeling. While Gould was famous for knowing a lot about all sorts of subjects, math was a notable gap in his expertise, at least as far as can be judged from his publications. I think this is a very important limitation of Gould's entire approach to science (of which the spandrels paper is merely one particularly famous example). Gould thinks big, but he doesn't think quantitatively (Darwin doesn't either, but Darwin had a good excuse). Reasoning with mathematics is very different than reasoning with analogies and metaphors. Math forces you to be much more explicit and precise in your assumptions and in the logic with which you derive your conclusions, and doesn't leave much room for you to show off your rhetorical skills or your superficial knowledge of architecture. Things like "developmental constraints" and "correlations of growth" have very precise definitions in the context of mathematical models like those of Gunter Wagner, which makes metaphors and analogies superfluous. I don't claim that mathematical modelling necessarily eliminates conceptual arguments or alleviates conceptual confusion (witness ongoing conceptual debate about the interpretation of quantum mechanics), but it often helps a lot, by forcing precision and explicitness.
I'll conclude with a final irony. A hot area of current research is explaining where "correlations of growth" actually come from in the first place. After all, these correlations exhibit some striking quantitative patterns (which I won't review here; take my word for it). Why these particular patterns and not some others? This isn't about whether selection can or cannot modify pre-existing correlations--it's about why these particular correlations exist at all. And were he alive, Gould probably wouldn't have liked the (tentative) answer. All the best-supported current hypotheses argue that observed correlations of growth (i.e. the allometric relationships Gould loved to appeal to so much as an alternative to adaptive explanations) are themselves ultimately a by-product of natural selection on some other organismal feature. For instance, allometric scaling of metabolic rate with body size may well be a by-product of selection for optimizing the functioning of closed circulatory systems (see West et al. 1994 Science). So maybe the biggest constraint on adaptive evolution is ultimately...adaptive evolution. Oh the irony.