More Articles on Evolution
The Scope of Natural Selection
Daniel C. Dennett
When Professor Orr published his hostile
review of Darwin's Dangerous Idea in the biology journal Evolution last February,
I was not pleased. His review was full of falsehoods and misconstruals, but
I had no recourse; that journal, like most academic journals, does not permit
authors to respond to reviews. Luckily for me, Orr was so eager to warn the
world of my errors that he restated his attack, with embellishments, in Boston
Review, which has invited me to respond. Months have passed, the damage has
been done, but at least I get to set the record straight.
I am grateful to Orr for embellishing his attack for the benefit
of his lay readers, since these additions vividly expose his own errors and
confusion, which were somewhat masked in the more professional version. In
what follows I will concentrate on his criticism of my understanding of his
field: biology. I trust that the readers of Boston Review have seen the flaws
in his philosophical arguments without my help, but a non-biologist might
well suspect that, on his home ground, Orr's complaints are as authoritative
and devastating as he makes them out to be. They are not.
Before getting to the meat of his criticism, Orr warns readers
that my book is:
marred with factual errors, some scientific and some
historical. Population genetic theory, for instance, does not prove that evolution
by random change is faster than evolution by natural selection, as Dennett
I did indeed misspeak (p. 126), but the result was ambiguity,
not error. The issue is complicated: it depends on whether you're measuring
the (average) speed of departure from a starting point in genetic space, or
the speed of attainment of some particular evolutionary product. I meant the
former, as the context ought to have made clear (I was in the process of observing
that there has been plenty of time in evolutionary history for the sheer genomic
diversity observed to accumulate). But I did couch my claim in terms that
permitted Orr to put the false reading on it. Score one half point for Orr.
He goes on:
And it was Darwin's theory of sexual, not natural,
selection that he called an 'awful stretcher.'
I made no such claim. I entitled a chapter section "Natural
Selection-an Awful Stretcher." The vivid epithet comes from Darwin, but
it was I who applied it to natural selection. Orr, having already stretched
too far in his effort to find an historical error to impute to me, topples
over when he tries to elaborate: "Darwin branding natural selection a
stretcher is about as plausible as Chomsky branding Universal Grammar a howler."
What? Darwin branded sexual selection an awful stretcher in the very act of
endorsing it-it is, after all, a special kind of natural selection, not an
alternative! Here's what Darwin said (as I noted, when I introduced the term):
"It is an awful stretcher to believe that a peacock's tail was thus formed;
but, believing it, I believe in the same principle somewhat modified applied
to man." Surely Darwin would have agreed, if asked, that natural selection
was another awful stretcher he believed in. Orr's remark about Chomsky makes
it appear that he thinks Darwin was dismissing sexual selection when he called
it an awful stretcher, but I suspect he knows better and simply got carried
away by his desire to rub it in about my nonexistent error.
These are the only two "factual errors" he comes
up with, and he judiciously grants them the status of "peccadilloes."
There are some factual errors in my book that others have pointed out, but
since Orr does not brandish any of them in his essay, I think we can conclude
that they eluded his scrutiny. Let's turn, then, to his main concern: my book's
alleged deep misunderstandings of biology.
Orr concentrates on my criticism of Gould and Lewontin's famous
attack on adaptationism, "The spandrels of San Marco and the Panglossian
paradigm." Evolution-what Darwin called "descent with modification"-is
an uncontroversial fact, as secure as any in science, but there is still controversy
among biologists about just how many facts about evolution are explained by
natural selection, the mechanism Darwin proposed as the most important force
shaping evolution. To see what the issue is, consider the popular retort of
software engineers when somebody finds a flaw in their program: "It isn't
a bug, it's a feature." In other words, it's supposed to be that way;
it was designed to work like that. Now is everything observable in the biosphere
a "feature," an "adaptation," as an evolutionary biologist
would call it? Are there no bugs, no undesigned bits, no historical accidents?
Of course there are. Everybody agrees on that. Everybody does not agree on
how important these non-adaptations are; the greater the role you give to
natural selection, the more "adaptationist" you are, and the Gould/Lewontin
essay was an attempt to swing opinion the other way.
Orr begins by usefully perpetuating some of the misunderstandings
I exposed in my discussion of Gould and Lewontin's central architectural metaphor
for a non-feature: a "spandrel"-one of the curving wedges of masonry
that serve as the transition from the domes of the Basilica di San Marco in
Venice to the arches that hold them up. (I find this useful because some critics
have wondered if any biologists have actually been confused on this score;
my answer is yes-see, e.g., H. Allen Orr.) This is what he says:
Although spandrels are often decked out with mosaics,
no one would seriously argue that spandrels are there because they provide
such swell surfaces for mosaics. Instead, spandrels are there because they
have to be-they are, it turns out, the inevitable by-product of putting a
dome on rounded arches.
"Inevitable by-product"? This is either flat false
or true but irrelevant. If by "spandrel" Gould and Lewontin mean
the particular structure used in San Marco (properly called a pendentive)
then what they say is false; pendentives are one of many options, but they
are probably the optimal engineering solution to the problem of supporting
the dome-what I would call a Forced Move. In this sense, spandrels (pendentives)
are adaptations par excellence. If, on the other hand, Gould and Lewontin
mean by "spandrel" just "whatever you put in that place between
the dome and the arches," then spandrels are trivially inevitable-you
have to put something there. Gould himself has recently1
opted for this reading, but as I had already pointed out in my book, pp. 272-3,
this interpretation also disqualifies spandrels for their role as lead metaphor
in Gould and Lewontin's biological argument. In this sense, architectural
constraints present a problem, not a solution, in biology as much as in building,
and that is where natural selection comes in.
This has been much discussed recently, and the architectural
engineer Robert Mark, in a fascinating article in the July issue of American
Scientist, shows that my own amateur engineering (and architectural history)
led me to underestimate the structural requirements of putting a large dome
on arches. The pendentives are in fact not just aesthetic adaptations, as
I had claimed, but also structural adaptations. So the San Marco "spandrels"
are a doubly poor choice for eponymous non-adaptations. The mosaics on the
pendentives are aesthetic design choices enabled by prior structural design
choices. An example of one adaptation exploiting a prior adaptation hardly
strikes a telling blow against adaptationism.
But why are we quibbling about these fine points of architectural
history? Why, indeed, do I pay such obsessive attention to Gould and Lewontin's
attack on adaptationism? Orr expresses his bafflement:
Why, after all, should a man hoping to export Darwinism
from biology be so obsessed with defending the minutiae of adaptive storytelling
within biology? I would not think the legitimacy of, say, cultural Darwinism
hangs critically on whether selection is very important in biological evolution
(as everyone, including Gould and Lewontin, believe) or is almost exclusively
important (as Dennett believes).
First, Orr strains to create a gulf that does not exist between
me and the run of biologists. What, we may ask, is the difference between
"very important" and "almost exclusively important"-the
presumably dissenting view he attributes to me? It is Gould and Lewontin,
in fact, who have had trouble clarifying their charge. Everybody agrees, as
Orr says, that selection is very important, so just what are Gould and Lewontin
saying? That it is not as important as some people think? If that had been
their message, I would agree with John Maynard Smith that their paper was
"on the whole, welcome." Orr quotes this verdict with approval,
and so did I (p. 278), noting, however, that this was not their message-or
at least it has not been Gould's version of their message.
I think Orr's professed perplexity must be disingenuous. I
make it crystal clear why I have to go to all this trouble clarifying these
minutiae: Gould has persistently misrepresented the import of the Gould/Lewontin
paper outside biology, and many have been taken in. My task was to show the
non- biologists that they have been seriously misled by Gould about this.
In my book I list four propositions that are widely believed by non-biologists
to have been demonstrated by Gould. The first two are relevant here:
If you believe: 1) that adaptationism has been refuted
or relegated to a minor role in evolutionary biology, or (2) that since adaptationism
is 'the central intellectual flaw of sociobiology' (Gould, 1993a, p. 319),
sociobiology has been utterly discredited as a scientific discipline . . .
then what you believe is a falsehood. (p. 265)
Well, are these truths or falsehoods? They are widely believed.
Many non-biologists are under the weird misapprehension, thanks to Gould's
rhetoric, that one is under no obligation to provide an adaptive account of
the evolution of a complex competence or organ-such as human language. In
some misguided quarters, indeed, adaptationist explanations of anything are
automatically suspect! For more than a year before the publication of my book
and on several occasions since then, I have repeatedly requested that he clarify
his position on these propositions. Steve Gould is undeniably a Great Communicator.
If these propositions are not what he meant, if over-eager readers have misunderstood
him, he should find it both obligatory and easy to correct these widespread
misapprehensions. If he meant them, he should either defend them against my
charge that they are false, or concede that he has misled his readers. He
has not accepted my invitation to clarify his position, so it falls to me
to explain to the world, at whatever length it takes, why these are not the
take-home messages from Gould and Lewontin's article.
The centerpiece of Orr's critique of my defense of adaptationism
is his feigned astonishment that I omit any discussion of Motoo Kimura's neutral
theory, "the most serious and famous of all challenges to selectionist
story-telling." Here he plays a cute rhetorical trick: Boy oh boy, Dennett
must be really out of it to have overlooked Kimura's work! Why on earth did
he omit this? Was it because he never heard of it? Naw. Or because Kimura's
book was too difficult for him to read? Naw. Or because he "doesn't want
to let the cat out of the bag" about the fact that biologists have found
out that "non-Darwinian" evolution is common? Naw. Having relieved
himself of these snide but indefensible suggestions-as he himself so generously
acknowledges-he arrives at the obvious truth: "Dennett is interested
in Design . . . and the neutralist/selectionist controversy has nothing to
say here." Right. It wasn't relevant. Kimura's theory is about typographical
change, visible at the molecular level of the genome. As Orr says, "most
evolution at the molecular level [my emphasis] is not caused by natural selection,
but by 'genetic drift.'" Before Kimura, theorists hugely underestimated
the possible role of the accumulation of random mutations that are (as Orr
notes) functionally equivalent. Kimura's work is wonderful, surprising, and
still controversial; it claims (among other things) that the diversity observed
at the molecular level could be much more the result of random genetic drift
than many theorists-"selectionists"-had thought, but this aspect
of selectionism has little or nothing to do with adaptationism, which is why
I left it out. Oh, I could have had some fun with the fact that there are
indeed some biologists- Gould and Lewontin come to mind-who have often suggested
to the lay public that Kimura's neutral theory is a rival to natural selection
as an account of morphologic evolution (the shaping of complex phenotypic
features), but as Orr as much as admits, this is not taken seriously among
From all Orr's hooting one might get the impression, by the
way, that I had nothing to say on these topics in my book, but in fact, I
discuss the role of random genetic drift (pp. 125-6) and the ubiquitous possibility
of typographically different but functionally equivalent genetic recipes for
proteins (pp. 195-6, p. 287). Where Orr mentions functionally irrelevant differences
in varieties of hemoglobin, I mention functionally irrelevant differences
in lysozyme. Same points made, right down the line. In fact I also make all
the other points about adaptation he goes on to explain for his lay readers,
saying "Dennett never confronts these legitimate worries." (All
of them? Yes. If you want to test my claim, you can go back and match page
30 of Orr's review with my pages 199, 248-249; I'll spare readers the tedious
One last point about Orr's understanding of central concepts
in his own field before we turn briefly to the evolution of culture. I say
that natural selection is an algorithmic process, and hence is "substrate
neutral"-it doesn't matter what material the algorithm is executed in,
so long as the recipe is followed. Orr says, correctly:
This substrate neutrality argument is supremely important
to Dennett. It-and nothing else-explains why selection can be lifted from
its historical base in biology. It is what makes Darwinism so dangerous.
He goes on: "But Dennett slips here." It is Orr who
has slipped, falling flat on his face with his proposed counterexample to
my claim that natural selection is a substrate neutral process: "blending
Evolution would quickly grind to a halt, for instance,
if inheritance were blending, not particulate. With blending inheritance,
the genetic material from two parents seamlessly blends together like different
colored paints. With particulate Mendelian inheritance, genes from Mom and
Dad remain forever distinct in Junior. This substrate problem [sic] was so
acute that turn-of-the-century biologists-all fans of blending inheritance-concluded
that Darwinism just can't work. Modern evolutionary genetics was born in 1930
when Sir Ronald Fisher cracked this problem: Population genetics shows that
particulate Mendelian inheritance saves the day. It is just the kind of substrate
[sic] needed for evolution by natural selection to work.
Orr is right that Mendel saved the day for Darwin (a point
of history I discuss in somewhat greater detail in my book), but this was
a triumph for substrate neutrality! Mendel-and Fisher, too, for that matter-was
clueless about what the substrate was. (It's DNA, of course, but that wasn't
discovered until 1953, by Crick and Watson.) Mendelian genes are a paradigm
case of substrate neutrality; they are pure data structures, whose material
composition is irrelevant just so long as they obey the combinatorial rules
Mendel laid down. The field of population genetics thrives to this day in
almost complete independence of any concern with the nitty-gritty biochemistry
of the substrate- just the way software engineering is conducted by people
who need know little about the electronics or physics of their silicon substrate.
The problem with blending inheritance is the code, not the substrate; it would
be just as hopeless a system to use in evolutionary software engineering as
it would be in carbon-based biological evolution.
Orr has apparently missed this fundamental point about the
abstract nature of evolutionary theory, and having missed it, he compounds
his confusion when he turns to my exploitation of Richard Dawkins' concept
of memes in defense of a Darwinian theory of cultural evolution. His "basic
problem" with any attempt at such a theory is that we are "very
ignorant of how humans hold ideas in their heads. . . So how can we possibly
conclude that the process 'must be' Darwinian?" Well, population geneticists
were very ignorant of how organisms held genes in their bodies until recently,
but this did not stop Fisher and his colleagues from establishing (in a substrate
neutral way) the soundness of the modern synthesis, did it? Orr comes close
to contradicting himself here, and doesn't notice.
But then his whole discussion of memes is inattentive to detail.
Orr says: "The fitness of memes is strangely tautological. While we can
often point to ecological reasons why certain genes are fitter than others,
a meme is deemed 'fit' only because it is common." This is obviously
false, belied by many of the examples discussed in my book. He goes on: "'Elvis
is alive' is certainly a fit meme, but it is neither true nor helpful."
Has he been paying attention at all? This independence of meme fitness from
truth and utility (to us) is the single most important feature of the meme's-eye
view of cultural evolution. He continues: "Last, Dennett confesses [sic]
that memes often show
a Lamarckian, not Darwinian, style of evolution, in that acquired traits get
passed on," ignoring the fact that I explain why the "charge"
that cultural evolution is Lamarckian misses the mark entirely (p. 355).
Given this obliviousness, his "astonishment" that
I would point out how difficult-and maybe impossible-it will be to establish
a predictively powerful science of memetics counts for little or nothing.
There are many uncontroversially evolutionary phenomena in biology that also
defy formal scientific inquiry by being either computationally intractable
or bereft of accessible data. Ancient memes, like ancient genes for soft body
parts, leave almost no fossil traces. The genetic contributions to many complex
abilities in many species are both practically impossible to research and
utterly indubitable. For instance, does Orr doubt that there is an evolutionary
account-probably forever inaccessible, alas-of the nest-building behaviors
in birds? He interprets my cautious discussion of the prospects for memetics
as "backing off" from stronger claims about memes, but this retreat
is a figment of his imagination. (Aficionados will recognize this trick as
an instance of the Gould two-step, which I exposed several years ago in the
New York Review of Books.)
Orr's other criticisms of memes follow the same pattern as
his earlier criticisms of my biology. Once again, he reiterates points I make
myself about important problems to be solved, but he treats these problems
as clearly insoluble, without giving any arguments or acknowledging that he
is following me. He claims that I have overlooked-or deliberately suppressed-trenchant
criticisms of memetics from within biology, but he doesn't mention any. I
deal with all the criticisms that I have encountered in the literature. If
he knows of others, he should have referred to them specifically.
One final point: Orr's failure to understand substrate neutrality
leads him to miss the otherwise obvious point that in any theoretical analysis
of the evolution of a cultural phenomenon-say, morality-there can be complex
interplay between genetic and memetic sources of design; substrate neutrality
means that the work of evolution can be accomplished in either medium (at
different rates, under the same or different selection pressures) and combined
in the final product. He apparently thinks that any account of the evolution
of morality must either be all genetic or all cultural. He asks, at one point:
"Is it obvious that genetic changes are required for such a thing? [No.
It is false.] Where are Dennett's trusty memes when we need them? [Right under
his nose.]" Orr might want to look at some of the literature I cited
on these topics, such as the recent theoretical work by Philip Kitcher, an
alumnus of Lewontin's lab. It's substrate neutral, all the way.
I will not comment on Orr's critical treatment of the more
purely philosophical parts of my book, beyond noting that the charitable interpretation
would be that he doesn't understand the difference between serious criticism
and the mug's game of quoting out of context. (Those who are curious may consider
as Exhibit A his abuse of my discussion of "bait-and-switch" in
One does not lightly undertake the task of dislodging heroes
from their pedestals so that their ideas can be critically assessed in the
same arena with the ideas of ordinary mortals. So I expected to be treated
fairly roughly by their fans, especially in their home town. Gould and Lewontin
and Chomsky have so far all chosen to leave the counter-attack to others,
my criticisms being too far beneath their notice, one gathers, to merit any
detailed public response. The attacks I have seen to date-of which Orr's is
the best, by the way-have been long on sneering and short on substance. It's
been surprisingly easy to take, since my task has been far from thankless.
Indeed, the thanks I have been receiving from biologists around the world
has been most gratifying.
1 After an address at Dartmouth, in April 1996, in response
to a question from me.