Structural Constraints, Spandrels, and Exaptation 1235
arguments of those who found the concept intelligible, even interesting, but doubted
that sufficient information could generally be obtained about the history of evidently
functional features to make a proper distinction based upon inferences about prior
states. Arnold argues that several advances, particularly the codification of cladistic
techniques for phyletic ordering, have made the distinction operational in a
sufficiently high percentage of cases: the formerly broad definition of "adaptation for
all useful traits whatever their origin was reasonable at a time when it was difficult to
find out how advantageous traits had arisen, unless this had been observed in recent
populations. However, phylogeny reconstruction now allows many individual
exaptations to be recognized with some certainty, and makes distinction of exaptive
and adaptive origin of performance advantage appropriate" (p. 126).
The relative timings for the origin of a form and for the inception of its current
function—as inferred either from the branching points of a cladistic analysis, or from
direct knowledge of historical sequences—provide the main criteria for distinction of
exaptation from adaptation. "For adaptation," Arnold writes (p. 132), "a hypothesis is
refuted if the new trait develops before the relevant selective regime. If the test is
passed, it is possible to check whether the new trait really confers an advantage in the
new regime that its plesiomorphic state does not."
Arnold then asks how we should interpret the opposite phenomenon "in which a
derived trait and a regime in which it gives a performance advantage first appear
concurrently on the same node on a lineage" (p. 133). This situation of coincidence in
cladogeny between form and function would seem to point to adaptation, but Arnold
notes (p. 133) that an imperfect record must fail to provide evidence for several
(probably most) events of speciation, and that traits may arise before selective
regimes at these missing nodes, and then be compressed into coincidence with a
current selective basis at the first recorded node of their joint occurrence. I accept this
point as evidently valid, but favorable for tests of the importance of exaptation. As
emphasized in other contexts within this book, ineradicable biases in testing a
hypothesis present little problem, and may even constitute a blessing in disguise,
when their direction works against the hypothesis under test—because the hypothesis
gains stronger affirmation by success in the face of such unfavorable odds. Since
missing nodes must, by Arnold's argument, induce an underestimate for the frequency
of exaptations by redirecting some genuine cases into the opposite category of
adaptations (sensu stricto), this bias does not pose problems for tests of exaptation.
For example, and using these twinned criteria, Arnold found 70 percent of 61
apomorphies in the lacertid lizard genus Meroles to be "concurrent with occupation
of the environmental situation" (p. 133) of their present function—therefore requiring
that they be ranked as adaptations.
Cases of multiple utility for a single feature offer special promise for resolution
by these criteria of temporal or cladogenetic sequencing. Arnold cites the case of "the
aberrant arboreal tropical African lacertid lizard, Holaspis guentheri," whose
extremely flattened head "allows it to hunt and hide in