Historical Constraints and the Evolution of Development 1035
than down other conceivable paths. "The meaning of the word would [then] vanish,"
Stearns notes (1986, p. 35). He therefore recommends: "We can preserve it in a
relative sense if we recognize that it only has meaning in a local context where one
concentrates on the possibilities latent in certain processes and views the limitations
on those possibilities as arising from outside that context." Therefore, in considering
revisions and expansions of Darwinian theory, ordinary natural selection becomes the
context and any force (like internally channeled variation) limiting its exclusive sway
in directing evolutionary change, becomes a constraint.
Antonovics and van Tienderen (1991), in an influential article that cleared away
much of the accumulating nonsense in definitional debates about constraint in
evolution, also favored this relative concept as a solution. They agreed with my
argument (Gould, 1989a) that "it is those factors that influence the process but are
external to the favored theory that should be termed constraints" (Antonovics and van
Tienderen, 1991, p. 167). But, choosing a terminology that struck them as more
consistent with the ethos of scientific neutrality, they preferred the term "null model"
to my "favored theory." (I would reply that we do not usually refer to strong
theories—like natural selection—based on particular and well-articulated causes, as
"null models." I would also argue that nothing negative attends the admission that
disciplines operate under favored theories—a "good thing" for science, so long as we
retain flexibility for change and do not equate "favored" with "established"; and,
especially, so long as we treat the status of "favored" as an impetus for challenge
rather than passive acquiescence, as we manifestly do when we invoke constraints to
rebut overly strict versions of natural selection.)
In any case, Antonovics and van Tienderen survey the literature and find, in
support of the argument developed here, that "the overall null model used by most
authors was one of evolution by natural selection (irrespective of the level of
selection)" (p. 167), and that nearly all explicit claims for "constraints" upon change
within populations and lineages "dealt with evolutionary constraints to adaptation by
natural selection" (p. 166). They also noted the "odd" feature of relative definitions
that strikes many scientists as paradoxical, but would not be so regarded if we
accepted the honorable and inevitable principle, so familiar to philosophers of science
and language, but still faced with discomfort by many scientific professionals, that all
terminology must be "theory bound"—specifically, in this case, that orthodox results
of one theory become constraints in other theories. They write (p. 167): "Given
evolution by random drift as a null model, natural selection now becomes a
constraint!" Yes, and appropriately so—with no exclamation point needed to register
surprise.
Although I disagree with his particular recommendation, Eble (1999) published
a thoughtful and conceptually innovative paper rooted in this important principle of
the inevitability and appropriateness of theory-bound terminology. Eble notes, and
brilliantly analyzes, two entirely distinct, but all too frequently conflated, meanings
of "chance" and "randomness" in evolutionary theory. His article, entitled "On the
dual nature of chance in evolutionary