1172 THE STRUCTURE OF EVOLUTIONARY THEORY
sharp (Burke et al. refer to the shifts of these boundaries to emplace more or fewer
vertebrae within any given region, as "transpositions"). In their most intriguing
conclusion, Burke et al. (1995) found that some of these phenotypic transitions
correlate precisely with anterior expression boundaries of particular Hox genes. For
example, Hoxc- 6 marks the transition between cervical and thoracic vertebrae,
despite the highly variable number of cervical vertebrae, ranging from 3 or 4 in frogs,
to 7 in mice (as in virtually all mammals, including giraffes), to 17 in geese. The
thoracic-lumbar transition generally correlates with the expression of Hoxa-9, Hoxb-
9, and Hoxc-9, whereas the Hoxd- 9 boundary tends to be shifted backwards to the
lumbosacral transition. Carroll (1995, p. 483) comments on these differences in the
ninth paralogy group: "This may be significant because the thoracic-lumbar
distinction is not general among tetrapods. It may be that shifts within the Hox- 9
group were important in the evolution of this transition from a more uniform trunk,
perhaps even in the evolution of the tetrapods from fish."
These regularities of Hox regionalization may help us to understand both the
limitations and flexibilities of vertebrate anatomy in terms of historical constraint. In
an early article, for example, Tabin (1992) suggested that tetra-pod limbs may now be
constrained to five digits per limb (despite the presence of up to 8 digits in the earliest
tetrapods of the Late Devonian Period— see Coates and Clack, 1990; and Gould,
1993e) because the Hoxd series that plays such a major role in patterning limbs may
now only generate five "addresses" for the development of distinct digits. Many
polydactylous mutants (and experimental manipulations) exist in vertebrates, but the
supernumeraries are always phenotypic replicates of one of the five distinct digits, so
the general hypothesis holds (see also Shubin et al., 1997, pp. 642-643).
A related classical question asks why tetrapods, honoring their name, never
grow more than four limbs, whereas the other major terrestrial group of arthropods
usually evolves phenotypes with more appendages (even though we might imagine,
on functional grounds, that an increased number of supports would be even more
valuable in large vertebrates with much lower area to volume ratios—for supporting
strength of bone scales as cross-sectional area). Coates and Cohn (1998, p. 379) note
that "the nearest approach to a third pair of lateral appendages may be the lateral
caudal keels of certain fishes, such as tuna and various sharks." But a true anatomical
third pair has never evolved in any tetrapod or extant fish. (The extinct acanthodian
fishes evolved the only vertebrate departure from the principle of two primary limb
pairs.)
But as Hox rules constrain, they can also be tweaked to win interesting
flexibility. Cohn and Tickle (1999), for example, studied Hox expression in the axial
skeleton of pythons, which can grow more than 300 essentially identical vertebrae,
and which retain hindlimb rudiments but express no forelimb development at all.
Except for the atlas, every vertebra anterior to the rudimentary hindlimb develops ribs
(a thoracic feature) as well as ventral hypopophyses (a cervical feature), suggesting to
Cohn and Tickle (1999, p. 474) that "information encoding thoracic identity may
have extended into