28
What, If Anything,
Is a Zebra?
REPRINTED W I T H PERMISSION OF:
W.W. Norton & Co., Inc.,
Hen's Teeth And Horse's Toes: Further Reflections 1
"Ch 28: What, If Anything, Is A Zebra?"
by S. Gould
©1QJH W W Norton <% To . lnc .
EACH
Pes. T55-165
Y E A R , professional scientists scan
thousands o f titles, read hundreds o f abstracts, and study a
f e w papers in depth. Since titles are the c o m m o n e s t , and
usually the only, f o r m o f contact between writers and potential readers in the great glut o f scientific literature, catchy
items are appreciated and r e m e m b e r e d , but unfortunately
rare. Every scientist has his favorite title. M i n e was coined
by paleontologist A l b e r t E. W o o d in 1957: " W h a t , I f Anything, Is a R a b b i t ? "
W o o d ' s question may have b e e n wry, but his conclusion
was ringing: rabbits and their relatives f o r m a coherent,
well-defined o r d e r o f mammals, not particularly close to
rodents in evolutionary descent. I was r e m i n d e d o f W o o d ' s
title recently when I read a serious challenge to the integrity
o f a personal favorite a m o n g mammals: the zebra. N o w don't
get t o o agitated. I am not trying to turn the w o r l d o f received
opinion upside d o w n . Striped horses manifestly exist. But
d o they f o r m a true evolutionary unit? W i t h "Stripes D o N o t
a Zebra M a k e " — a quite respectable title in its own r i g h t —
Debra K . Bennett has f o r c e d us to extend W o o d ' s question
to another g r o u p o f mammals. What, if anything, is a zebra?
Since evolutionary descent is our criterion f o r biological
o r d e r i n g , w e d e f i n e groups o f animals by their g e n e a l o g y .
W e d o not j o i n together t w o distantly related groups because their m e m b e r s have independently e v o l v e d s o m e similar features. Humans and bottle-nosed dolphins, f o r exam3 5 5
356
|HEN'S
TEETH
AND
HORSE'S
TOES
pie, share the pinnacle o f brain size a m o n g mammals. But
w e d o not, f o r this reason, establish the taxonomic group
Psychozoa to house both s p e c i e s — f o r dolphins are m o r e
closely related by descent with whales, and humans with
apes. W e f o l l o w the same principles in our own genealogies. A boy with D o w n ' s s y n d r o m e is still his parents' son
and not, by reason o f his affliction, m o r e closely related to
other D o w n ' s children, n o matter h o w l o n g the list o f similar features.
T h e potential dilemma f o r zebras is simply stated: they
exist as three species, all with black-and-white stripes to be
sure, but d i f f e r i n g notably in both numbers o f stripes and
their patterns. ( A fourth species, the quagga, became extinct early in this century; it f o r m e d stripes only o n its neck
and forequarters.) T h e s e three species are all members o f
the genus Equus, as are true horses, asses, and donkeys. (In
this essay, I use " h o r s e " in the generic sense to specify all
m e m b e r s o f Equus,
including asses and zebras. W h e n I
mean O l d D o b b i n
o r M a n o ' W a r , I will write
"true
h o r s e s . " ) T h e integrity o f zebras then hinges o n the answer
to a single question: D o these three species f o r m a single
evolutionary unit? D o they share a c o m m o n ancestor that
gave rise to them a l o n e and to n o other species o f horse?
O r are s o m e zebras m o r e closely related by descent to true
horses o r to asses than they are to other zebras? I f this
second possibility is an actuality, as Bennett suggests, then
horses with black-and-white stripes arose m o r e than once
within the genus Equus, and there is, in an important evolutionary sense, n o such thing as a zebra.
But h o w can w e tell, since n o o n e witnessed the origin o f
zebra species ( o r at least australopithecines weren't taking
notes at the time), and the fossil record is, in this case, too
inadequate to identify events at so fine a scale. During the
past twenty years, a set o f procedures has been codified
within the science o f systematics f o r resolving issues o f this
kind. T h e m e t h o d , called cladistics, is a formalization of
procedures that g o o d taxonomists f o l l o w e d intuitively but
did not p r o p e r l y express in words, leading to endless quibb l i n g and fuzziness o f concepts. A clade is a branch on an
WHAT,
IF
ANYTHING,
IS
A
ZEBRA?
| S57
evolutionary tree, and cladistics attempts to establish the
pattern o f branching f o r a set o f related species.
Cladistics has generated a fearful j a r g o n , and many o f its
leading exponents in A m e r i c a are a m o n g the most contentious scientists I have e v e r encountered. But behind the
names and nastiness lies an important set o f principles. Still
the clear formulation o f principles d o e s not guarantee an
unambiguous application in each case—as w e shall see f o r
our zebras.
I believe that w e can get by with just two terms f r o m the
bounty o f f e r e d by cladists. T w o lineages sharing a c o m m o n
ancestor f r o m which no other lineage has sprung f o r m a
sister group. My b r o t h e r and I f o r m a sister g r o u p (pardon the
confusion o f g e n d e r ) because he is my only sib and neither
o f my parents had any other children.
Cladists attempt to construct hierarchies o f sister groups
in o r d e r to specify temporal o r d e r o f branching in evolutionary history. F o r e x a m p l e : gorillas and chimpanzees f o r m a
sister g r o u p because n o other primate species branched
f r o m their c o m m o n ancestor. W e may then take the chimpgorilla sister g r o u p as a unit and ask which primate forms a
sister g r o u p with it. T h e answer, according to most experts,
is us. W e n o w have a sister g r o u p with three species, each
m o r e closely related to its two partners than to any other
species.
W e may extend this process indefinitely to f o r m a chart o f
Orang
Human
Gorilla
T h e cladistic pattern of great apes and humans,
NATURAL HISTORY. DRAWING BY JOE LE MONNIER.
Chimp
REPRINTED FROM
358
|HEN'S
TEETH
A^D
HORSE'S
TOES
branching relationships called a cladogram. But consider just o n e m o r e step: W h a t primate species is the sister
g r o u p to the human-chimp-gorilla unit? Conventional wisd o m cites the orangutan, and w e add it to our cladogram.
T h i s cladogram o f " h i g h e r " primates contains an interesting implication: there is n o such thing as an ape, at least as
usually defined. Several species o f primates may swing
through trees, eat bananas in zoos, and f o r m g o o d p r o t o types f o r science fiction o f various sorts. But orangs, chimps,
and gorillas (the " a p e s " o f our vernacular) are not a genealogical unit because orangs are cladistically m o r e distant
f r o m chimps and gorillas than humans a r e — a n d w e originally d e f i n e d the term ape to contrast s o m e lesser forms with
o u r exalted state, not to include us!
T h e zebra p r o b l e m can also b e placed in this context. I f
the three species o f zebras f o r m a sister g r o u p (as humans,
chimps, and gorillas d o o n our c l a d o g r a m ) , then each is
m o r e closely related to its two partners than to any species
o f horse, and zebras f o r m a true evolutionary unit. But if
zebras are like " a p e s , " and another species o f horse lies
within the cladogram o f zebras (as humans lie within the
cladogram o f traditional apes), then striped horses may
share s o m e striking similarities meriting a c o m m o n vernacular term (like zebra), but they are n o t a genealogical unit.
But h o w d o we identify sister groups correcdy? Cladists
argue that w e must search f o r — a n d here comes the second
term—shared derived characters (technically called synapom o r p h i e s ) . Primitive characters are features present in a
distant c o m m o n ancestor; they may b e lost o r m o d i f i e d
independently in several subsequent lineages. W e must be
careful to avoid primitive characters in searching f o r comm o n features to identify sister groups, f o r they spell nothing
but trouble and error. Humans and many salamanders have
five toes; horses have o n e . W e may n o t t h e r e f o r e state that
humans are m o r e closely related to salamanders than to
horses, and that the concept o f " m a m m a l " is therefore a
fiction. Rather, five toes is an inadmissible primitive character. T h e c o m m o n ancestor o f all terrestrial vertebrates had
five toes. Salamanders and humans have retained the origi-
WHAT,
IF
ANYTHING,
IS
A
ZEBRA?
|
S57
nal number. H o r s e s — a n d whales and cows and snakes and
a host o f other vertebrates—have lost s o m e o r all o f their
toes.
D e r i v e d characters, o n the other hand, are features present only in m e m b e r s o f an immediate lineage. T h e y are
unique and newly e v o l v e d . A l l mammals, f o r example, have
hair; n o other vertebrate does. H a i r is a d e r i v e d character
f o r the class Mammalia because it e v o l v e d but o n c e in the
c o m m o n ancestor o f mammals and t h e r e f o r e identifies a
true branch o n the family tree o f vertebrates.
Shared
derived characters are held in c o m m o n by two o r m o r e
lineages and may b e used to specify sister groups. I f we wish
to identify the sister g r o u p a m o n g tunas, seals, and bobcats,
we may use hair as a shared d e r i v e d character to unite the
two mammals and to eliminate the fish.
For zebras, the question then b e c o m e s : A r e stripes a
shared d e r i v e d character o f the three species? I f so, the
species f o r m a sister g r o u p and zebras are a genealogical
unit. I f not, as Bennett argues, then zebras are a disparate
g r o u p o f horses with s o m e confusing similarities.
T h e m e t h o d o f cladistics is both simple and sensible:
establish sequences o f sister g r o u p s by identifying shared
derived
characters.
Unfortunately,
conceptual
elegance
d o e s not guarantee ease o f application. T h e rub, in this
case, lies in d e t e r m i n i n g just what is o r is not a shared
derived character. W e have s o m e r o u g h guidelines, and
s o m e seat-of-the-pants feelings, but n o unerring formulas.
I f d e r i v e d characters are sufficiently " c o m p l e x , " f o r example, w e b e g i n to f e e l confident that they could not have
e v o l v e d independently in separate lineages and that their
mutual presence t h e r e f o r e indicates c o m m o n descent.
Chimps and gorillas share a set o f c o m p l e x and apparently independent modifications in several o f their c h r o m o somes (mostly " i n v e r s i o n s , " literally, the turning around o f
part o f a c h r o m o s o m e by breaking, flipping, and reattaching). Since these c h r o m o s o m a l changes are c o m p l e x and d o
not seem to represent " e a s y " modifications so adaptively
necessary that separate lineages might e v o l v e them independently, we mark them as shared d e r i v e d characters pre-
360
I HEN'S
TEETH
AND
HORSE'S
//
TOES
sent in the common ancestor of chimps and gorillas, and in
no other primate. H e n c e they identify chimps and gorillas
as a sister group.
Unfortunately, most derived characters are more ambiguous. T h e y tend to be either easy to construct or else so
advantageous that several lineages might evolve them independently by natural selection. Many mammals, for example, develop a sagittal crest—a ridge o f bone running along
the top o f the skull from front to back and serving as an
attachment site f o r muscles. Most primates d o not have a
sagittal crest, in part because large brains make the cranium
bulge and leave neither r o o m nor material for such a structure. But a general rule for scaling of the brain in mammals
holds that large animals have relatively smaller brains than
relatives o f diminished body size (see essays in Ever Since
Darwin and The Panda's Thumb). Thus, the largest primates
have a sagittal crest because their relatively small brains do
not impede its formation. (This argument does not apply
to the great oddball Homo sapiens, with an enormous brain
despite its large body.) T h e largest australopithecine, Australopithecus boisei, has a pronounced sagittal crest, while
smaller members of the same genus d o not. Gorillas also
have a sagittal crest, while most smaller primates do not. W e
would make a great error if, using the sagittal crest as a
shared derived character, we united an australopithecine
with a gorilla in a sister group and linked other, smallerbodied australopithecines with marmosets, gibbons, and
rhesus monkeys. T h e sagittal crest is a " s i m p l e " character,
probably part o f the potential developmental repertoire for
any primate. It comes and goes in evolution, and its mutual
presence does not indicate common descent.
Bennett bases her cladistic analysis of the genus Equus on
skeletal characters, primarily o f the skull. All horses are
pretty much alike under the skin, and Bennett has not found
any shared derived characters as convincing as the chromosomal similarities o f chimps and gorillas. Most of her
characters are, by her own admission, more like the sagittal
crest—hence the provisional nature o f her conclusions.
Bennett argues that the genus Equus contains two major
WHAT,
IF
ANYTHING,
IS
A
ZEBRA?
|
S57
cladistic g r o u p s — d o n k e y s and asses on o n e side and true
horses and zebras o n the other. T h u s , zebras pass the first
test f o r consideration as a genealogical unit. Unfortunately
(or not, according to your point o f v i e w ) , Bennett claims
that they fail the second test. She does identify the Burchell
and Grevy zebras (Equus burchelli and E. grevyi) as a sister
group. But in her scheme, the third species, the mountain,
or Hartmann, zebra (E. zebra) d o e s not j o i n its cousins to
f o r m a larger sister group. Instead, the sister species o f the
mountain zebra is our close compatriot in farm and track,
the true horse (E. caballus)\ T h u s , mountain zebras j o i n with
true horses b e f o r e they connect with other zebras. O l d D o b bin is inextricably intercalated within the cladogram o f zebras—and since he is not a zebra by any definition, then
what, if anything, is a zebra?
But Bennett's analysis is based upon only three characters,
n o n e very secure. All are potentially simple modifications o f
shape or p r o p o r t i o n , not presences or absences o f c o m p l e x
structures. All, like the sagittal crest, could c o m e and go.
O n l y one potential shared derived character unites true
horses with E. zebra: the " o r i e n t a t i o n o f postorbital bars
relative to horizontal p l a n e " (a relatively less slanted position f o r a bar o f b o n e located o n the skull behind the e y e s —
not exactly the stuff o f which confident conclusions are
made). Only two potential shared derived characters unite
362
|HEN'S
TEETH
A^D
HORSE'S
TOES
Burchell and G r e v y zebras: the presence o f frontal d o m i n g
(inflation o f the t o p part o f the skull) and relative skull
breadth (these two zebras have l o n g and narrow snouts).
Unfortunately, w e know that at least o n e o f these characters
d o e s n ' t work well f o r Bennett's cladistic scheme because she
admits that a m e m b e r o f her other l i n e a g e — a horse with the
peculiar moniker o f the Asiatic half-ass (E.
hemionus)—has
independently e v o l v e d a l o n g , narrow snout. I f twice, why
not three times?
W h e n w e look f o r c o r r o b o r a t i o n to an obvious s o u r c e —
numbers o f c h r o m o s o m e s — w e are again disappointed. A s I
discussed in essay 26, the various species o f horses, despite
their marked similarities o f f o r m , d i f f e r greatly in number o f
c h r o m o s o m e s . Fusion o r fission o f c h r o m o s o m e s may b e a
m a j o r mechanism o f speciation in mammals, and these differences may t h e r e f o r e have great evolutionary significance.
A l l zebras, and only zebras, have f e w e r than fifty pairs o f
c h r o m o s o m e s (thirty-two in Hartmann's zebra to forty-six in
G r e v y ' s zebra). A l l other horses have m o r e than fifty ( f r o m
fifty-six in Equus hemionus to sixty-six in Przewalski's horse).
T h e l o w number o f zebras might mark them as a genealogical g r o u p if the character is shared derived, and not either
primitive o r e v o l v e d m o r e than once. Bennett's hypothesis
may still b e maintained by arguing either that small numbers
are primitive f o r all horses and that asses and true horses
acquired
larger
numbers
by
independent
evolutionary
routes; o r that different lineages o f zebras e v o l v e d small
numbers a l o n g separate evolutionary paths. Still, since we
have n o reason to associate stripes with small numbers of
c h r o m o s o m e s , their c o n j o i n e d presence in all zebras might
best b e interpreted as a sign o f g e n e a l o g y . T h e m o r e complex characters that a g r o u p shares, the m o r e likely that the
g r o u p is genealogical—unless w e have g o o d reason to reg a r d all the characters as primitive (and we d o not in this
case).
I conclude that Bennett's proposal is interesting, but very
much u n p r o v e n . Suppose, h o w e v e r , that she is right. What
then w o u l d a zebra be? O r m o r e specifically, h o w did cladistically unrelated horses get black-and-white stripes? T h e r e
WHAT,
IF
ANYTHING,
IS
A
ZEBRA?
|
S57
are two possibilities. Either the c o m m o n ancestor o f zebras
and true horses had stripes and true horses lost them, while
the three species o f " z e b r a s " passively retained them; or
else, striping is an inherited d e v e l o p m e n t a l capacity o f all
horses and not so c o m p l e x a character as it appears. In this
case, several separate lineages could acquire stripes independently. Zebras would then be horses that have realized
a potential pathway o f d e v e l o p m e n t probably c o m m o n to
many or all m e m b e r s o f the genus Equus (see next essay).
T h i s particular tale o f zebras may not hold, but the radical
messages o f cladistic o r d e r i n g are secure in many cases.
S o m e o f our most c o m m o n and c o m f o r t i n g groups
no
l o n g e r exist if classifications must b e based on cladograms.
W i t h a p o l o g i e s to Mr. W a l t o n and to so many coastal c o m patriots in N e w England, I regret to r e p o r t that there is
surely no such thing as a fish. A b o u t 20,000 species o f
vertebrates have scales and fins and live in water, but they
d o not f o r m a coherent cladistic g r o u p . S o m e — t h e lungfishes and the coelacanth in particular—are genealogically
close to the creatures that crawled out o n land to b e c o m e
amphibians, reptiles, birds, and mammals. In a cladistic
o r d e r i n g o f trout, lungfish, and any bird or mammal, the
lungfish must f o r m a sister g r o u p with the sparrow or elephant, leaving the trout in its stream. T h e characters that
f o r m our vernacular concept o f " f i s h " are all shared primitive and d o not t h e r e f o r e specify cladistic groups.
At this point, many biologists rebel, and rightly I think.
T h e cladogram o f trout, lungfish, and elephant is undoubtedly true as an expression o f branching o r d e r in time. But
must classifications b e based only on cladistic information?
A coelacanth looks like a fish, tastes like a fish, acts like a
fish, and t h e r e f o r e — i n s o m e legitimate sense b e y o n d hidebound tradition—is a fish.
N o debate in evolutionary b i o l o g y has been m o r e intense
during the past d e c a d e than the challenge raised by cladistics
against traditional schemes o f classification. T h e p r o b l e m
arises f r o m the complexity o f the w o r l d , not f r o m the fuzziness o f human thought (although woolliness has made its
usual contribution as well). W e must r e c o g n i z e two rather
364
j HEN'S
TEETH
different c o m p o n e n t s
ALTY)
HORSE'S
to o u r vernacular
TOES
conception
of
" s i m i l a r i t y " b e t w e e n organisms—and classifications are designed to reflect relative d e g r e e s o f similarity. O n
the
o n e hand, w e must consider g e n e a l o g y , o r branching order.
Cladistics works with branching o r d e r alone, rigorously excluding any other notion o f similarity. But what about the
admittedly vague and qualitative, but not therefore unimportant, n o t i o n o f overall similarity in f o r m , function, o r
biological role? T h e coelacanth, to say it again, looks and
acts like a fish even if its closer cladistic relatives are mammals. A n o t h e r theory o f classification, called phenetics—
f r o m a G r e e k w o r d f o r appearance—focuses o n overall
similarity alone and tries to escape the charge o f subjectivity
by insisting that phenetic classifications b e based upon large
suites o f characters, all expressed numerically and processed by c o m p u t e r .
Unfortunately, these two types o f information—branching o r d e r and overall similarity—do not always yield congruent results. T h e cladist rejects overall similarity as a snare
and delusion and works with branching o r d e r alone. T h e
pheneticist attempts to work with overall similarity alone
and tries to measure it in the vain pursuit o f objectivity. T h e
traditional systematist tries to balance both kinds o f information but o f t e n falls into hopeless confusion because they
really d o conflict. Coelacanths are like mammals by branching o r d e r and like trout by biological role. Thus, cladists buy
potential objectivity at the price o f i g n o r i n g biologically
important information. A n d traditionalists curry confusion
and subjectivity by trying to balance two legitimate, but
o f t e n disparate, sources o f information. W h a t is to b e done?
I cannot answer this question, f o r it raises issues o f style,
mores, and m e t h o d o l o g y m o r e than demonstrable substance. But I can at least c o m m e n t o n the source o f this bitter
d e b a t e — a rather simple point that s o m e h o w g o t lost in the
heat. I n an ideal w o r l d , there w o u l d b e n o conflict a m o n g the
three schools—cladistics, phenetics, and traditional—and
all w o u l d p r o d u c e the same classification f o r a given set of
organisms. In this p i p e - d r e a m world, w e would find a perfect
correlation b e t w e e n phenetic similarity and recency o f com-
WHAT,
IF
ANYTHING,
IS
A
ZEBRA?
|
S57
mori ancestry (branching o r d e r ) ; that is, the l o n g e r a g o two
groups o f organisms separated f r o m a c o m m o n ancestor, the
m o r e unlike they w o u l d n o w b e in appearance and biological
role. Cladists w o u l d establish an o r d e r o f branching in time
by cataloging shared d e r i v e d characters. Pheneticists w o u l d
crunch their numerous measures o f similarity in their favorite computers and find the same o r d e r because the most
dissimilar creatures w o u l d have the most ancient c o m m o n
ancestors. Traditionalists, finding c o m p l e t e c o n g r u e n c e between their t w o sources o f i n f o r m a t i o n , w o u l d j o i n
the
chorused harmony o f a g r e e m e n t .
But let the reverie halt. T h e w o r l d is much m o r e interesting than ideal. Phenetic similarity o f t e n correlates very
poorly with recency o f c o m m o n ancestry. O u r ideal w o r l d
requires a constancy o f evolutionary rate in all lineages. But
rates are enormously variable. S o m e lineages change not at
all f o r tens o f millions o f years; others u n d e r g o marked
alterations in a m e r e thousand. W h e n the forebears o f terrestrial vertebrates first split o f f f r o m a c o m m o n ancestry
with coelacanths, they w e r e still unambiguously fish in appearance. But they have e v o l v e d , a l o n g numerous lines during s o m e 250 million years, into f r o g s , dinosaurs,
flamin-
gos, and rhinoceroses. Coelacanths, on the other hand, are
still coelacanths. By branching o r d e r , the m o d e r n coelacanth may b e closer to a rhino than a tuna. But while rhinos,
on a rapidly e v o l v i n g line, are n o w markedly different f r o m
that distant c o m m o n ancestor, coelacanths still look and act
like fish—and w e might as well say so. Cladists will put them
with rhinos, pheneticists with tunas; traditionalists will h o n e
their rhetoric to d e f e n d a necessarily subjective decision.
Nature has i m p o s e d this conflict upon science by decreeing, through the workings o f evolution, such unequal rates
o f change a m o n g lineages and such a p o o r correlation between phenetic similarity and recency o f c o m m o n ancestry.
I d o not believe that nature frustrates us by design, but I
rejoice in her intransigence nonetheless.