finding whales’ sisters
Although finding and identifying the astragalus (figure 39) seems a sem-
inal moment in our thinking of the relatives of whales, it will not be
The Skeleton Puzzle | 133
Dog
the mesonychid ungulate
Pig
generalized
a modern artiodactyl
ankle shape
Dissacus
MNHN BR 211
trochlea
trochlea
head
head
more or less
has the shape
flat or convex
of a pulley (trochlea)
in al directions
in all artiodactyls
the fossil whale
the fossil whale
Pakicetus
Ambulocetus
the fossil artiodactyl
H-GSP 98148
H-GSP 18507
Indohyus
RR 224
trochlea
trochlea
head
was broken in this fossil,
its shape not known
head
has the shape
head
of a trochlea, just
as in living artiodactyls
figure 39. The astragalus is the bone on which the ankle pivots in mammals. The
dog astragalus shows the primitive condition, where the head is more or less convex
in all directions, while the top part, on which the ankle pivots, is a trochlea (pulley).
In artiodactyls, the head also has the shape of a trochlea. Whales that still have hind
limbs, such as Pakicetus, have an astragalus similar to artiodactyls. For
Ambulocetus, that part of the bone was not found. Indohyus is a close relative of Khirtharia.
enough to convince the world. That will require an explicit considera-
tion of all of the morphology of the cetaceans and all of their potential
relatives: a cladistic analysis. In a cladistic analysis, all differences
between animals are compiled in a table called a character matrix, and
all of those differences are explicitly described. For instance, the shape
of the astragalus is a character of relevance, and one could describe that
character as having two states: “astragalar head has the ball-shape of a
condyle” and “astragalar head has the pulley-shape of a trochlea.”
Numbers are then assigned to these states, usually zero and one (more
if it is a complex character), and the computer maps those on different
134 | Chapter 10
cladograms and calculates how many evolutionary changes would take
place (figure 40). Our character matrix for the whale work takes its
characters from our own work, but also from that of colleagues in the
whale field such as Zhe-Xi Luo, Mark Uhen, Jonathan Geisler, and
Maureen O’Leary.3 The addition of a pakicetid skeleton to the matrix in
a cladistics analysis showed indeed that mesonychians should be evicted
from the extended family of cetaceans.4
Determining How Animals Are Related
Our character matrix has 105 characters that are columns of numbers,
mostly zeros and ones. The twenty-nine species studied are the rows
in the matrix. They include pakicetids and Ambulocetus, as well as
artiodactyls from hippos to mouse deer, and several mesonychians.4
The computer makes sense of the matrix by trying out possible com-
binations of proposed relationships and calculating how many evo-
lutionary changes each would take. For instance, the computer will
propose that Ambulocetus and pakicetids are sister groups, and that
their next-closest relative is one of the artiodactyls, and this can be
summarized in a cladogram (simplified version in figure 40, top).
The computer then determines where on the cladogram each char-
acter would change given the particular relationship proposed in the
cladogram and taking into account what the state of that character
is in a group that is the most distant relative of all of them (the out-
group). For instance, we can plot the astragalar character on the top
an astragalar head in the shape of a condyle. Hence, at the base of
the cladogram, the astragalar character is in the zero state. Moving
to the next branch on the cladogram, artiodactyls have a head that
looks like a trochlea, so that means that an evolutionary change took
place at that line segment from zero to one, as indicated by the short
dash and the arrow between zero and one. Since Pakicetus is similar
to artiodactyls, no change took place at the next branch, or any other
branch.
To reason through this for multiple characters instead of one is too
complicated for a human brain, but the computer does it by trying
other hypotheses of relationships, as for instance in the second clado-
gram of figure 40, where mesonychians, not artiodactyls, are the sister
group to Ambulocetus and Pakicetus. In this cladogram, the change
leading to the astragalar head in the shape of a trochlea takes place
on the branch between primitive ungulates and artiodactyls (change
from zero to one), and then that character reverses to its original state
(change from one to zero) at the branch between artiodactyls and mes-
onychians, to reappear once more (zero to one) at the branch between
The Skeleton Puzzle | 135
the mesonychians and the cetaceans. This evolutionary hypothesis
takes three steps. If evolutionary change is rare, then the relationships
suggested by this cladogram are less likely than those of the first clado-
gram, which only took one step.
But wait, we can maintain the relationships of the second clado-
gram and yet decrease the number of evolutionary steps. The third
cladogram proposes that the re-emergence of a condyle-shape could
occur on the line to mesonychians instead of in the common ancestor
of mesonychians, pakicetids, and Ambulocetus. Even though the sec-
ond and third have identical branching patterns, they make different
statements about the evolution of the shape of the astragalus, and the
third takes only two evolutionary steps. That is still more than the
arrangement of the first cladogram, so the computer will point to the
first cladogram as the most likely reflection of what happened in real
life. It is easy to imagine that this gets more complicated if we have to
try all possible branch
ing patterns for twenty-nine species, and impos-
sible to do by hand if there are 105 characters that do not evolve in
unison and often point in conflicting directions. However, the compu-
ter can keep all of this straight and figure out the branching pattern
that requires the fewest evolutionary changes. That is called the most
“parsimonious” cladogram.
The field of systematics studies the relationships among animals by
using the cladistics analyses explained in the sidebar. Esoteric as it seems
to most laypeople, it is one of the most contentious areas of the study of
whale origins, and some of the most argumentative scientists are sys-
tematists. Having said all of that, figuring out the relationships among
the animals that you study is important for just about any other aspect
of biology. The publication of the pakicetid skeleton with a cladistics
analysis on all the whales5 coincides with the publication of another
Eocene whale skeleton from Pakistan by Philip Gingerich and col-
leagues,6 and those papers seal the issue for most scientists: whales are
related to artiodactyls. That does not mean that the fossil data are
totally in agreement with the DNA data. The fossil data show that
some artiodactyl (as opposed to a mesonychian) is the closest living
relative of cetaceans, but it does not point to a particular artiodactyl as
being in that position. A mountain of DNA data indicate that hippos
are the closest living relatives of whales;7 the fossils are just not that
specific.
This bothers me. DNA data can never address the possibility that
some fossil artiodactyl is even more closely related to whales than hip-
pos are, because it is not possible to get DNA out of such old fossils. For
136 | Chapter 10
Other ungulates,
including mesonychids
artiodactyls
Astragalar
Astragalar head shape
head shape:
0: flat to bal -shaped
1: pulley-shaped (trochlea)
0 1
Pakicetus
Ambulocetus
Other ungulates
Astragalar
artiodactyls
head shape:
0 1
mesonychians
1 0
0 1
Pakicetus
Ambulocetus
Other ungulates
Astragalar
artiodactyls
head shape:
0 1
1 0
mesonychians
Pakicetus
Ambulocetus
figure 40. Three cladograms that show to which group of
mammals cetaceans may be related. Most scientists support the top
cladogram. Each of these cladograms has implications for how the
astragalus evolved. A zero indicates that the astragalar head was
convex; a one indicates that it was a trochlea; and an arrow indicates
that an evolutionary change took place in one direction or the other.
now, I have to settle for less: the new evidence has routed the mesony-
chians in favor of artiodactyls as cetacean relatives. That is a big deal.
Now we can focus on how pakicetids lived. In the future, I will be pay-
ing more attention to fossil artiodactyls as I think about whales, but for
now, I indulge in a part of science for which I have had a weak spot ever
since my first brush with whales, a long time ago: hearing.
Chapter 11
The River Whales
hearing in whales
The new pakicetid skulls can really help with learning about hearing. It
was clear already that cetacean hearing changed when the ancestors of
cetaceans went underwater. Land ears work poorly underwater, because
sound in air differs from sound underwater. The fossils showed it too:
that first pakicetid incus did not resemble modern whales or modern
land mammals (figure 3); that thick involucrum must have done some-
thing to sound transmission (figure 2); and the mandibular foramen
grew bigger over the course of the Eocene (figure 25).
In general, all the anatomical parts of the organ of hearing in whales
can be found in land mammals too, but the shapes are different (figure
41). Land mammals have a canal in the side of the head that gives entry
to sound: the external auditory meatus. It ends at the eardrum. Behind
the eardrum are the three ossicles already mentioned in figure 3: malleus
(hammer), incus (anvil), and stapes (stirrup). In most mammals, the
ossicles are loosely suspended within an air-filled cavity, the middle ear
cavity, which is protected by a protective bony shell, the tympanic bone
in whales. The malleus looks like a club, its narrow handle firmly
attached to the eardrum, and its wide part having a joint with the incus.
As sounds make the eardrum vibrate, the malleus vibrates, and the
vibrations are passed on to the incus. The incus has two arms, the crus
breve and the crus longum. The former is anchored into the wall and
137
ear canal wall of skull
Land mammals
Incus large
Pakicetidae
(external
incus
and dense
malleus
such as Artiodactyla
(malleus not
auditory
partly rotated
known)
meatus)
cranial cavity
eardrum
middle
canal for nerve
(tympanic
ear
to brain
membrane)
cavity
(internal
acoustic meatus)
wall of middle ear
inner ear
(tympanic bone)
stapes
petrosal
involucrum
bone
mandibular foramen
lower jaw
ossicles further enlarged Remingtonocetidae
Ambulocetidae
reoriented in middle ear
(no ossicles
malleus fused to tympanic
known)
tympanic plate
mandibular
large and thin
partial
foramen
enlarged
further enlarged
isolation
mandibular
of ear region
bony contact between foramen
from skull
mandible and tympanic
mandibular
eardrum drawn out
wall thin
into cone shape
further isolation of
ear region from skul
Odontoceti
tympanic ring
greatly reduced
in size
external auditory
Basilosauridae
meatus lost
petrosal completely
fat pad located in jaw
unattached
passing through
to skul in some
mandibular foramen
odontocetes
to the tympanic bone
figure 41. The ear in land mammals and whales. The
diagram at the top left identifies
all of the parts. Labels in other diagrams indicate which changes took place at each
evolutionary step leading to modern whales. Dashed lines indicate bones not known for
the group in question; their shape has been inferred from other groups.
The River Whales | 139
helps in keeping the ossicles suspended and able to pivot. The crus
longum has a joint with the stapes. As the incus pivots, the stapes is
pushed in and pulled out of a small hole in yet another bone, the oval
window in the petrosal bone. Behind the oval window is a cavity in the
shape of a snail shell (the cochlea of the inner ear) that is filled with
fluid. The pumping causes movements in the fluid, and that stimulates
modified nerve cells that are arranged in a row along the length of the
cochlea, passing the signal on to the brain.
In modern odontocetes (last diagram of figure 41), there is no open
external auditory meatus; the duct is closed off by the tissues around it.
The most sound-sensitive part of the face of a dolphin is actually the
skin over the lower jaw, the mandible,1 and sound travels from there
through that large fat pad housed in the mandibular foramen of the
lower jaw (figure 25). Sound constitutes vibrations that pass through a
material, and these vibrations are passed on to the very thin part of the
tympanic bone, the tympanic plate. Since it is made of bone, the tym-
panic plate has unique vibrational properties that are needed for the
high-frequency sounds that odontocetes echolocate with. The eardrum
is still present, but it is not a flat membrane. It looks like a folded-in
umbrella. It may not have a function in hearing at all.2 In whales also,
the malleus is connected by bone to the edge of the tympanic plate;
sounds are transmitted by the ear ossicles to the cochlea; and the latter
works the same as in other mammals. The function of the involucrum is
not well understood. It has been proposed that it is a counterweight
during sound transmissions of the tympanic plate,3 but the exact sound-
transmission mechanism through the odontocete middle ear remains
controversial, and sophisticated computer modeling of this area sug-
gests that mechanisms may be different for different cetacean species
and even at different frequencies.4 The ossicles are much heavier in
whales than in land mammals. That is strange—sound does not carry
much energy, and it would be easier for faint sounds to make those
ossicles vibrate if they were lighter. Possibly, the ossicles do not vibrate