draw a line between 55 and 56 percent while not separating
54 and 55 percent. There are methods that at least establish
statistically distinguishable groupings within observed
proportions (e.g., gap-coding), but even doing so does not
satisfy the severest critics of proportional characters (e.g.,
Farris, 1990). Sander (2000) also failed to provide justifications
for his divisions of proportions. Therefore, it is unlikely
that the character-states designated by him represent
synapomorphies. For these reasons, Sander’s (2000)
tree is not incorporated in the higher taxonomy presented
in this study.
Maisch & Matzke (2000B) published yet another review
of ichthyosaurs, containing a phylogenetic analysis
(Fig. 100D). Their data matrix was extended from previous
studies. However, there were some methodological problems
in the analysis that compromised the validity of the
result obtained. The most serious problem was character
polarization. Their data matrix does not contain any outgroup
per se, but only employs an all-zero ancestor (i.e., an
imaginary taxon with subjectively determined state 0 for
all characters). The authors stated that the character polarity
was judged based on Captorhinus, Limnoscelis, and
Youngina, but it is not explained how this was achieved,
e.g., how the outgroup topology was defined, how the
ancestral characters were optimized, and whether the data
were obtained from the literature or from observation of
specimens. Also, it is not clear how characters specific to
ichthyosaurs were polarized. When removing the all-zero
ancestor from the data matrix, different tree topologies
than those given by Maisch & Matzke (2000B), even as
unrooted trees, were obtained. Therefore, the tree topologies
given by Maisch & M atzke (2000B) entirely depended
upon the assumption of an imaginary all-zero ancestor.
(Sander [2000] also used an all-zero ancestor in his data
matrix, but its removal has no effect on the tree topology).
It is more advisable to include the three outgroup taxa in
the data matrix and specify the outgroup topology to ensure
objective optimization of the character states at the
ancestral node. Other problems include the use of unreliable
(i.e., schematic or preliminary) figures for character
coding. For example, sacral rib head character was coded
for Utatsusaurus based on a schematic reconstruction of the
pelvic girdle by Motani et al. (1998). Furthermore, Thaisau-
rus was coded almost exclusively based on the published
preliminary figure (the type specimen is as yet unprepared,
and the outlines of the exposed part, given in the
preliminary description by Mazin et al. [1991], do not
necessarily reflect the true bone outlines, judging from a
color photograph kindly provided by Dr. Mazin; see discussion
of Thaisaurus). The inclusion of such a poorly
known basal taxon, coding character-states from an illustration,
clearly compromised character optimization, as
the above reported removal of the all-zero ancestor indicates.
While improvements are expected in the future by
refining the data matrices of Motani (1999B), Sander
(2000), and Maisch & Matzke (2000B), Motani’s (1999B)
framework for higher taxonomy is adopted in the present
review (Table 5) because it seems to contain the least
number of methodological problems at this stage.
Functional morphology and biomechanics
Basic Principles
Behaviors of extinct animals cannot be directly observed,
and our inferences of their body functions are usually
made from their shapes as preserved in fossils. The central
assumption involved is that physical laws have not changed
between now and the time when the animal in question
was alive. Accordingly, objects sharing similar shapes and
sizes probably behave similarly, given similar environments.
It follows that the quality and testability of the
inferences thus made largely depends on how explicitly
the underlying physical principles are integrated into the
inference procedures.
Functional inferences in paleobiology share a common
pathway. First, a correlation between a certain shape and
function is established, using extant animals for which
behaviors can be observed. Then, using this correlation,
functions are inferred for extinct animals, whose shapes
are preserved in fossils. Functional studies can be roughly
categorized by how a correlation is established in the first
step. For example, the most basic level may be called
“intuitive functional morphology”, which is based on observed
yet unquantified correlation (Level 1 inference).
The inference can be slightly enhanced by using quantitative
techniques to establish correlations, and this may be
called “quantitative functional morphology” (Level 2 inference).
Level 2 inferences, although quantitative, lack physical
principles to explain the cause of the observed correlation,
and therefore remain strictly phenomenological. Finally,
the inference can be improved by using mechanical
or optical principles to establish correlations. Such biomechanical
or bio-optical inferences (or Level 3 inferences)
may still contain some phenomenological aspects, but are
more explicit than Level 2 inferences. The differences
among the three will be demonstrated in the next section,
using Stenopterygius as an example.
Swimming of Stenopterygius
Based upon the tuna-like body outlines preserved in some
specimens of Stenopterygius, most biologists would suspect
that this ichthyosaur was an efficient swimmer (see also
Introduction). At an implicit level, they correlate the thun-
niform body plan with efficient swimming, and then apply
this correlation to interpret the function of this fossil vertebrate.
This represents a very basic level of functional morphological
inference, based on intuitive observations rather
than quantification (Level 1).
Slightly more advanced inferences can be made by
using quantitative techniques, as exemplified by Motani et
al. (1996). These authors statistically established a correlation
between swimming styles and body proportions
(body fineness ratio and caudal fin aspect ratio to be precise)
in extant sharks, and used this relationship to infer the
swimming styles of various ichthyosaurs. This is a case of
quantitative functional morphology (Level 2). Stenopterygius
was again found to be a thunniform swimmer. The
weakness of this study (and many other functional morphological
studies) is the lack of explanation for the causes
behind the observed correlation. There is no reason why a
correlation in sharks should also apply to the phylogenet-
ically distant ichthyosaurs, unless the observed correlation
is shown to have been caused by physical principles that
act on the body when performing a function (i.e., external
constraints) rather than by phylogenetic factors.
The obvious solution is to incorporate physical principles
when establishing a correlation, to give physical (e.g.,
mechanical or optical) reasoning to the model. As long as
the model is rooted in physics, it is possible to establish its
validity across phylogenetically diverse groups. Such a
study is truly in the realm of biomechanics, and is still rare
in the paleobiological literature. (There are some studies on
flight and terrestrial locomotion.) A few studies on marine
reptiles claimed to be biomechanical while making only
Level 1 inferences; such a misconception would eventually
degrade the field and should be avoided. The study by
Massare (1988) probably represents the first attempt to use
fluid mechanics to infer some aspects of swimming in
marine reptiles from body shape. She calculated likely
maximum cruising speeds of Mesozoic marine reptiles
using an equation that defines the equilibrium between the
amounts of energy spent and that required for overcoming
hydrodynamic drag during steady swimming. The model
itself is still used in neontological studies (e.g., Hind &
Gurney, 1997), and thus is viable. However, this pioneering
work of Massare (1988) was not widely accepted because
it failed to predict the swimming speed when applied
to extant cetaceans. It is now known that a simple
error in the constant used in the equation caused this
problem (Motani, in press). An updated version of the
model is capable of predicting the optimum cruising speed
of extant swimming vertebrates, and, therefore, is useful in
inferring the swimming capabilities of Mesozoic marine
reptiles (Motani, in press). Massare’s (1988) accomplishments
deserve better acknowledgement in the literature.
Despite the viability of the energetics-based approach
of Massare (1988,1994), and Motani (2002B), these methods
have one shortcoming: they require the metabolic
rates to be known (or to be assumed) a priori. To solve this