Nature 486, 523–526 (28 June 2012) doi:10.1038/nature11124
Received
Accepted
Published online
The origin of
tetrapods and the transition from swimming to walking was a pivotal step
in the evolution and diversification of terrestrial vertebrates. During
this time, modifications of the limbs—particularly the specialization
of joints and the structures that guide their motions—fundamentally
changed the ways in which early tetrapods could move1, 2, 3, 4.
Nonetheless, little is known about the functional consequences of limb
anatomy in early tetrapods and how that anatomy influenced locomotion
capabilities at this very critical stage in vertebrate evolution. Here
we present a three-dimensional reconstruction of the iconic Devonian
tetrapod Ichthyostega and a quantitative and comparative analysis of limb mobility in this early tetrapod. We show that Ichthyostega
could not have employed typical tetrapod locomotory behaviours, such as
lateral sequence walking. In particular, it lacked the necessary rotary
motions in its limbs to push the body off the ground and move the limbs
in an alternating sequence. Given that long-axis rotation was present
in the fins of tetrapodomorph fishes5, 6, 7, it seems that either early tetrapods evolved through an initial stage of restricted shoulder8, 9 and hip joint mobility or that Ichthyostega was unique in this respect. We conclude that early tetrapods with the skeletal morphology and limb mobility of Ichthyostega were unlikely to have made some of the recently described Middle Devonian trackways10.
Early tetrapods, and the fish that gave rise to them, were
originally interpreted as being terrestrially capable animals with
load-bearing fins or limbs11, 12, 13.
Since the 1990s, however, new fossil discoveries and anatomical
interpretations have demonstrated that the first limbed vertebrates were
primarily aquatic in habit and that limbs evolved before the ability to
‘walk’ on land1, 2.
More recently, work has suggested that hindlimb-powered locomotion
first evolved in sarcopterygian fish, well before the origin of
digit-bearing limbs or terrestriality14; this model implies that a muscularly supported pelvis and sacrum are not compulsory for fin/limb–substrate interactions10. Others have even unconventionally proposed that some early tetrapods may have been more seal-like in their mode of locomotion3, 4,
rather than moving in a primitively salamander-like fashion. This
apparent conflict surrounding the timing of events that gave rise to
modern tetrapod locomotory styles has left our understanding of the
evolution of terrestriality uncertain.
To illuminate the evolution
of early tetrapod locomotion, we conducted a computer-aided assessment
of limb joint mobility in one of the best known Devonian tetrapods, Ichthyostega. To achieve this goal, we used micro-computed tomography (μCT) to scan suitable Ichthyostega specimens, created a digitally rendered three-dimensional skeletal model (Fig. 1),
and quantified maximum range of motion in the shoulder, elbow, hip and
knee in three orthogonal planes of movement. To interpret joint mobility
in a locomotor context, we compared the data of Ichthyostega
with those of five morphologically and phylogenetically distinct modern
tetrapod analogues with varying joint morphologies and locomotion
behaviours. These include a salamander (Ambystoma tigrinum), crocodile (Crocodylus niloticus), platypus (Ornithorhynchus anatinus), seal (Halichoerus grypus) and otter (Lutra vulgaris).
Moreover, we validated our methodology through dissection and joint
manipulation to determine the effect of soft tissues (or lack thereof)
on joint mobility (see Supplementary Information).
Figure 1: Three-dimensional reconstruction of Ichthyostega from μCT scan data.
a, Anterolateral view. b, Dorsal view. c, Lateral view. d,
Ventral view. The forelimbs and hindlimbs are shown in their resting
pose from which ranges of motion were calculated. A list of the
specimens used to create the model and the procedure followed for model
construction can be found in Supplementary Information, as can a comparison with the most recent two-dimensional reconstruction presented in ref. 4. Scale bar, 10cm.
Data from the hindlimb demonstrate that the hip joint in Ichthyostega
has a comparable degree of mobility, in terms of flexion/extension and
adduction/abduction, to that of the modern tetrapods. However, unlike
the modern tetrapods, the hip joint of Ichthyostega has a minimal degree of long-axis rotation (pronation/supination) capacity (Fig. 2a,
but note that the seal, whose hindlimb is modified as a flipper, has a
reduced range of long-axis rotation in comparison with the other modern
taxa). When range of mobility is partitioned into positive (increasing
angle between limb and girdle) and negative (decreasing angle between
limb and girdle) angular movements (Fig. 3a), the hip joint of Ichthyostega shows a relatively equal distribution of mobility, with the pattern approaching that seen in the modern tetrapods (see Supplementary Table 5); however, there is somewhat more abduction capacity. In terms of knee joint movements, Ichthyostega displays the most restricted mobility in flexion/extension (see Supplementary Table 6),
but no discernible differences were recovered for the other two planes
of movement, presumably because soft tissues rather than osteology are
primary limits on knee mobility, as in many other tetrapods.
Figure 2: Maximum ranges of mobility in the limb joints of Ichthyostega and five modern tetrapod analogues.
a, Hip joint. b, Shoulder
joint. Mobility was examined in three orthogonal planes of movement
including flexion/extension, adduction/abduction and
pronation/supination (or long-axis rotation). The most obvious
difference between Ichthyostega and the modern tetrapods analysed
is a distinct lack of long-axis rotation. A validation test of the
method used to calculate range of mobility can been found in Supplementary Information.
Figure 3: Partitioned range of mobility in the hip joint and shoulder joint of Ichthyostega.
a, Hip joint mobility partitioned into positive and negative angular movements. b, Movement of the hip in flexion/extension. c, Movement of the hip in adduction/abduction. d, Shoulder joint mobility partitioned into positive and negative angular movements. e, Movement of the shoulder in flexion/extension. f,
Movement of the shoulder in adduction/abduction. Asterisks indicate the
resting pose. Specific details of the method used to calculate joint
range of mobility are provided in Supplementary Information, as are animations of maximum range of mobility in the shoulder and hip of Ichthyostega. Scale bars, 10cm.
The hip joint in Ichthyostega forms a condyloid-like
articulation with greatly enlarged dorsal and ventral bony buttresses
surrounding an anteroventrally-to-posterodorsally elongated acetebulum13.
In addition to this, the femoral head is boomerang-shaped with a large
ventral intertrochanteric fossa enveloping the ventral bony buttress of
the acetabulum, forming a locking mechanism. This type of hip morphology
permits the femur to rock along the primary and secondary axes of the
acetabulum but prevents any major long-axis rotary movements. Because
the primary axis of the acetabulum (and associated femoral head) in Ichthyostega
is tilted anteriorly, movement of the hindlimb would have occurred at
an angle about 45° from the main horizontal axis of the body.
Consequently, the hindlimb would primarily have moved in an
anteroventral-to-posterodorsal arc during hip extension and an
anterodorsal-to-posteroventral arc during hip adduction (Fig. 3b, c).
The
limited range of hip long-axis rotation, in combination with a joint
offset of about 45° from the horizontal, implies that the plantar
surface of the pes in Ichthyostega was unable to contact the
substrate. In particular, the femur would have been prevented from
attaining a horizontal orientation and the pes would not have been
capable of pointing anteriorly, a limb pose conventionally considered
plesiomorphic for terrestrial tetrapods15, 16 (see Supplementary Figs 5 and 6).
This means that the pelvis could not have been lifted free of the
ground and that the hindlimbs were more critical during swimming, with a
more passive or stabilizing function during land/substrate movement
(analogous to phocid seals17). This proposed model of hindlimb movement in Ichthyostega is in accordance with the development of broad, paddle-shaped, distal limb bones and expanded pedes1.
It is also further supported by a reduced range of flexion/extension in
the knee, which is an essential component of terrestrial locomotion in
living tetrapods18, 19, 20, 21.
With respect to the forelimb, the shoulder joint in Ichthyostega has the most restricted range of angular motions in all three planes of movement (Fig. 2b),
with the most striking characteristic again being a distinct lack of
long-axis rotation (pronation/supination). Partitioning the range of
motion into positive and negative angular movements further reveals that
the shoulder joint in Ichthyostega is primarily restricted to negative angular movements, with the majority of mobility occurring in flexion and adduction (Fig. 3d). In contrast to Ichthyostega, the modern tetrapods examined show appreciable shoulder mobility in all three planes of movement (Fig. 2b) and a more equal distribution between positive and negative angular movements (see Supplementary Table 3). Elbow joint mobility in Ichthyostega
is, as far as can be judged from the ulna articulation, comparable to
that in the modern tetrapods. It is relatively flexible in all degrees
of freedom and shows a pattern intermediate between more sprawling
animals and those that use more upright limb postures (see Supplementary Table 4).
The restricted range of shoulder mobility in Ichthyostega is primarily the product of an anteroposteriorly elongated and dorsoventrally flattened glenoid fossa and humeral head13.
As with the hip, this type of joint morphology produces a
condyloid-like joint articulation, permitting flexion/extension and
adduction/abduction but preventing humeral long-axis rotation. Further
to this, the glenoid fossa in Ichthyostega is somewhat twisted
along its primary axis, with its anterior portion facing ventrally and
its posterior portion facing dorsally. The morphology of the glenoid
fossa in Ichthyostega guides the forelimb to move in a slight
anteroventral-to-posterodorsal plane during shoulder joint flexion,
whereas during adduction it directs the forelimb to move in an inclined
plane from the vertical, causing the humerus to trend posteriorly (Fig. 3e, f). In addition to joint shape, the pectoral girdle of Ichthyostega
also has a large bony buttress surrounding the anterodorsal border of
the glenoid itself, further restricting movements in extension and
abduction.
Given that symmetrical gaits (for example lateral
sequence walking, trotting) require a large degree of limb retraction
and rotation, in addition to girdle rotation by means of bending of the
vertebral column15, 16, 18, 19, 20, 21, 22, 23, it is unlikely that Ichthyostega employed such gaits with its forelimbs. Indeed, the pattern of shoulder joint mobility in Ichthyostega, in combination with a rigid pectoral girdle and thorax (due to large overlapping ribs13)
and terrestrially ineffectual hindlimbs, indicates that the most likely
mode of forelimb movement on land/substrate involved synchronous
mudskipper-like ‘crutching’ motions24. This proposed locomotory behaviour for the forelimbs of Ichthyostega concurs with previous conjectures3, 4
and is further supported by highly developed elbow extensor musculature
(implied by large dorsally extending olecranon processes)9
attached to a relatively mobile elbow joint. In addition to
land/substrate movement, however, the large amount of shoulder adduction
and elbow extension in the forelimb of Ichthyostega would have enabled both station-holding and lifting of the head out of the water to breathe and potentially feed25.
On the basis of our study of limb joint mobility, combined with rib and vertebral morphology4, 13, we conclude that Ichthyostega
could not use ‘normal’ quadrupedal gaits. The ability to rotate the
humerus and femur longitudinally and use symmetrical gaits (for example
lateral sequence walking) must have evolved in other early tetrapod
species. Given that a similar type of shoulder and/or hip joint
morphology presents itself in some other early tetrapod species2, 8, 26, limited limb joint mobility—particularly long-axis rotation—may have been more widespread (see Supplementary Information).
However, the use of symmetrical gaits, or lack thereof, in other early
tetrapods would need to be tested through further three-dimensional
investigations of both the limbs and axial skeleton. Nevertheless,
long-axis rotation capacity was present in the fins of tetrapodomorph
fishes5, 6, 7, indicating that some tetrapods evolved through a phase of restricted shoulder8, 9
and hip joint mobility before acquiring the ability to perform rotary
motions and the associated locomotory behaviours that persist in extant
taxa.
In addition, our data indicate that the tetrapod forelimb
was the first to gain a role in land/substrate locomotion, with the
pelvis and hindlimb becoming robust more as a muscular and propulsive
adjunct to the tail for swimming and only later being exapted for
walking on land (contra ref. 14).
The divergent functional roles of early tetrapod limbs and the late
onset of hindlimb-powered land/substrate locomotion are consistent with
both the diminutive pelvic fins of tetrapodomorph fishes compared with
their pectoral fins, and with the genetic mechanisms controlling the
formation of fin and limb musculature. In particular, studies
demonstrate a distinct evolutionary lag between the developmental modes
of pectoral and pelvic fin musculature in bony fishes27, 28. Therefore, in addition to hip joint constraints, it is conceivable that Ichthyostega
did not yet have the pelvic musculature necessary for hindlimb-driven
land/substrate movement, but rather gradually transformed the existing
musculature from a more aquatic to a more terrestrial propulsive role.
Given our results, could an Ichthyostega-like early tetrapod have produced similar trackways to some of those recently described from the Middle Devonian10? All available evidence from limb joint mobility and axial anatomy4, 13
indicates that such animals could not have made symmetrical gait ‘foot’
prints. In particular, these early tetrapods probably lacked the
necessary rotary motions in their limbs (and perhaps lateral flexion of
the vertebral column) to push the body off the substrate and progress
using alternating limb movements. Maybe as yet unknown tetrapod species
(or known taxa that currently lack postcranial material) with different
joint mobility and axial anatomy made these traces; available data
cannot yet answer this conundrum. Nonetheless, the analysis presented
here supports the possibility that Ichthyostega-like animals
could produce synchronous (parallel) trackways, as our findings indicate
that such a trace should consist of a series of bilateral forelimb
impressions.
A large selection of Ichthyostega specimens (see Supplementary Information), particularly focusing on postcranial material, and five modern analogues including a Tiger salamander (Ambystoma tigrinum; TNHC 17991), Nile crocodile (Crocodylus niloticus; RVC ‘Flunch’ specimen in J.R.H.’s research collection), Platypus (Ornithorhynchus anatinus; USNM 221110), Grey seal (Halichoerus grypus; UMZC K.7943) and European otter (Lutra vulgaris; UMZC K.2768), were scanned with a medical CT or μCT
(depending on size; settings varied widely) to capture
three-dimensional bone geometry. All files were segmented in Mimics
software (Materialise), exported as high-resolution .stl files and then
reconstructed in the three-dimensional modelling, animation and
rendering software Autodesk 3D Studio Max. Because articular cartilage
is no longer present on the bones of Ichthyostega, all joints in
the extant taxa were also reconstructed without any articular cartilage,
ensuring that modelling conditions were kept consistent; however, we
investigate this assumption further below.
Model alignment
All animals and models (Supplementary Fig. 1)
were articulated in a stepwise fashion. First, the skull and vertebral
column of each animal studied were aligned in a straight line; however,
because of the ventrally directed sacral ‘hip fan’ in Ichthyostega,
the tail was left sloping ventrally. The pectoral and pelvic girdles
were then positioned. Unlike the pelvic girdle, which attaches to the
sacrum, the placement of the pectoral girdle is slightly more
subjective. The salamander, platypus and crocodile were all scanned as
cadaveric specimens; positioning of the pectoral girdle was therefore
fairly straightforward. The seal and otter, however, were scanned as
disarticulated skeletons; in this case, the pectoral girdle (or scapula)
was considered to sit over the anterior ribs with the straight edge of
the scapular blade pointing dorsally and not extending above the level
of the neural spines. In Ichthyostega the pectoral girdle was
aligned with an impression of the left cleithrum on the thoracic ribs of
MGUH VP 6115. The limbs were then positioned and their centres of
rotation and joint axes aligned (see below).
Limb joint axes
With
respect to the shoulder and hip, spheres were placed in the glenoid and
acetabulum to estimate each joint’s centre of rotation, assuming a
ball-and-socket (three degrees of freedom) joint. The humerus and femur
were then positioned in an anatomical resting pose, which was determined
by moving the bones until they were gently resting—with no bone
contact—in their respective sockets/spheres. As a result, the resting
pose inherently accounts for a certain thickness of soft tissue (for
example articular cartilage). However, because of the elongated and
narrow glenoid and acetabulum in Ichthyostega, the centre of
rotation in the shoulder and hip was determined by first aligning and
scaling a cylinder along the primary axis of the glenoid and acetabulum
and then aligning a sphere, with the same cross-sectional diameter, to
the cylinder’s centre of rotation; the humerus and femur were then
positioned into a resting pose, and the size of the sphere was adjusted
to ensure that it entirely enveloped the humeral or femoral head. In
addition, because the glenoid in Ichthyostega twists along its
primary axis, a plane was also placed through the glenoid to estimate,
more realistically, the orientation of the humerus.
After
establishing each joint’s centre of rotation, the three orthogonal axes
of each sphere were aligned to the anatomical axes of the humerus and
femur, first by determining the long axis of the bone, and then
orienting the craniocaudal axis and dorsoventral axis (in sprawling
animals) or mediolateral axis (in parasagittal animals). Then the
centres of rotation and joint axes of the humerus and femur were aligned
with their respective spheres (Supplementary Fig. 2).
The same procedure was followed for aligning the antebrachium (ulna and
radius), manus, crus (tibia and fibula) and pes, but this time spheres
were placed on the distal ends of the humerus and femur to estimate
joint axes in the elbow and knee or the ulna/radius and tibia/fibula to
estimate joint axes in the wrist and ankle (which were not examined in
this study, particularly because of the absence of a manus and the lack
of a clear ankle joint in Ichthyostega).
Bone segments
To
reduce processing time, the high-resolution Autodesk 3D Studio Max
models were partitioned into two files, one composed of the pectoral
girdle and left forelimb and another of the pelvic girdle and left
hindlimb. Individual bones were grouped to create four segments: the
girdle segment (pectoral or pelvic), followed by the proximal limb
segment (humerus or femur), the distal limb segment (antebrachium or
crus) and finally the manus or pes segment. However, because the ulna
and radius in Ichthyostega do not articulate in close proximity
to each other, and the radial condyle is partly missing, the distal
segment of the forelimb in Ichthyostega consists solely of the
ulna. To preserve positional relationships, segments (and associated
spheres) were linked hierarchically such that if a segment was moved,
all associated distal segments moved in the same direction.
SIMM model
To
quantify maximum range of movement, each limb model was recreated in
the biomechanical modelling software SIMM (Musculographics, Inc.)29.
To build a SIMM model, each bone’s positional relationship in a
Cartesian coordinate system was determined in Autodesk 3D Studio Max.
First, the centre of rotation or pivot point of the girdle segment was
aligned to x,y,z coordinates of 0,0,0. The x,y,z
coordinates of the proximal segment were recorded and then the pivot
point of the proximal segment (and associated sphere) was realigned to
0,0,0, such that it was overlapping the girdle segment. This procedure
was repeated for the distal segment and then finally the manus/pes
segment. The joint axes were then determined using each segments
associated sphere. To do this, the pivot point of each sphere was moved a
short distance along its local xaxis (craniocaudal direction) and then the new x,y,z
location (compared with 0,0,0) was recorded. The pivot point was then
returned to 0,0,0 and the same procedure was repeated for the local yaxis (dorsoventral or mediolateral) and zaxis
(long axis of bone). The coordinates of the joint axes were then
normalized by dividing each by its standard deviation. Finally, all
0,0,0 aligned segments were exported as .stl files to create a ‘bone
file’ and the positional relationships were combined to construct a
linked ‘joint file’ (.jnt).
Range of movement
In SIMM, each joint was modelled with three degrees of freedom: adduction/abduction (about the xaxis), flexion/extension (about the yaxis) and pronation/supination (that is, long-axis rotation; about the zaxis) (Supplementary Fig. 2).
The terms used to describe movement in this study are specific to joint
angles rather than net limb/segment movement during active locomotion
(for example, protraction/retraction, elevation/depression). In terms of
angular movements, the resting pose of each segment was set to a
starting angle of 0°. Movements in adduction, flexion and pronation were
considered to be negative angular movements (decreasing angle between
segments) and ranged from a minimum of 0° to a maximum of −180°.
Conversely, movements in abduction, extension and supination were
considered to be positive angular movements (increasing angle between
segments) and ranged from a minimum of 0° to a maximum of +180° (Supplementary Fig. 2).
Finally, to obtain the maximum range of motion three assumptions were
followed: first, joints were only permitted to move along one plane at
any time (that is, movements were uncoupled); second, joints were
permitted to translate or slide within their presumed joint capsule30; and third, joints were moved in 5° increments until there was either bone-to-bone contact or the joint became disarticulated.
Institutional abbreviations
MGUH,
Geological Museum, Copenhagen; RVC, Royal Veterinary College, London;
TNHC, Texas Natural History Collections; UMZC, University Museum of
Zoology, Cambridge; USNM, Smithsonian Institution National Museum of
Natural History.
Clack, J. A.Devonian tetrapod trackways and trackmakers: a review of the fossils and footprints. Palaeogeogr. Palaeoclimatol. Palaeoecol.130, 227–250 (1997)
Shubin, N. H., Daeschler, E. B. & Jenkins, F. A., JrThe pectoral fin of Tiktaalik roseae and the origin of the tetrapod limb. Nature440, 764–771 (2006)
Ahlberg, P. E.Humeral homology and the origin of the tetrapod elbow: a reinterpretation of the enigmatic specimens ANSP 21350 and GSM 104536. Spec. Pap. Palaeontol.86, 17–29 (2011)
Niedzwiedzki, G., Szrek, P., Narkiewicz, K., Narkiewicz, M. & Ahlberg, P. E.Tetrapod trackways from the early Middle Devonian period of Poland. Nature463, 43–48 (2010)
King, H. M., Shubin, N. H., Coates, M. I. & Hale, M. E.Behavioural evidence for the evolution of walking and bounding before terrestriality in sarcopterygian fishes. Proc. Natl Acad. Sci. USA108, 21146–21151 (2011)
Reilly, S. M. & Elias, J. A.Locomotion in Alligator mississippiensis: kinematic effects of speed and posture and their relevance to the sprawling-to-erect paradigm. J. Exp. Biol.201, 2559–2574 (1998)
Jayne, B. C. & Irschick, D. J.Effects of incline and speed on the three-dimensional hindlimb kinematics of a generalized iguanian lizard (Dipsosaurus dorsalis). J. Exp. Biol.202, 143–159 (1998)
Ashley-Ross, M. A., Lundin, R. & Johnson, K. L.Kinematics of level terrestrial and underwater walking in the California newt, Taricha torosa. J. Exp. Zool.311A, 240–257 (2009)
Jenkins, F. A., Jr & Goslow, G. E., JrThe functional anatomy of the shoulder of the savannah monitor lizard (Varanus exanthematicus). J. Morphol.175, 195–216 (1983)
Pridmore, P. A.Submerged walking in the epaulette shark Hemiscyllium ocellatum (Hemischyllidae) and its implications for locomotion in rhipidistian fishes and early tetrapods. Zoology98, 278–297 (1994)
Markey, M. J. & Marshall, C. R.Terrestrial-style
feeding in a very early aquatic tetrapod is supported by evidence from
experimental analysis of suture morphology. Proc. Natl Acad. Sci. USA104, 7134–7138 (2007)
Delp, S. L. & Loan, J. P.A graphics-based software system to develop and analyze models of musculoskeletal structures. Comput. Biol. Med.25, 21–34 (1995)
We thank J. Molnar for assistance with segmentation, model
construction, and movie generation; L. Witmer for access to platypus
scan data; Digimorph for access to salamander scan data (National
Science Foundation grant IIS-9874781 and IIS-0208675 to T. Rowe); M.
Lowe for collections support at the University Museum of Zoology,
Cambridge; G. Cuny for access to collections housed in the Geological
Museum at the University of Copenhagen; J. Rankin for musculoskeletal
modelling support; and P. Ahlberg for commenting on an earlier draft of
this manuscript. For access to μCT scanning
equipment in their care, we also acknowledge R. Abel, C. Martin and A.
Heaver. This research was supported by Natural Environment Research
Council grants NE/G005877/1 and NE/G00711X/1.
Department of Veterinary Basic Sciences and Structure and
Motion Laboratory, The Royal Veterinary College, Hawkshead Lane,
Hatfield AL9 7TA, UK
Stephanie E. Pierce &
John R. Hutchinson
University Museum of Zoology, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK
Stephanie E. Pierce &
Jennifer A. Clack
Contributions
All authors contributed to project concept and design. S.E.P.
collected and analysed the data and wrote the manuscript, including main
text, figures and Supplementary Information. J.A.C. and J.R.H. provided a critical review of all aspects of manuscript development. All authors approved the final draft.
Competing financial interests
The authors declare no competing financial interests.
This
proof contains Supplementary Text, which includes details of how the 3D
model of Ichthyostega was constructed, Supplementary References,
Supplementary Figures 1-6 and Supplementary Tables 1-6.
![Three-dimensional reconstruction of Ichthyostega from [mgr]CT scan data.](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_v__mHGJirnIcIx1SWQfNFyTMFbNm_isKKXDmGnIDcPGQZC6YfZr5tTF3IHgUb2m7e22QQkrTmPykIIMDbWE9E6JgBfq7LuO4gRTdhBc8RIaacO4rajlH7h5J0tPNKcqjW1mPXHLMbBWVnHeIect0Mr=s0-d "Figure 1: Three-dimensional reconstruction of Ichthyostega from [mgr]CT scan data.")
![Three-dimensional reconstruction of Ichthyostega from [mgr]CT scan data.](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_uPE79lFdVAISMzw6iN94Jjv6fvt77gXCKDhCXzA-iaJ2Ee8fd6hQ4hqt2YZjri2RE_HRU9lMsy5wJ2gSMwcugFiG_aBbtN5xTAQvSsA3AmFgUmh_rdZmxMKpOKN6zzgGZz6i_uP3E-f7WIXea2mU4tJpNkqhPM=s0-d)
