Content uploaded by Hans-Peter Schultze
Author content
All content in this area was uploaded by Hans-Peter Schultze on Jan 17, 2019
Content may be subject to copyright.
Vertebral Column and Associated Elements in Dipnoans
and Comparison With Other Fishes:
Development and Homology
Gloria Arratia,
1
* Hans-Peter Schultze,
1
and Jorge Casciotta
2
1
Museum fu¨ r Naturkunde, Humboldt Universita¨ t, Berlin, Germany
2
Museo de La Plata, Departamento Cientı´fico de Zoologı´a, La Plata, Argentina
ABSTRACT A vertebral column consisting of a persis-
tent notochord and ossified arcocentra is the primitive
condition for Gnathostomata; it still persists in primitive
actinopterygians and sarcopterygians. Advanced acti-
nopterygians and sarcopterygians develop numerous
types of centra that include, among others, the presence of
holocentrum, chordacentrum, and autocentrum. The chor-
dacentrum, a mineralization or calcification of the fibrous
sheath of the notochord, is only found in actinopterygians,
whereas an autocentrum is a synapomorphy of teleosts
above †Leptolepis coryphaenoides. The chordacentrum,
formed by migration of cartilaginous cells from the arches
into the fibrous sheath of the notochord and usually cov-
ered by a thin calcification, is a unique feature of chon-
drichthyans. The actinopterygian chordacentrum and the
chondrichthyan chordacentrum are not homologous. The
postcaudal cartilaginous centrum is only known in post-
caudal vertebrae of living dipnoans. The holocentrum is
present in certain fossil dipnoans and actinopterygians,
where it has been independently acquired. It is formed by
proliferation of cartilage cells around the elastica externa
of the notochord. These cells later ossify, forming a com-
pact centrum. A vertebral column formed by a persistent
notochord without vertebral centra is the primitive pat-
tern for all vertebrates. The formation of centra, which is
not homologous among vertebrate groups, is acquired in-
dependently in some lineages of placoderms, most ad-
vanced actinopterygians, and some dipnoans and rhipidis-
tians. Several series of structures are associated with the
vertebral column such as the supraneurals, interhaemals,
radials, and ribs. In living dipnoans median neural spine,
‘supraneural’, and dorsal radial result from growth and
distal differentiation of one median cartilage into two or
three median bones during ontogeny. The median neural
spine articulates with the neural arch and fuses with it in
the caudal vertebrae early in ontogeny. Two bones differ-
entiate in the anterior abdominal vertebrae, i.e., the prox-
imal neural spine and the distal ‘supraneural.’ Three
bones differentiate in front of the dorsal fin, i.e., the prox-
imal neural spine, the middle ‘supraneural,’ and the distal
radial; the same pattern is observed in front of the anal fin
(the proximal haemal spine, the middle interhaemal, and
the distal radial). Considering that the three dorsal (and
also the three ventral) bones originate from growth of only
one cartilage, they cannot be serial homologs of the neural
spines, or ‘supraneural.’ They are linear homologs of the
median neural cartilage in living dipnoans. The develop-
ment of these elements differs within osteichthyans from
sarcopterygians to actinopterygians, in which the neural
spine originates as a continuation of the basidorsal arcua-
lia and in which the supraneural and radial originate from
independent cartilages that appear at different times dur-
ing early ontogeny. The ribs of living dipnoans are unique
in that they are not articulated with parapophyses, like in
primitive fossil dipnoans, but a remnant of the ventral
arcuale surrounded by a small arcocentrum remains at its
base. A true caudal fin is absent in living dipnoans. The
postcaudal cartilages extend to the caudal tip of the body
separating dorsal and ventral rays (or the camptotrichia).
Actinotrichia are present in young dipnoans. They are also
known in extant actinistians and actinopterygians. They
probably represent the primitive state for teleostomes. In
contrast, the camptotrichia are unique for extant dipno-
ans (and probably Carboniferous and younger dipnoans).
Lepidotrichia apparently developed many times among
osteichthyans. J. Morphol. 250:101–172, 2001.
© 2001 Wiley-Liss, Inc.
KEY WORDS: vertebrae; ‘supraneurals’; supraneurals;
radials; ontogeny; homology
Vertebrae and their formation have been studied
for a long time. One hundred years ago Gadow and
Abbott (1895) published a very influential hypothe-
sis on the formation and composition of the verte-
brae; however, the study of vertebrae still raises a
number of questions concerning the formation and
homology of the centrum, arch elements, and asso-
ciated bony elements (e.g., spines, supraneurals,
ribs) that are not yet solved. There are differences in
the origin and development of vertebrae in fishes
that make it questionable that they can be explained
by one developmental hypothesis. Thus, according to
Rosen et al. (1981) the vertebrae of actinopterygians
(with true or complete centra) have arisen on at
least four separate occasions. Laerm (1979) pro-
*Correspondence to: Gloria Arratia, Museum fu¨r Naturkunde, In-
valienstr. 43, D-10115 Berlin, Germany.
E-mail: gloria.arratia@rz.hu-berlin.de
JOURNAL OF MORPHOLOGY 250:101–172 (2001)
©2001 WILEY-LISS, INC.
posed three distinct patterns of formation of verte-
bral centra or central elements in primitive acti-
nopterygians alone (e.g., Australosomus, Birgeria,
Polypterus, Pteronisculus), and Arratia (1991)
showed that there are at least five patterns of for-
mation of the vertebral centrum in teleosts. Among
sarcopterygians, two vertebral patterns, each with
variations, were proposed for rhipidistians (Laerm,
1979).
There are numerous controversial ideas concern-
ing the formation of lungfish vertebrae and their
evolutionary trends, despite the fact that different
aspects of the dipnoan vertebrae of recent lungfishes
(e.g., Neoceratodus and Protopterus) have been stud-
ied (e.g., Hasse, 1893; Klaatsch, 1895; Gadow and
Abbott, 1895; Fu¨ rbringer, 1904; Goodrich, 1909;
Kerr, 1909; Miller, 1930; Mookerjee et al., 1954;
Percy, 1962; Schaeffer, 1967; Schultze, 1970; Shute,
1972; Bartsch, 1989; Schmitz, 1998), and that there
is available information on fossil forms such as
†Griphognathus and †Uranolophus (e.g., Schultze,
1969, 1975a; Campbell and Barwick, 1988). The
available information suggests much diversity
within lungfishes. For instance, no bony centra are
laid down in Neoceratodus, where the perichorda
thickens to produce a tough semicartilaginous
sheath and the cartilaginous bases of the arches
expand on its surface (Gu¨ nther, 1871; Klaatsch,
1895; Fu¨ rbringer, 1904; Goodrich, 1909). In con-
trast, autocentral components forming part of the
centrum are described in Protopterus (Mookerjee et
al., 1954; Bartsch, 1989). (The material studied by
Mookerjee and collaborators is lost, according to the
intensive search by B. Schaeffer in the 1960s; pers.
comm. to H.-P. Schultze.) Few fossil dipnoans (e.g.,
†Griphognathus,†Jarvikia,†Soederberghia) have
ossified centra (Jarvik, 1952; Schultze, 1969, 1970;
Campbell and Barwick, 1988).
The goals of this contribution are to describe in
detail, for the first time, the ontogenetic develop-
ment of vertebrae and associated elements in one
living lungfish, Lepidosiren paradoxa, to compare it
with those in other living and fossil lungfishes, and
to compare the dipnoan vertebral column with those
of other sarcopterygians, actinopterygians, and also
elasmobranchs. Finally, the homology of the differ-
ent elements of the vertebral column and associated
elements and the modes of formation of the lungfish
vertebrae and other fishes are discussed. Remane’s
(1952; e.g., criterion of position, criterion of special
quality, criterion of constancy or continuity) and
Ax’s (1987:158–178) concepts of homology are fol-
lowed. Consequently, “homologous features are fea-
tures in two or more evolutionary species which go
back to one and the same feature of a common stem
species” (Ax, 1987:159). “Non-homologous features
in two or more evolutionary species are features
which were not present in the common stem species;
they were evolved independent of each other” (Ax,
1987:159).
MATERIALS AND METHODS
The specimens studied (see Appendix) belong to
the following institutions and individuals: ANSP,
Academy of Natural Sciences, Philadelphia, Penn-
sylvania; BMNH, The Natural History Museum,
London, England; BGHan, Bundesanstalt fu¨ r Geo-
wissenschaften und Rohstoffe, Niedersa¨ chsisches
Landesamt fu¨ r Bodenforschung, Hannover, Ger-
many; BSPG, Bayerische Staatssammlung fu¨ r Pala¨-
ontologie und historische Geologie, Mu¨ nchen, Ger-
many; CAS(SU) California Academy of Sciences
(formerly cataloged as Stanford University), San
Francisco, California; FMNH, Field Museum of Nat-
ural History, Department of Geology, Chicago, Illi-
nois; GOE, Institut und Museum fu¨ r Geologie und
Pala¨ ontologie, Georg-August Universita¨t, Go¨ ttin-
gen, Germany; JM, Jura Museum, Naturwissen-
schaftliche Sammlungen Eichsta¨ tt, Germany; KU,
Division of Fishes and KUVP, Division of Vertebrate
Paleontology, Natural History Museum, The Uni-
versity of Kansas, Lawrence, Kansas; LACM, Divi-
sion of Paleontology, Los Angeles County Museum,
California; MB, Institut fu¨ r Pala¨ ontologie, Museum
fu¨ r Naturkunde der Humboldt Universita¨ t, Berlin,
Germany; MCSNIO, Civico Museo Insubrico di Sto-
ria Naturale di Induno, Olona, Italy; MHNM, Site
fossilife` re de Miguasha, Que´ bec, Canada; MNHN,
Muse´ um national d’Histoire naturelle, Institut de
Pale´ ontologie, Paris, France; MNHN-I, Muse´ um na-
tional d’Histoire naturelle, Institut de Ichtyologie,
Paris, France; MRCA, Muse´ e Royal de l’Afrique
Centrale, Tervuren, Belgium; NRM, Department of
Zoology, Fish Collection, Swedish Museum of Natu-
ral History, Stockholm, Sweden; OS, Department of
Fisheries and Wildlife, College for Agricultural Sci-
ences, Oregon State University, Corvallis, Oregon;
PC, private collection of G. Arratia; PCJC, private
collection of J. Casciotta; SHL, collection of Mr. Hel-
mut Leich, Bochum, Germany; SIO, Scripps Institu-
tion, University of California, La Jolla, California;
SMNH, Department of Paleozoology, Swedish Mu-
seum of Natural History, Stockholm, Sweden;
TCWC, Texas Cooperative Wildlife Collection, De-
partment of Wildlife and Fisheries Science, Texas
A&M University, College Station, Texas; UCLA,
University of California at Los Angeles, Department
of Biology, Los Angeles, California; UMMZ, Museum
of Zoology, The University of Michigan, Ann
Arbor, Michigan; ZMB, Institut fu¨ r Systematische
Zoologie, Fisch Sammlung, Museum fu¨ r Naturkunde
der Humboldt Universita¨ t, Berlin, Germany; ZMUC,
Universitets Zoologiske Museum, Copenhagen, Den-
mark. After we finish the studies of the dipnoan
material, specimens belonging to the private collec-
tions of G. Arratia and J. Casciotta will be deposited
in a museum collection.
102 G. ARRATIA ET AL.
Methods
Twenty-six specimens of Lepidosiren paradoxa
and 10 of Neoceratodus forsteri were cleared and
double-stained following the procedure described in
Arratia and Schultze (1992). Numerous alcoholic
specimens of Lepidosiren, Protopterus, and Neocera-
todus were dissected. Two specimens of Lepidosiren
were used for histological serial cross sections of the
whole fish. These cross sections (7–10 m thickness)
were stained following the Picro-Ponceau method.
The serial cross sections (10 m thickness) of 13
specimens of Protopterus sp. were stained with the
Azan-Mallory method.
Serial cross sections (7–10 m thickness) of nu-
merous teleosts (e.g., Elops saurus, Hiodon
alosoides, Dorosoma cepedianum, Oncorhynchus
mykiss, Thymallus thymallus, Esox americanus)
were also stained with the Picro-Ponceau method
and those of Amia calva and Lepisosteus osseous
with the Heidenhain’s Azan method. Thin sections
of certain fossil specimens were also studied; these
were prepared like petrographic thin sections.
Compound microscopes (Olympus and Leica) with
high resolution power, phase contrast and with po-
larized light optics and Wild M5 and M8 stereodis-
secting microscopes with camera lucida were used
for microscopical studies. The photographs of young
specimens were taken with a Nikon camera at-
tached to the compound Olympus microscope or with
a Leica camera attached to a Leica microscope (with
Agfaorthochromatic film). Drawings were executed
under the Wild stereodissecting microscope with
camera lucida attachment.
Some dipnoan specimens have damaged tails.
Nevertheless, we have taken the total length of spec-
imens as preserved; the preanal length was taken
also from all specimens for a comparative reference.
A high level of variability is present in the number
of vertebrae of modern lungfishes. The variation
greatly affects the anteriormost neural arches and
the postcaudal cartilages. Incomplete neural arches
may be present between the posterior margin of the
cranium and the first complete neural arch bearing
a spine. On the other hand, the number of postcau-
dal cartilages increases with age. To avoid misun-
derstandings, we give the total number of indepen-
dent vertebrae (not counting the vertebrae included
in the occipital region), and in addition provide
counts for the abdominal, caudal, and postcaudal
regions.
OBSERVATIONS AND RESULTS
Nomenclature
We explain our use of terms because identical
structures have received different names in previous
works on lungfishes. In addition, terms have been
used differently depending on the fish group. Our
use of terms in lungfishes, and other fishes as well,
is based on the development and homology of the
structures (see Fig. 1A–C for lungfish terminology).
Vertebra. The term “vertebra” includes all seri-
ally repeated ossified, cartilaginous, and ligamen-
tous elements around the notochord, consisting of
centrum, neural arch and spine, and haemal arch
and spine (Schultze and Arratia, 1988).
Fish vertebrae form from the appearance of suc-
cessive elements that fuse with each other. The pri-
mary elements are the arches (arcualia) that are
always present and the secondary elements are the
centra (centralia), which may be absent.
There are different types of vertebrae according to
their structure. For instance, three different kinds of
vertebrae can be distinguished based on the rela-
tionships of the arches and centrum: 1) aspondylous
vertebra (Fig. 2A), represented only by its arches or
arcocentra (e.g., in extant dipnoans and in some
fossil dipnoans such as †Dipterus, †Scaumenacia,
†Fleurantia, †Conchopoma; in fossil and extant ac-
tinistians; in cheirolepiforms, acipenseriforms, and
some primitive neopterygians); 2) dysospondylous
vertebra (Fig. 2B,C,E–I), characterized by arches
that are autogenous from the chordacentrum or os-
sified centrum (e.g., in fossil dipnoans such as
†Griphognathus, †Rhynchodipterus; in the riphidis-
tian †Megalichthys; in some Devonian palaeonisci-
forms, in polypteriforms, and some teleosts such as
Elops); and 3) holospondylous vertebra (Fig. 2D),
where arches and ossified centrum are fused into
one unit (e.g., in Lepisosteus; caudal vertebrae of
most advanced teleosts).
Another classification of the vertebrae concerns
the number of elements forming each centrum. For
instance, a monospondylous vertebra (Fig. 2B–F) is
formed by only one centrum. In contrast, a vertebral
centrum formed by more than one element is termed
diplospondylous or polyspondylous (Fig. 2G–I).
Centrum. The vertebral centrum is represented
by mineralized, calcified, or ossified portions that
surround the notochord. The centrum is termed
chordacentrum, arcocentrum, autocentrum, etc., de-
pending on its origin. We define these elements as
follows.
An arcocentrum is formed by ossification of carti-
lage extending from the arcualia around the noto-
chord (in placoderms and teleostomes).
The chordacentrum may have different origins.
The chondrichthyan chordacentrum is formed by
cartilaginous cells that migrate from the arches into
the fibrous sheath of the notochord that grows into a
typical amphicoelous cartilage. It is usually covered
by a thin calcification. The actinopterygian chorda-
centrum is formed as mineralization (e.g., calcifica-
tion) in the middle fibrous part of the chordal
sheath.
Aholocentrum is formed by proliferation of carti-
lage cells around the elastica externa of the noto-
chord, producing a compact centrum. The layer of
cartilage cells grows into a typical amphicoelous os-
103VERTEBRAL COLUMN IN DIPNOANS
sified centrum that usually has cavities of different
sizes and shapes filled with cartilage (in a variety of
fishes such as some Paleozoic dipnoans and palaeo-
nisciforms).
The autocentrum is a centrum or parts of the
centrum formed outside the elastica externa by di-
rect ossification (in actinopterygians such as †Lep-
tolepis coryphaenoides and more advanced teleosts).
In modern dipnoans the centrum is represented
by the arcocentra (or neural arches) (Fig. 1A–C).
However, autocentra have been mentioned for
Protopterus by Mookerjee et al. (1954) and Bartsch
(1989) for one specimen (but see below). Fossil
dipnoans in which centra are known have arcocen-
tra and one ossified centrum per vertebral seg-
ment.
Postcaudal cartilaginous centrum. A hyaline
cartilaginous centrum that forms as a continuous
unit with the hyaline cartilaginous neural and
haemal arches is only known in postcaudal verte-
brae of living dipnoans. We name this unique struc-
ture the postcaudal cartilaginous centrum. This
kind of structure should not be confused with the
centrum present in elasmobranchs (see below).
Fig. 1. Nomenclature of dif-
ferent components of vertebrae
and associated elements in cross
sections in dipnoans. Diagram-
matic sections of an anterior
abdominal (A), a posterior ab-
dominal (B), and a caudal
(C) vertebra.
104 G. ARRATIA ET AL.
Supradorsal cartilage. A small rod of cartilage
(or fibrocartilage in large specimens) positioned be-
tween both distal ends of the arcocentrum or neural
arch and the proximal end of the neural spine is
identified here as the supradorsal cartilage in dip-
noans (Fig. 1A,B) by comparison with a similar ele-
ment present in extant actinopterygians.
Neural spine. An independent, median elongate
bone that articulates or fuses ventrally with the
neural arch and articulates dorsally with the supra-
neural in the abdominal region, but is fused to the
neural arch in the caudal region; the distal portion
of the neural spine of the caudal vertebrae also
articulates with a ‘supraneural’ in dipnoans. The
neural spines lie in the median, epaxial skeletoge-
nous septum caudad to the posterior margin of the
cranium.
The element identified here as neural spine (Fig.
1A–C) was named by Gu¨ nther (1871); the name is in
current use. Fu¨ rbringer (1904) used the Latin name
processus spinosus. However, the dorsal part of the
structure that we term neural arch was named neu-
ral spine by Campbell and Barwick (1988) in
†Griphognathus and proximal supraneural by Ahl-
berg and Trewin (1995) in †Dipterus. Our develop-
mental and comparative studies do not support
these identifications.
‘Supraneural’ and supraneural bones. The
‘supraneurals’ are independent, median, elongate,
rod-like bones that articulate proximally with the
neural spine and distally with the dorsal radial (Fig.
1B,C), or lack articulation distally with another
bone (Fig. 1A). The ‘supraneural’ or supraneural
series of bones lies in the median, epaxial skeletog-
enous septum caudad to the beginning of the verte-
bral column.
The ‘supraneural’ bone has received different
names in the dipnoan literature. It was named in-
terneural 1 by Gu¨ nther (1871) and Flossenstu¨ tzen
(pterygiophores) by Fu¨ rbringer (1904) for Neocera-
todus. It was identified as the (proximal) median
radial by Goodrich (1909) for Protopterus annectens.
It was named the proximal radial by Cloutier (1996)
for †Scaumenacia and †Fleurantia and distal supra-
neural by Ahlberg and Trewin (1995) for †Dipterus.
Epineural spine was used by Andrews and Westoll
(1970a) and Andrews (1977) for rhipidistians and
the coelacanth Latimeria, respectively.
Fig. 2. Types of vertebrae.
A: Aspondylous vertebra.
B,C,E–I: Dysospondylous verte-
brae. D: Holospondylous verte-
bra. B–F: Monospondylous ver-
tebrae. G–I: Diplospondylous
vertebrae; auc, autocentrum; bv,
blood vessels; chc, chordacen-
trum; ha (varc), haemal arch
(ventral arcocentra); h.chc,
hemichordacentrum; na (darc),
neural arch (dorsal arcocentra);
nc, spinal cord; no, notochord.
105VERTEBRAL COLUMN IN DIPNOANS
The ‘supraneural’ bone originates by distal growth
of the cartilaginous arcuale as a continuation of the
neural spine, from which it detaches as a separate
bone in ontogeny in dipnoans and other sarcoptery-
gians. In contrast, the supraneural bone develops
from an independent cartilage, dorsal to the neural
spine in actinopterygians.
Interhaemal or infrahaemal bone. The inter-
haemal bone is an independent, median elongate
rod-like element that articulates with the haemal
spine proximally and the ventral radial distally (Fig.
1C). The infrahaemal series of bones lies in the
median, hypaxial skeletogenous septum caudal to
the beginning of the caudal region.
This bone was identified as interhaemal 1 by
Gu¨ nther (1871) in Neoceratodus and haemal spine
by Fu¨ rbringer (1904), who identified the long
haemal arch plus haemal spine of Neoceratodus as
one simple element, the haemal arch. Goodrich
(1909), Mookerjee et al. (1954), and Bartsch (1989)
did not name the haemal elements of Protopterus.It
was named epihaemal spine by Andrews and
Westoll (1970a) and Andrews (1977) in rhipidistians
and the coelacanth Latimeria, respectively. In gen-
eral fish terminology, it was named apophyses
e´ pineuses infe´ rieures (⫽lower or distal spinous ap-
ophyses) by Cuvier (1829) and as Unterflossentrae-
ger (⫽infrahaemal bone) by Stannius (1839).
Pterygiophore or radial. This is an indepen-
dent, median, rod-like bony element on which the fin
rays attach (not articulate) (Fig. 1B,C). The so-called
dorsal radial of current literature is the bone to
which dorsal and caudal epiaxial fin rays are at-
tached. The series of dorsal radials lies in the me-
dian, epaxial skeletogenous septum caudad to about
half the length of the body. This bone was identified
first as interneural 2 or upper interneural by
Gu¨ nther (1871) in Neoceratodus, and an element
supporting rays or radial by Fu¨ rbringer (1904). It
was named (distal) median radial by Goodrich
(1909).
The so-called ventral radial of current literature is
the bone to which the ventral or anal and caudal
hypaxial fin rays are attached. The series of ventral
radials lies in the median, hypaxial skeletogenous
septum along the caudal region of the body.
Postcaudal cartilage. This element is part of
the postcaudal region of modern dipnoans and is
formed by a hyaline cartilaginous centrum (see
above) that may bear rudimentary neural and
haemal arches and spines. It belongs to the series of
elements (without name) identified as “last elements
of the confluent neural and haemal elements” by
Gu¨ nther (1871:561).
This element was called radial by Goodrich (1909),
who used the name radial to explain that in modern
dipnoans the ventral caudal lobe is supported by
separate radials, and not by rigid unjointed haemal
spines. There is no support for this terminology be-
cause there is no difference between dorsal and ven-
tral lobes of the caudal fin in modern dipnoans. The
structure of the postcaudal cartilages is unique to
modern dipnoans and has no similarities to the ra-
dials in other fishes such as actinopterygians. Bar-
tsch (1989) named the series of postcaudal cartilages
of modern dipnoans the urostyle, but there is no
similarity to the urostyle of teleosts.
Fin rays. The fin rays in Neoceratodus were
called dermoneurals (dorsal rays) and dermo-
haemals (ventral or anal rays) by Gu¨ nther (1871).
However, the fin rays in modern dipnoans were
given a different name, camptotrichia, by Goodrich
(1904, 1909). According to this author, the campto-
trichia become fibrous and little calcified, somewhat
resembling the ceratotrichia of elasmobranchs. They
are formed of a bony substance containing osteo-
cytes.
Our observations of serial cross sections of fin rays
of modern dipnoans differ from those of Goodrich
(1904, 1909). Nevertheless, we agree that the rays
have a structure different from lepidotrichia found
in fossil dipnoans and other fishes such as acti-
nopterygians (see below and description of rays in
living dipnoans and Discussion). We use the term fin
rays in a general way because fossil and living dip-
noans have fin rays that are formed differently, as
explained below.
According to Goodrich (1904), sarcopterygian and
actinopterygian fins are supported by different skel-
etal elements, the dermotrichia, which are the so-
called actinotrichia, the camptotrichia, and the lepi-
dotrichia. Another type of fin ray is the
ceratotrichia.
Dermotrichia: This general name has been used
for the fin rays in acanthodians (Miles, 1970; Zidek,
1975, 1976). However, Reis (1896), Watson (1937),
and Heyler (1962) named them ceratotrichia. The
proximal portions of these unjointed dermotrichia
are ossified and their distal unossified portions are
occasionally observed. However, according to Ge´r-
audie and Meunier (1980) the distal unossified por-
tions might be actinotrichia.
Actinotrichia: These rays are known in some re-
cent dipnoans (see below Figs. 14A, 15A,B; and Ge´r-
audie, 1984; contra Goodrich, 1904, and Ge´ raudie
and Meunier, 1982), in the extant actinistian Latim-
eria (Ge´ raudie and Meunier, 1980), and in acti-
nopterygians. They are the main support of the fin-
folds in young stages and the most distal supporting
elements in adults. They were given the name by
Ryder (1886). They are elastoidin-formed elements.
Camptotrichia: These rays are known only in re-
cent dipnoans. They are rays whose fine structure
more resembles lepidotrichia than actinotrichia.
Camptotrichia are formed by two different regions: a
superficial region of typical acellular fibrous bone
and a deep region continuous with the previous. The
deep region does not constitute a typical bony tissue
because it lacks calcification according to Ge´ raudie
and Meunier (1982:327). According to our observa-
106 G. ARRATIA ET AL.
tions, three regions can be distinguished in Pro-
topterus: the two described by Ge´ raudie and
Meunier and a third one included in the internal
part which may correspond to the enclosed acti-
notrichia (see below Fig. 16A,B).
Lepidotrichia: These rays are known in fossil dip-
noans, in actinistians, and in fossil and recent acti-
nopterygians. Lepidotrichia are dermal rays com-
posed of two hemisegments on which the
actinotrichia overlap distally and internally. Com-
monly, they are segmented; however, certain lepi-
dotrichia lack segmentation (see below for discus-
sion).
Ceratotrichia: These fin rays are known from fos-
sil and living elasmobranchs (e.g., Goodrich, 1904;
Dean, 1909; Zangerl, 1973, 1981; Bendix-Almgren,
1975) and holocephalians. They are formed by large
collagen fibers that develop in bilateral rows within
the dermis. Surrounding each ceratotrichium is a
layer of peritrichial fibroblasts containing secretory
vesicles. Ceratotrichia grow by apposition of colla-
gen fibrils from the peritrichial matrix (Kemp,
1977). According to Ge´ raudie and Meunier (1980),
the fine structure of the ceratotrichia is reminiscent
of that of actinotrichia. The ceratotrichia of elasmo-
branchs were proposed as homologous to the acti-
notrichia by Bouvet (1974) and Kemp (1977).
Lepidosiren
The vertebral column (Fig. 3) of Lepidosiren para-
doxa comprises the abdominal (formed by vertebrae
with neural arches, but without haemal arches), the
caudal (vertebrae with neural and haemal arches),
and the postcaudal (formed by a series of postcaudal
cartilages) regions. Eighty-five to 90 serial elements
form the vertebral column of adult specimens; 55–59
form the abdominal region (the first neural arch
included in the neurocranium is not counted; see
below Fig. 13), 28–32 the caudal region, and 15 or 16
cartilaginous elements form the postcaudal region.
There are differences in structure and relationship
among vertebrae of these regions.
Vertebral centra, either autocentra or chordacen-
tra, are absent in all stages of growth; therefore, the
notochord is the only support for the dorsal and
ventral arcocentra and associated elements (e.g.,
neural and haemal spines, ‘supraneurals,’ interh-
aemals or infrahaemals, and dorsal and ventral ra-
dials).
Abdominal and caudal vertebrae of adult Lepido-
siren are represented by ossified elements retaining
cartilaginous regions dorsal and ventral to the noto-
chord (Figs. 1A–C, 4A–C).
The ossified paired dorsal vertebral elements form
the neural arches (Figs. 1A–C, 4B, 5A,B). There are
85–90 neural arches in total; their ventral parts
partially surround the notochord and correspond to
the arcocentra. Each arcocentrum bears ventrally a
broad cartilage, a remnant of the arcuale. The dorsal
parts of the neural arches remain unfused medially;
nevertheless, they are united by a single, median
cartilage that also articulates with the ossified neu-
ral spine in the abdominal region. In the caudal
region, right and left arcocentra fuse with the neural
spine to form a single element. The neural spines
articulate with the ‘supraneural’ bones along the
vertebral column and the ‘supraneurals’ articulate
with the dorsal radials or dorsal pterygiophores
from the middle region of the vertebral column pos-
teriorly (Figs. 4C, 5A,C,D). The anteriormost dorsal
fin rays (camptotrichia) lie above. They do not artic-
ulate with the small cartilaginous radials at the
distal ends of the ‘supraneurals’ (Fig. 4A). Posteri-
orly, the last abdominal vertebrae are associated
with ‘supraneurals’ articulated with ossified radials
(Fig. 5A,C,D) that are connected to dorsal rays. The
neural arches or dorsal arcocentra do not bear
epineural processes nor epineural bones.
The ossified paired ventral vertebral elements
form the small ventral arcocentra that are fused
with short and sharp ribs in the abdominal region
(Figs. 1A,B, 4A,B). Each ventral arcocentrum bears
a broad cartilage, a remnant of the arcuale. The
well-developed haemal arches fuse with the median
single elements or haemal spines in the caudal re-
gion (Figs. 1C, 4C). In the caudal region the ventral
arcocentra develop from ventral arcualia. The
haemal arches partially surround the caudal blood
vessels and the notochord. They fuse with the
haemal spine in early ontogeny. The haemal spines
articulate with the interhaemal (infrahaemal) bones
along the caudal region and the interhaemal bones
articulate with the ventral radials (ventral pterygio-
phores) (Figs. 1C, 4C).
Notochord. The notochord extends along the
length of the body, except for the short caudal end
(here called the postcaudal region) (Figs. 3, 6) in
small specimens. In contrast, it ends in front of a
long postcaudal region in larger individuals (Fig.
Fig. 3. Lateral view of vertebral column and associated elements of Lepidosiren paradoxa
(specimen of 43.6 mm total length; PC 1996.1).
107VERTEBRAL COLUMN IN DIPNOANS
7A,B). It is a cylinder filled with cells belonging to
the notochordal epithelium and is surrounded by
three notochordal sheaths, the elastica interna, the
elastica media or fibrous sheath, and the elastica
externa. The notochord does not present deep in-
vaginations along its length; however, their outlines
can be observed occasionally.
According to Kerr (1909:11), “By stage 14 in Lepi-
dosiren the mesoderm has become marked off on
each side by a split, the notochordal rudiment now
forming a ridge-like projection of the archenteric
roof. About stage 16 the chorda becomes split off
from the thin layer of endoderm beneath it which
persists as the enteric roof. By about stage 23 the
notochord has become cylindrical and a delicate pri-
mary sheath is formed on its surface. About stage 32
the secondary sheath makes its appearance. It rap-
idly increases in thickness and about stage 36 begins
to be colonized by immigrant amoeboid cartilage
cells from the arcualia, becoming eventually con-
verted into a continuous cylinder of cartilage.”
Later in development, the notochordal epithelium
(Fig. 8A,B) bears large polyhedric cells with large
eccentric nuclei with numerous granules; the inner
cells are bigger than those at the periphery. No
vacuoles are observed in the notochordal epithelium.
In a specimen of 36 mm preanal length, the noto-
chordal sheaths are partially formed at the most
anterior part of the notochord and the elastica ex-
terna appears incomplete in cross sections. Attached
to the elastica interna there are several nuclei be-
longing to the notochordal epithelium.
The elastica externa is a conspicuous, refringent
sheath that surrounds the notochord. This mem-
Fig. 4. Lateral view of ele-
ments forming the vertebrae
and associated structures in
Lepidosiren paradoxa (specimen
of 400 mm total length; PC
1996.3). A: Abdominal vertebrae
32–36. B: Diagrammatic cross
section of one abdominal cen-
trum and its associated ele-
ments. C: Caudal vertebra 67;
bv, blood vessels; ?c.ra, cartilag-
inous radial?; da.pr, anterior
dorsal process; dra, dorsal radial
or dorsal pterygiophore; d.ry,
dorsal fin ray; ha⫹hs, haemal
arch plus haemal spine forming
one element; iha, interhaemal
element or infrahaemal element;
na, neural arch; na⫹ns, neural
arch plus neural spine forming
one element; no, notochord; nc,
spinal cord; ns, neural spine;
r.da, remnant of dorsal arcuale;
ri, rib; r.va, remnant of ventral
arcuale; ‘su,’ ‘supraneural;’
?sud.c, supradorsal cartilage?;
va.pr, anterior ventral process;
vra, ventral radial or ventral
pterygiophore; var⫹ri, ventral
arcuale at base of rib and rib.
108 G. ARRATIA ET AL.
brane is complete except for perforations (Figs. 8B,
9A,D). The perforations of the elastica externa are of
different diameters and occur in variable number at
the base of neural and haemal arches along the
whole notochord, but there are not more than four
perforations per side at the base of an arch. Carti-
laginous cells invade the fibrous sheaths or the elas-
tica media through these perforations. Three areas
with different distributions of perforations of the
elastica externa can be distinguished: 1) The most
anterior one is below the neurocranium; in this re-
gion only the laterodorsal portion of the elastica
externa presents perforations. 2) In the abdominal
region, behind the neurocranium, perforations of
the elastica externa are found where the bases of the
dorsal arches attach to the elastica externa. 3) In the
caudal region the elastica externa has perforations
where both neural and haemal arches attach to the
elastica externa.
We have never observed perforations of the elas-
tica externa in regions between arches, even though
cartilaginous cells are found between two arches
inside the elastica media. We interpret the presence
of cartilaginous cells between two arches as the re-
sult of migration of cells. This migration inside the
elastica media apparently follows a helicoidal move-
ment, due to the position that the cells have along
the fibrous sheath of the notochord.
The elastica media or fibrous sheath (Figs. 8B,
9A–C) is the thickest of the three sheaths. This
sheath increases its thickness tremendously dur-
ing ontogeny but never produces chordacentra. In
the anterior region of the notochord the fibers of the
elastica media react differently to staining of the
Picro-Ponceau method and show three bands, e.g.,
two marginal, strongly pigmented bands and an al-
most whitish middle one. Caudad, the elastica me-
dia becomes homogeneous in aspect. In addition to
the fibers, the elastica media (Figs. 8B, 9A–D) bears
cartilaginous cells that have migrated from the car-
tilage of the arcualia positioned outside the elastica
externa.
The elastica interna is placed between the noto-
chordal epithelium and the elastica media. It is thin-
ner than the elastica externa and is difficult to ob-
serve in cross sections of young specimens. It stains
usually darker than the elastica media with the
Picro-Ponceau method.
The beginning of notochord formation is indicated
by a center of epithelial and mesenchymal cells.
Once the notochord appears in the head region, it is
framed laterally by two broad masses of hyaline
Fig. 5. Lateral view of ele-
ments forming the vertebrae and
associated structures of Lepidosi-
ren paradoxa.A: Abdominal ver-
tebra 43. B–D: Diagrammatic
composition of the neural or dor-
sal region of an anterior (B), pos-
terior abdominal (C), and caudal
vertebra(D);dp.na, dorsalprocess-
(ess) of the neural arch; dra, dor-
sal radial or dorsal pterygiophore;
na, neural arch; na⫹ns,neural
arch⫹neural spine; no, notochord;
ns, neural spine; ‘su,’ ‘supraneu-
ral;’ ?su.c, supradorsal cartilage?;
var⫹ri, ventral arcuale at base of
rib and rib.
109VERTEBRAL COLUMN IN DIPNOANS
cartilage, the occipital cartilage. A thin mass of con-
nective tissue is present between the parasphenoid
and the notochord. At this point the notochord is
oval in shape and the elastica media shows the three
bands mentioned above, at least close to the ventral
and dorsal regions. Cartilaginous cells migrate into
the elastica media through perforations of the elas-
tica externa. These cells originate from the occipital
cartilage.
Posterior to the cranium, the dorsal arcualia begin
to ossify as neural arches (dorsal arcocentra) early
in ontogeny (e.g., specimens of about 40 mm total
length already have thin arcocentra). Shortly after
the arcualia begin to ossify, perforations are ob-
served in the elastica externa.
After the beginning of the ossification of the first
and/or second dorsal arches, ventral arcualia ap-
pear. In the abdominal region the bases of the ven-
tral arcocentra are smaller than the dorsal ones and
the elastica externa is unperforated. Although car-
tilaginous cells are attached to the elastica externa,
these are not able to migrate into the elastica media
because the elastica externa remains complete all
along the abdominal region in front of the ventral
arcualia. In the caudal region the ventral arcualia
become larger than those of the abdominal region
because of proliferation of cartilaginous cells. Thus,
ossification of the ventral arcualia produces well-
developed ventral arcocentra. The elastica externa
exhibits perforations and the cartilaginous cells are
able to migrate into the elastica media.
Dorsal arcualia and dorsal arcocentra or
neural arches. The specimens examined have be-
tween 85–90 neural arches, 55–59 in the abdominal
region, 28–32 in the caudal region, and a variable
number in the postcaudal region. There is the same
number of neural arches, neural spines, ‘supraneu-
rals,’ and radials, except for the postcaudal region,
where the sequence is lost.
The neural arches (Figs. 1A–C, 4A–C) are paired
structures that result from ossification of small and
elongate dorsal arcualia; the dorsal arcualia grow
dorsad and surround the spinal cord. Left and right
arcualia fuse dorsally and continue growing distally
as a single, median cartilage. The distal growth and
differentiation of each dorsal median cartilage into
separate ossifications produce first the neural spine,
second the ‘supraneural’ bone in the anterior abdom-
inal region, and finally the dorsal radial in the mid-
caudal region (Figs. 10A–C, 11A,B). These single
Fig. 6. Caudal and postcaudal region of the vertebral column with associated elements and fin rays of Lepidosiren paradoxa
(specimen of 140 mm total length; PC 1996.2); da.pr, anterior dorsal process; dra, dorsal radial or dorsal pterygiophore; d.ry, dorsal
fin ray; ha⫹hs, haemal arch ⫹haemal spine; iha, interhaemal element or infrahaemal element; na⫹ns, neural arch ⫹neural spine;
no, notochord; ptc.c, postcaudal cartilage; ‘su,’ ‘supraneural;’ va.pr, anterior ventral process; vra, ventral radial or ventral pterygio-
phore; v.ry, ventral fin ray.
110 G. ARRATIA ET AL.
elongate cartilages are constricted at certain levels
where joints will form. Early in ontogeny, the artic-
ular surfaces between the above-mentioned ele-
ments are formed mainly by hyaline cartilage, but
through growth the hyaline cartilage is replaced by
thick and well-developed fibrocartilaginous articu-
lar surfaces (compare Figs. 4A,C, 5A–D, and 10A–C,
11A,B).
Dorsal interarcual (⫽intercalar) cartilages were
only observed in the caudal region in one specimen
of 96.0 mm preanal length, having three small car-
tilages. They consist of hyaline cartilage and are
attached dorsally to the notochord, between two
basidorsal arcualia or two neural arches.
As stated above, each neural arch (arcocentrum) is
formed by a pair of ossified, somewhat expanded
bases that fuse medially to form a median, slightly
elongate dorsal process, and two membranous out-
growths, one rostral and another caudal, that
project anteriad and posteriad from the median pro-
cess. In the abdominal region the rostral membra-
nous outgrowth is larger than the caudal one (Figs.
4A, 5A). Close to the end of the abdominal region
both anterior and posterior outgrowths are almost
the same size, and in the caudal region, once the
neural arch is fused with the neural spine, the cau-
dal outgrowth is larger than the rostral one. In spec-
imens larger than 53.5 mm preanal length, the first
and sometimes the second arch lack the mentioned
membranous outgrowths or they are small. In spec-
imens between 43–60 mm total length, the out-
growths are still not developed (Fig. 10A,B). The
rostral membranous outgrowth may bear a foramen
placed closer to its anterior edge. Sometimes the
foramen is incompletely enclosed; thus, it is only
represented by a notch at the anterior margin of the
outgrowth. The distribution of this foramen along
the vertebral column is variable. It is usually
present from the 3rd or 4th to the 43rd arch, and is
lacking in some arches in between.
Fig. 7. Caudal and postcaudal region of the vertebral column with associated elements and fin rays of Lepidosiren paradoxa.
A: Specimen of 400 mm total length (PC 1996.3). B: Caudal end of the notochord and its relationships to the postcaudal cartilages
(specimen of 180 mm total length; PCJC1990.1); c.dra, cartilaginous dorsal radial; c.ha, cartilaginous haemal arch or ventral arcuale;
chb, chondral element; dra, dorsal radial or dorsal pterygiophore; d.ry, dorsal fin ray; ha⫹hs, haemal arch ⫹haemal spine; iha,
interhaemal element or infrahaemal element; iha⫹ra?, cartilage from which the interhaemal and the ventral radial originate?; na,
neural arch; no, notochord; ns, neural spine; ptc.c, postcaudal cartilage; ‘su,’ ‘supraneural;’ ‘su’⫹ra, ‘supraneural’ plus cartilaginous
radial still forming one element; vra, ventral radial or ventral pterygiophore; v.ry, ventral fin ray.
111VERTEBRAL COLUMN IN DIPNOANS
Although the neural arches are well ossified all
along the body, the basal portion of the arcocentra
remains cartilaginous even in specimens up to 124
mm preanal length. The quantity of cartilage is
larger in the posterior abdominal region and in the
caudal region than in the anterior abdominal region.
In specimens of about 400 mm preanal length, the
neural arches of the abdominal region are appar-
ently totally ossified and remnants of the dorsal
arcualia are not evident at their ventral margin in
cleared and stained specimens. However, cross sec-
tions show vestiges of the dorsal arcualia (Fig. 9A,D)
almost in continuation with the elastica media of
the notochord. The neural arches have irregular
forms, usually with an anterior projection that
does not bear cartilage as does the anterior pro-
cess of the caudal region. Specimens between
43–60 mm preanal length bear some of the most
posterior neural arches as cartilage (Fig. 10D).
These observations point out that the direction of
ossification of the neural arches develops rostro-
caudad. In addition, the depth of the neural arches
gradually decreases from the abdominal to the
caudal region.
The neural arches of the whole vertebral column
are articulated with a single neural spine in our
smallest cleared and stained specimen (43.6 mm
total length). The neural arches of the abdominal
region and the first two caudal arches bear a neural
spine articulated to their dorsal region (median pro-
cess). However, the third neural arch of the caudal
region fuses early in ontogeny with the neural spine
(in specimens of 60 mm total length or more). This
fusion appears between neural arches 57 and 60
(Fig. 10C). At the first arch with such fusion a con-
spicuous anterior process develops. This process is a
short, rod-like structure, placed parallel to the noto-
chord. From the very beginning this anterior process
is ossified, except for its anteriormost tip, which
remains cartilaginous.
The most posterior neural arches (Figs. 6, 7A) of
the caudal region, in addition to being smaller in
size, may present an additional dorsally directed
process between the anterior process of the neural
Fig. 8. Lepidosiren paradoxa of 50 mm total length. Cross section (PC 1996.5) through the anterior half of the body to illustrate the
composition and relationships of an abdominal vertebra (A) and enlargement of neural arches and notochord and its sheaths
(B). Arrows point to perforations of the elastica externa. Picro-Ponceau staining technique; bv, blood vessel; c.ha, cartilaginous haemal
arch or ventral arcuale; c.na, cartilaginous neural arch or dorsal arcuale; ee, elastica externa of notochord; fsh, fibrous sheath of
notochord; nc, spinal cord; no, notochord.
112 G. ARRATIA ET AL.
Fig. 9. Lepidosiren paradoxa of 220 mm total length (PC 1996.6). A: Cross section of the anterior half of the body to illustrate the
composition of the dorsal part of an abdominal vertebra and its relationships to surrounding structures. B: Ventral part of the
notochord and its sheaths and remnant of the ventral arcuale surrounded by the ossification of the base of the rib. C: Enlargement of
the ventral arcuale surrounded by the ossification at the base of the rib (indicated by arrows). D: Enlargement of the region with a large
perforation of the notochord. Note the continuation of the remain of the dorsal arcuale into the fibrous sheath (indicated by arrows)
and the extension of the cartilaginous cells. Picro-Ponceau staining technique; c.na, cartilaginous neural arch or dorsal arcuale; ee,
elastica externa of notochord; fsh, fibrous sheath of notochord; na, neural arch; nc, spinal cord; var⫹ri, ventral arcuale at base of rib
and rib.
arch and the neural spine. In the postcaudal region,
the regularity presented along the vertebral column
is lost and a kind of “disarray” is observed, e.g.,
neural spines may articulate with the anterior pro-
cess of the next neural arch (compare Figs. 10A–D,
6, and 7A).
Neural spines. The independent neural spines of
the abdominal region (Figs. 1A, 4A, 5A) are rod-like,
elongate bones that are slightly constricted at their
mid-length in adults. They are positioned dorsal to
the dorsal median processes of the neural arches in
adults. The neural spines fuse to the neural arches
in the caudal region (Figs. 4C, 5D, 6, 10C). The
length of the neural spines increases from the 1st to
the 4th vertebra in the abdominal region. From this
point on, up to the beginning of the caudal region,
Fig. 10. Vertebral column and associated structures of Lepidosiren paradoxa (specimen of 43.6 mm total length; PC 1996.1).
A: Anteriormost vertebral elements. B: Middle section of the abdominal region. C: Abdominal–caudal region. D: Posteriormost caudal
and postcaudal regions; c.dra, cartilaginous dorsal radial; c.ha, cartilaginous haemal arch or ventral arcuale; c.na, cartilaginous neural
arch or dorsal arcuale; c.vra, cartilaginous ventral radial; da.pr, anterior dorsal process; d.ry, dorsal fin ray; fu.l, fusion line; ha⫹hs,
haemal arch plus haemal spine forming one element; iha, interhaemal element or infrahaemal element; na, neural arch; na⫹ns, neural
arch plus neural spine forming one element; nc, spinal cord; no, notochord; ns, neural spine; ns⫹‘su’, neural spine plus supraneural
still forming one element; ptc.c, postcaudal cartilage; r.da, remnant of dorsal arcuale; r.va⫹ri, remnant of ventral arcuale plus rib; ‘su,’
‘supraneural;’ ‘su’⫹c.ra, ‘supraneural’ plus cartilaginous radial still forming one element; va.pr, anterior ventral process; v.ry, ventral
fin ray.
114 G. ARRATIA ET AL.
Figure 10. (Continued.)
115VERTEBRAL COLUMN IN DIPNOANS
Fig. 11. Dorsal part of the vertebral column and associated structures of Lepidosiren paradoxa (specimen of 43.6 mm total length;
PC 1996.1; cl. and st.). A: Middle part of the abdominal region. B: Posterior part of abdominal region. Black arrow points anteriorly.
White surrounded black arrows point to regions where cartilage continues to form between bony elements; c.dra, cartilaginous dorsal
radial; na, neural arch; nc, spinal cord; no, notochord; ns, neural spine; ‘su,’ ‘supraneural.’
116 G. ARRATIA ET AL.
the length of spines remains almost the same. In the
caudal region the neural spines become gradually
shorter and more inclined ventroposteriorly (Figs.
6, 7A).
As with the neural arches, the direction of ossifi-
cation of the series of neural spines is rostrocaudad.
The ossification process in each spine progresses
from the middle region dorsad and ventrad, simul-
taneously. The neural spines are already ossified in
specimens between 43 and 60 mm total length.
‘Supraneural’ bones. The ‘supraneurals’ are
placed dorsal to the neural spines (Figs. 1A–C, 4A,C,
5A,B). Dorsal radials follow in the posterior abdom-
inal and caudal regions (Figs. 4C, 5A,C,D, 6, 7A). A
‘supraneural’ appears at the level of the first neural
arch in three specimens, but at the second neural
arch in 12 specimens. The length of the ‘supraneu-
rals’ increases considerably from the first to the
fifth; however, from the sixth ‘supraneural’ the
length increases gradually up to the posterior caudal
region where the ‘supraneurals’ become shorter.
The ‘supraneurals’ are very similar in shape to the
neural spines. However, they can be longer than the
neural spines. Each possesses a well-developed dis-
tal cartilage (Fig. 4A) in the anterior abdominal
region.
Formation and ossification of ‘supraneurals’ fol-
lows two directions, starting at the beginning of the
caudal region in anterior and posterior directions
(Fig. 10B,C). In specimens of 43–60 mm total length
the first ‘supraneural’ is not yet formed and the
second and third are cartilaginous. In a specimen of
about 60 mm total length the first ‘supraneural’ is
still cartilaginous.
A ‘supraneural’ is not separated from the second
and third neural spines in the caudal region of a
specimen of 43 mm total length (Fig. 10D) and they
are absent at the last three formed neural spines
(Fig. 6). They do not appear as separate elements in
the postcaudal region of larger specimens (Fig.
7A,B).
Dorsal radials. Relatively small, elongate ele-
ments, always associated with the lepidotrichia and
distal to the ‘supraneurals,’ are identified here as
dorsal radials or pterygiophores (Figs. 1B,C, 4C,
5A,C,D). The appearance of the cartilaginous and
ossified dorsal radials is variable. The first dorsal
radial appears at the level of neural arches 28–32.
The number of radials is highly variable, depending
on the size of the specimen. It may vary between 47
and 68 bony elements.
The dorsal radials gradually increase in length up
to the beginning of the posterior half of the caudal
region; further posterior they become smaller (Figs.
3, 6, 10B,C). The posteriormost dorsal radials can be
still cartilaginous in specimens of about 400 mm
preanal length, where they are irregular in pres-
ence, shape, and size (Fig. 7A,B).
In specimens of 36 and 59 mm preanal length, the
ossification center of each dorsal radial is at the
mid-length of the cartilage from which the radial
originates. The beginning of ossification of the dorsal
radials is variable from specimen to specimen and it
seems that it is not strictly associated with the size
of the animal (Fig. 12).
Ventral arcualia and ventral arcocentra or
haemal arches. In the abdominal region, the ven-
tral arcocentra develop from the arcualia placed
ventrolateral to the notochord. The first ventral ar-
cocentrum is placed at the level of the third or fourth
neural arch (dorsal arcocentrum). In specimens
43–60 mm total length, the first pair of ventral
arcocentra of the abdominal region is already ossi-
fied. However, those situated posteriorly remain car-
tilaginous. Each is represented by a slightly rounded
and expanded base enclosing a small node of carti-
lage and each is continuous with the rib (Figs. 4A,B,
10A). Cartilage is still present in the posterior ab-
dominal ventral arcocentra in specimens of 124 mm
preanal length. The ossification process of the ven-
tral arcocentra follows a rostral-caudal direction
similar to that of the neural arches.
In the caudal region the dorsal arcocentra (neural
arches) and the ventral arcocentra (haemal arches)
closely resemble each other in shape and size (Figs.
1C, 4C, 6, 7A). The site of contact between the
haemal arches and the notochord is placed ventrally
in the caudal region; in contrast, the ventral arches
of the abdominal region have a more lateral position.
In the caudal region the size of the haemal arches
decreases caudad and becomes very irregular. In
specimens between 43–60 mm total length the last
five to eight haemal arches remain cartilaginous
(Fig. 10D).
The two first haemal arches are fused to the
haemal spines even in the smallest specimens exam-
ined (Fig. 10C). Contrary to the condition in the
Fig. 12. Plots of the beginning of the ossification of the dorsal
radials versus preanal length in young individuals of Lepidosiren
paradoxa. Note that the beginning of ossification is not strictly
correlated with size.
117VERTEBRAL COLUMN IN DIPNOANS
neural arches, the two first haemal arches lack the
anterior process that is present in the dorsal region.
The anterior process occurs from the third arch pos-
teriorly. There is no difference between this process
and the one present in the neural arches.
Ribs. There is a total of 54–57 pair of ribs asso-
ciated with the vertebral column (⫽rib below) that
develop directly from very small, rounded, or almost
rounded ventral arcualia, plus a large, well-ossified
rib (⫽cranial rib) that articulates with the cranium.
In specimens between 43–60 mm total length, ribs
are already ossified; however, a remnant of the arc-
ualia is found at the base of each rib (Figs. 9B,C,
10A–C), even in cross sections of large specimens.
The ribs are short structures that curve ventrally.
They are extensions of the ventral arcocentra, not
separated from the latter. The first pair of ribs is
slightly thinner than the others and is situated in
front of the third and fourth neural arches in the
smallest specimens (Figs. 3, 10A). In larger speci-
mens the first rib corresponds to the second vertebra
(Fig. 13). The first rib is missing (Fig. 13). The cra-
nial rib is a large, well-developed bone that articu-
lates with the cranium, and it is not associated with
any neural arch or parapophyses (Fig. 13; Dalquest
et al., 1989: fig. 7).
Fu¨ rbringer (1904), studying the ribs and their
relationships to the cranium and neural arches, con-
cluded that the cranial rib is the third rib in Neo-
ceratodus, because he found vestiges of two cartilag-
inous arches that would correspond to vertebrae
whose arches and ribs are included within the cra-
nial wall or are lost. We have not found vestiges of
anterior neural arches and/or spines in Lepidosiren
to support this interpretation in Neoceratodus (but
see Protopterus below).
Interhaemal or infrahaemal elements. The
interhaemal elements are similar in morphology to
the ’supraneurals’ of the dorsal region (Figs. 1C, 4C,
6, 7A). They are mirror images of each other. In front
of the first haemal arch there can be up to four free
interhaemal elements positioned below and between
the posteriormost ribs (Fig. 10C). The interhaemal
elements articulate with the distal tip of the haemal
spines and the ventral radials. The length of the
articulated interhaemal elements decreases cau-
dally, although the first free elements can be shorter
than the others. Ossification follows a rostrocaudal
direction; however, the first free elements can be
completely cartilaginous.
In one of the smallest specimens, an elongate car-
tilage appears separately from the first haemal
spine, the interhaemal element plus the radial
forms independently (Fig. 10C). Also, the second and
third haemal arch plus spine are separated from the
interhaemal elements. The fourth haemal spine ar-
ticulates with the interhaemal element.
Ventral radials. The ventral radials develop in
early ontogeny from the distal cartilaginous portion
of the interhaemal (Fig. 10C); therefore, the first
median cartilage is considered a complex structure
from which both the future interhaemal and ventral
radial will form. After their differentiation, the ra-
dials articulate with the distal articular surface of
the interhaemal elements and are associated with
the ventral rays (⫽ventral camptotrichia, see below;
Figs. 6, 7A, 10C).
There can be up to four free cartilaginous radials
in front of the first ventral arch. The ventral radials
are longer forward than backward; however, the free
elements, when they are present, can be shorter.
Because there are cartilaginous elements anteriorly
Fig. 13. Anterior vertebrae and
associated elements of Lepidosiren
paradoxa (216 mm total length; PC
1996.4) to illustrate relationships of
anterior ribs to neural arches. The
counts in brackets follow the num-
bering proposed by Fu¨rbringer
(1904) for Neoceratodus; na2-5, neu-
ral arch 2-5; ns, neural spine; ri1-5,
rib 1-5; ‘su,’ ‘supraneural.’
118 G. ARRATIA ET AL.
and posteriorly, it is assumed that the ossification
process progresses bidirectionally. Nevertheless,
there are fewer unossified elements posteriad than
anteriad. Their number is highly variable, depend-
ing on the size of the specimen, varying between 27
and 40 bony elements.
Accessory elements. In the caudal region, small
nodules of cartilage dorsal and ventral to the noto-
chord can be present. These nodules occur in spaces
between neural arches and ‘supraneurals’ and in
between haemal arches and interhaemal elements.
Dorsally, a maximum of 12 nodules were found in
one specimen of 123.9 mm preanal length. Ven-
trally, 15 nodules were found in one specimen of
106.6 mm preanal length; one nodule was partially
ossified.
Postcaudal cartilages. Postcaudal cartilages
(Figs. 6, 7A,B, 10D, 14A–C, 15A,B) appear immedi-
ately behind the notochord. They are almost rectan-
gular structures, completely cartilaginous, and be-
come elongate posteriad. The anterior margin of the
first block is concave and the distal tip of the noto-
chord abuts against it (Fig. 7A,B).
The postcaudal region (Figs. 6, 10D) of young
specimens is formed by one elongate cartilage
placed just posterior to the notochord, whereas in
larger individuals the postcaudal region (Fig.
14A–C) is formed by a series of 15 or 16 cartilag-
inous elements which are separated from each
other by narrow spaces. It is assumed that the
postcaudal cartilages grow by distal additions.
The postcaudal cartilages form centra-like ele-
ments to which variable numbers of neural and
haemal elements are added.
The first and only postcaudal cartilage (Figs. 6,
10D) of young individuals lacks arches; however,
they are present in large individuals. The first post-
caudal cartilage (Figs. 7A,B, 14A–C) may present
complete or incomplete neural and haemal arches,
some of which may become ossified. Some postcau-
Fig. 14. The postcaudal region of Lepidosiren paradoxa (specimen of 180 mm preanal length; PCJC 1990.1). A: Lateral view
illustrating the postcaudal cartilages, fin rays, and actinotrichia. B: Last two caudal vertebral segments and first two postcaudal
cartilages. C: Fourth and fifth postcaudal cartilages and their relation to dorsal and ventral cartilages; act, actinotrichia; da.pr,
anterior dorsal process; d.ry, dorsal fin ray; ha⫹iha⫹ra, cartilage from which the haemal arch, interhaemal, and ventral radial
originate?; ha⫹sp, haemal arch plus rudimentary spine; na, neural arch; na⫹sp, neural arch plus rudimentary spine; ptc.c, postcaudal
cartilage; v.ry, ventral fin ray.
119VERTEBRAL COLUMN IN DIPNOANS
Fig. 15. Posterior part of the body of Lepidosiren paradoxa illustrating presence of actinotrichia and fin rays. A: Specimen of 180
mm preanal length (PCJC 1990.1). B: Specimen of 400 mm total length (PC 1996.3). Note the bifurcation of some rays at their bases
and the orientation of the actinotrichia and fin rays caudally. C: Dorsal fin ray showing segmentation; act, actinotrichia; d.ry, dorsal
fin ray; ptc.c, postcaudal cartilage; v.ry, ventral fin ray.
120 G. ARRATIA ET AL.
dal cartilages have two neural arches dorsally and
two haemal arches ventrally. Other have three. Cau-
dally, the cartilaginous neural and haemal arches
are incomplete. The most posterior postcaudal car-
tilages lack arches. Some specimens show that
some of the cartilages were regenerated after par-
Fig. 16. Cross sections of fin rays to illustrate differences between dipnoans and actinopterygians. Arrows point to the halves of a
ray. A,B: Dorsal fin rays of Protopterus sp. Note the region stained differently inside the half-ray (Azan-Mallory staining technique).
B: Detail of a dorsal ray in cross section of Protopterus sp. C: Caudal fin ray of Lepisosteus osseus.D: Caudal fin ray of Amia calva.
(C,D, Heidenhain’s Azan staining technique). E: Caudal fin rays of Elops saurus (Picro-Ponceau staining technique).
121VERTEBRAL COLUMN IN DIPNOANS
Figure 17
tial destruction by mechanical action or predation.
Regenerated postcaudal cartilages do not form
arches.
An elongate median cartilage (Fig. 14B) usually
lies dorsal to the rudimentary neural spine of the
anterior postcaudal cartilages. It is assumed that
the median cartilage represents a ‘supraneural’ plus
a dorsal radial because this cartilaginous element
supports camptotrichia. The most posterior dorsal
median cartilages are free (Fig. 14C). A similar pat-
tern is present ventrally and the elongate median
cartilage is assumed to be an interhaemal plus a
ventral radial.
Serial cross sections show that the notochord does
not continue into the postcaudal cartilages. The end-
ing of the notochord is marked by a transitional area
containing a mixture of notochordal tissue, hyaline
cartilage, and parts of the elastica externa. Posteri-
orly, each postcaudal cartilage is a compact struc-
ture formed only by cartilage up to the end of the
body (see below). The spinal cord lies dorsal to the
cartilage and the blood vessels lie ventrally. The
postcaudal cartilages may produce neural arches
dorsally and haemal arches ventrally, which are also
built only of cartilage. These arches are not the
result of arcual development, but structures origi-
nated from the postcaudal cartilage. We interpret
these to be compound cartilaginous centra because
each postcaudal cartilage (Fig. 14B,C) may have
more than one arch.
In large specimens the postcaudal elements are
hard, dense fibrocartilages (Fig. 7A), except the last
ones, which are formed by hyaline cartilage.
Actinotrichia, rays (camptotrichia), and un-
paired fins. The vertebral column continues to the
end of the “caudal finfold”; consequently, there are
dorsal and ventral actinotrichia and rays (campto-
trichia). In young and adult Lepidosiren, the rays do
not continue up to the end of the body; the most
caudal tip of the body lacks rays (Figs. 6, 7A, 14A,
15A,B). This is quite evident in specimens of about
15–20 cm length. Therefore, a question arises: Is
there a true caudal fin in Lepidosiren? (see Discus-
sion, below).
Early in ontogeny the dorsal-caudal-anal finfold is
supported by actinotrichia, which are replaced by
rays. Specimens of about 45 mm total length show
that ossification starts around the proximal part of
the actinotrichia, but the distal portion is still a
slender rod of collagen. Slender rods of collagen can
be observed irregularly along the finfold in young
and adult individuals. However, cross sections of
camptotrichia show no evidence of actinotrichia in
the element or at its distal end, which suggests that
somehow actinotrichia disappear during growth.
Ossified dorsal and ventral camptotrichia stop a
distance before the end of the body in all specimens.
They are simple, paired elements, completely sepa-
rated from each other, and do not articulate with the
cartilaginous or ossified radials. There are no mor-
phological differences between dorsal and ventral
rays. Cross sections of camptotrichia show that they
are circular, completely separated from each other,
and usually are regularly located, one in front the
other, but separated by extensive mesenchyme (like
the condition present in Protopterus) (Fig. 16A).
Medium-sized specimens show that each camptotri-
chium is formed by acellular bone that is sur-
Fig. 18. Notochord and ribs
in ventral view (A), and lateral
(B) and dorsal view (C) of a rib of
Protopterus amphibius (speci-
men of 556 mm total length;
ZMB 15797); c.bv, canal for
blood vessels; no, notochord; ri,
rib; r.va, remnant of ventral ar-
cuale.
Fig. 17. Elements forming the vertebrae and associated struc-
tures of Protopterus amphibius (specimen of 556 mm total length;
ZMB 15797). A: Abdominal vertebrae 15–17. B: Caudal vertebra
45. C: Diagrammatic cross section of one abdominal vertebra and
its associated elements; ?c.ra, cartilaginous radial?; da.pr, ante-
rior dorsal process; dra, dorsal radial or dorsal pterygiophore;
d.ry, dorsal fin rays; f.ri, articular fossa for rib; ha⫹hs, haemal
arch plus haemal spine forming one element; iha, interhaemal
element or infrahaemal element; na, neural arch; na⫹ns, neural
arch plus neural spine forming one element; no, notochord; ns,
neural spine; ri, rib; r.da, remnant of dorsal arcuale; r.va, rem-
nant of ventral arcuale; ‘su,’ ‘supraneural;’ ?sud.c, supradorsal
cartilage?; va.pr, anterior ventral process; vra, ventral radial or
ventral pterygiophore.
123VERTEBRAL COLUMN IN DIPNOANS
rounded by a layer of cells, the scleroblasts. How-
ever, the camptotrichia show discontinuous regions
along their length: some are ossified but others prob-
ably correspond to unossified areas. That would
agree with the observation that the rays react dif-
ferently to alizarin red along their length, leaving
unstained sections.
There is no close correlation in the number of
camptotrichia and radials. The number of rays per
radial can vary from two to six (Figs. 6, 7). In addi-
tion, the length of the rays can be very irregular in
the postcaudal region. In the largest specimen ex-
amined there is a change in the aspect of some rays;
an irregular “segmentation” appears along some
rays, whereas others show small, bony swellings
(Fig. 15B,C).
In the posterior postcaudal region, young indi-
viduals present a variable number of actinotrichia
(Fig. 15A), whereas the most posterior tip com-
pletely lacks actinotrichia and rays. In the largest
specimen examined (Fig. 15B), there are two se-
ries of actinotrichia at the most posterior tip; the
posteriormost series, with a slightly different ori-
entation, is probably an indication of a late forma-
tion of “caudal” rays. These last actinotrichia are
only present dorsally to the last postcaudal carti-
lages.
Lepidosiren does not have basal fulcra and fring-
ing fulcra associated with any fin. Basal scutes are
also absent.
Protopterus
The vertebral column of Protopterus has been the
subject of numerous studies, such as those by Goo-
drich (1909), Remane (1936), Mookerjee and collab-
orators (for a list of references, see Mookerjee et al.,
1954), and more recently by Bartsch (1989). The
ultrastructure of the notochord has been recently
described by Schmitz (1998).
According to previous information based on young
individuals (e.g., Gegenbaur, 1867; Gu¨ nther, 1871;
Gadow and Abbott, 1895; Schauinsland, 1906; Good-
rich, 1909; Remane, 1936), the vertebral column of
the dipnoans is aspondylous, i.e., it is not differen-
tiated into individual centra. However, Mookerjee et
al. (1954: fig. 1A,B) showed that the vertebral col-
umn of Protopterus is divided into a number of cen-
tra that form short tubular structures without the
constriction in the intravertebral portion. The au-
thors interpreted the vertebral portion of the cen-
trum as bony.
According to Mookerjee et al. (1954), large Pro-
topterus have bony centra, a statement that has not
been accepted and that is not supported by our ob-
servations. In the caudal region of a young cleared
and stained specimen of P. dolloi, Bartsch (1989:
429) observed a series of irregularly shaped struc-
tures that stained red with alizarin and that he
interpreted as perichordal autocentra that develop
in the dorsal region of the centrum. However, an
autocentrum has a characteristic mode of appear-
ance and structure within actinopterygians (see
Fig. 19. Most anterior vertebrae
and associated elements of Pro-
topterus amphibius (specimen of
556 mm total length; ZMB 15797) to
illustrate relationships of anterior
ribs to posterior neurocranium and
neural arches. The counts in brack-
ets follow the numbering proposed
by Fu¨rbringer (1904) for Neocerato-
dus; f.ri, articular fossa for rib;
na1-4, neural arch 1-4; ns, neural
spine; ri4, rib 4.
124 G. ARRATIA ET AL.
Franc¸ois, 1967; Schultze and Arratia, 1988, 1989;
Arratia and Schultze, 1992). Recently, Dr. P. Ba-
rtsch prepared cross sections of the specimen pre-
viously described as having autocentra and ob-
served that these structures are inside the fibrous
sheath of the notochord; consequently, they are
not autocentra. We have been unable to see an
autocentral component in the Protopterus exam-
ined.
According to our observations, the vertebral col-
umn of Protopterus comprises the abdominal, cau-
dal, and postcaudal regions, as in Lepidosiren.
Vertebral centra, either autocentra or chordacen-
tra (Fig. 17A–C), are absent in large specimens;
therefore, the notochord is the only support for the
dorsal and ventral arcocentra and associated ele-
ments. Abdominal and caudal vertebrae of adult
Protopterus amphibius examined are represented
Fig. 20. Cross sections of the postcaudal region of Protopterus sp. (specimen of 30 mm total length; NRM 33876). Azan-Mallory
staining technique. A,B,C: At the level of the first postcaudal cartilage showing progressive caudal views. Note the presence of the
notochordal sheaths and the notochordal matrix and cartilaginous cells. D: Posterior part of first postcaudal cartilage. E,F,G: Posterior
postcaudal cartilages. Note the presence of cartilaginous neural and haemal arches in continuation with the cartilaginous centra
(D,E,F); bv, blood vessels; c.ha, cartilaginous haemal arch or ventral arcuale; c.na, cartilaginous neural arch or dorsal arcuale; ee,
elastica externa of notochord; nc, spinal cord.
125VERTEBRAL COLUMN IN DIPNOANS
only by ossified elements with cartilaginous centers
dorsal and ventral to the notochord, as in Lepidosi-
ren (compare Figs. 17A–C and 4A–C).
The ossified paired dorsal vertebral elements form
the neural arches (Fig. 17C). Their ventral parts
partially surround the notochord. Usually the neu-
ral arches follow each other so closely that the small
anterior process of the neural arch laterally overlaps
the neural arch of the anterior vertebra or overlaps
the articular cartilage placed between the neural
arch and the neural spine of the anterior vertebra.
The dorsal parts of the neural arches remain un-
fused medially. As in Lepidosiren, they are joined by
a single median cartilage that also contacts the neu-
ral spine in the abdominal region. In the caudal
region, right and left arcocentra fuse with the neural
spine, producing a single element (Fig. 17B). The
neural spine articulates with the ‘supraneural’
through a rod-like cartilage or fibrocartilage (in
large individuals), and the ’supraneurals’ articulate
with the dorsal radials or dorsal pterygiophores
from about half the length of the abdominal region of
the body posteriorly.
The ossified paired ventral elements are probably
absent or represented by very small arcocentra that
are fused with the ribs (Fig. 17C) in the abdominal
region. A small cartilage, a remnant of the arcuale,
lies in a small cavity placed in the proximal portion
of the arcocentrum plus rib or at the base of the rib
alone (Fig. 18B,C). Contrary to the situation found
in Lepidosiren, the base of the rib or arcocentrum
plus rib fits in an oval cavity formed by the external
surface of the notochord (Fig. 17A).
The well-developed haemal arches fuse with the
median elements or haemal spines in the caudal
region (Fig. 17B). The haemal spines articulate with
the interhaemal bones along the caudal region and
the interhaemal bones articulate with the ventral
radials or ventral pterygiophores.
We have not observed cartilaginous interdorsal
and interventral elements and there is no evidence
that the swelling of fibrous tissue underlying part of
the base of the arches has developed from interdor-
sal and interventral arcualia. Therefore, we dis-
agree with the interpretation of Mookerjee et al.
(1954) and Bartsch (1989) that these swellings can
be interpreted as probable interdorsal and interven-
tral elements. Remane (1936) illustrated a section of
the vertebral column of Protopterus showing small
cartilaginous dorsal intercalar elements. This condi-
tion is probably occasionally present in certain spec-
imens and/or certain vertebrae of Protopterus like
the occasional intercalar elements described above
for Lepidosiren. Intercalaries have been occasion-
ally observed in the Australian Neoceratodus (see
below).
Fig. 21. Vertebrae and associ-
ated elements of Neoceratodus
forsteri (specimen of 92.4 mm
total length; PC 1998.15).
A: Abdominal-caudal region.
B: Middle section of caudal re-
gion. C: Posteriormost caudal and
postcaudal regions; dar, dorsal ar-
cuale (including the basidorsal
and probably the interdorsal);
dra, dorsal radial or dorsal ptery-
giophore; d.ry, dorsal fin ray; fu.l,
fusion line; hs, haemal spine; id,
interdorsal arcuale; iha, interh-
aemal element or infrahaemal el-
ement; iha⫹vra, cartilage from
which the interhaemal and the
ventral radial originate; na, neu-
ral arch; na⫹ns, neural arch plus
neural spine forming one element;
no, notochord; ptc.c, postcaudal
cartilage; ‘su,’ ‘supraneural;’ sud.c,
supradorsal cartilage; var, ventral
arcuale; var⫹ri, ventral arcuale
at base of rib and rib; vra, ventral
radial or ventral pterygiophore;
v.ry, ventral fin ray.
126 G. ARRATIA ET AL.
There are no significant differences in the struc-
ture of the notochord and its sheaths between Lepi-
dosiren and Protopterus; therefore, see above de-
scription and Figures 8A,B and 9A–C.
There are 33–36 ribs, including one pair of cranial
ribs. The vertebral ribs are robust and well ossified
and bear an elongate groove laterally (Fig. 18A–C).
The first vertebral rib corresponds to the third neu-
ral arch bearing a spine; the two anterior arches with
neural spines do not have ribs (Fig. 19). The first
neural arch is attached to the endocranium. An ele-
ment within the occipital region of the endocranium,
an incomplete arch, that is ossified only in the left side
(Fig. 19) lies anterior to the first neural arch. If we
consider this element as a neural arch, then the first
vertebral rib corresponds to neural arch 4. The strong,
well-ossified cranial rib fits into two fossae of the cra-
nium, one below the other. According to the number of
arches and ribs present in our material, we consider
that the cranial rib in Protopterus corresponds to arch
1 and not 3, as proposed by Fu¨ rbringer (1904).
The postcaudal region of Protopterus (Miller,
1930; Percy, 1962; Bartsch, 1989) is similar to that
described above for Lepidosiren. We will describe its
histological structure in detail so that we can dis-
cuss some aspects of the caudal notochord and the
postcaudal region of dipnoans that have been ig-
nored or misinterpreted in the literature.
Figure 21. (Continued.)
127VERTEBRAL COLUMN IN DIPNOANS
As shown by all serial cross sections examined,
the notochord in Protopterus as well as in Lepidosi-
ren does not extend into the postcaudal region. The
notochord ends in the anterior part of the first post-
caudal cartilage and the notochord and its sheaths
begin to disappear and are replaced by cartilage
(Fig. 20A–D). The first postcaudal cartilage is
formed for a combination of notochordal material
and cartilage (Fig. 20A–C). In the posterior part of
the first postcaudal cartilage and in all posterior
ones the compact cartilaginous centrum is continu-
ous with the neural and haemal arches (Fig. 20D–F).
Caudad, the postcaudal cartilages become smaller
and reach the end of the tail (Fig. 20G). Conse-
quently, with the exception of the first postcaudal
cartilage, the postcaudal vertebral column is exclu-
sively formed by cartilaginous centra. This kind of
centrum formation cannot be confused with that of
the elasmobranchs (see Discussion), as suggested by
Gadow and Abbott (1895), Mookerjee et al. (1954),
and more recently by Bartsch (1989).
The dorsal and ventral camptotrichia are exactly
as above-described and figured for Lepidosiren (see
Figs. 7A, 14A, 15A–C). We have not seen caudal rays
posterior to dorsal and ventral rays, contrary to the
illustration by Goodrich (1909: fig. 205) for Pro-
topterus annectens. The last postcaudal cartilage,
reaching the posterior end of the body, separates
both series of rays that do not reach the end of the
body. Therefore, true caudal fin rays have not been
observed in the material examined.
Ossified dorsal and ventral camptotrichia are ob-
served in all specimens examined. Like in Lepidosi-
ren, they are simple, paired elements, completely
separated from each other and not articulated with
the cartilaginous or ossified radials. Cross sections
of camptotrichia (Fig. 16A,B) show that they are
circular and are completely separated from each
other by mesenchymal cells under the stratified epi-
dermis. The cross sections of small specimens show
two distinct regions within the acellular campto-
trichia that reveal differences in staining: the super-
ficial region with mineral deposits and an inner re-
gion apparently free of any calcification (remnant of
the actinotrichia?) that surrounds an oval–round
small area that reacts to the staining, like the su-
perficial region. (A similar description was provided
for Protopterus annectens by Ge´ raudie and Meunier,
1982.) We have not seen similar structures in other
dermal bones, nor in fin rays of other fishes exam-
ined (compare Fig. 16A,B with C–E).
Protopterus, like Lepidosiren, does not have basal
fulcra and fringing fulcra associated with any fin.
Basal scutes are also absent.
Neoceratodus
The vertebral column of Neoceratodus forsteri (⫽
Ceratodus) is very well known, thanks to the de-
tailed work of Gu¨ nther (1871), Klaatsch (1895), and
Fu¨ rbringer (1904). We concur with Gu¨ nther (1871:
526) that “Neoceratodus agrees perfectly with Lepi-
dosiren in the structure of the vertebral column.”
Therefore, we will address mainly differences be-
tween both genera.
Most specimens reported in the literature, as well
as ours, are small or middle-sized specimens. The
specimen studied by Gu¨ nther was 90 cm long (the
length of the specimen studied by Fu¨ rbringer is un-
known), and our specimens are smaller. The longest
total length known for Neoceratodus is about 110 cm
(A. Kemp, pers. comm., 1999). The inferred young
age of the studied specimens probably explains the
large quantity of cartilage (e.g., large basidorsals
and basiventrals; Fig. 21A–C) that the fishes
present; however, the large specimen studied by
Gu¨ nther still retains large quantities of cartilage.
Therefore, this is apparently the condition found in
Neoceratodus. The large quantity of cartilage
present in Neoceratodus is in contrast to the state
shown by large individuals of Lepidosiren and Pro-
topterus, which exhibit less cartilage.
Large remnants of dorsal and ventral arcualia are
retained during the life of the individuals. The basi-
dorsal and basiventral arcualia (Fig. 21A–C) are
comparatively larger than those in other living dip-
noans. Due to their enlargement, it is unclear
whether each arcuale is a simple element in the
caudal region (Fig. 21C). The doubt arises because in
some of the last vertebral segments the arcualia
resemble basidorsal plus interdorsal and basiven-
tral plus interventral. In addition, the last arcualia
may form large cartilaginous plates that are irreg-
ular in size and shape. The supradorsal cartilages
(Fig. 21A,B), which are elongate, large elements at
the beginning of the caudal region, lose their iden-
tity caudally and also appear as part of the enlarged
arcualia. In contrast, a nodular cartilage is present
in Lepidosiren (Fig. 4B) and Protopterus (Fig. 17C).
The presence of intercalaries is a variable feature
of Neoceratodus. Intercalaries are present in some
specimens (Fig. 21B,C) and absent in others. Dorsal
intercalaries were mentioned and figured by
Klaatsch (1895:151, pl. 7, fig. 5), and ventral inter-
calaries by Fu¨ rbringer (1904:32, pl. 37, fig. 16). The
young specimens examined have interdorsal and in-
terventral cartilaginous intercalaries irregularly
present.
According to Gu¨ nther (1871, pl. 30, fig. 2), the
notochord of Neoceratodus ends some distance be-
fore the tail ends, like the condition described above
in Protopterus and in Lepidosiren. Irregularly seg-
mented elements follow in the postcaudal region,
each with many neural and haemal arches and
spines (Gu¨ nther, 1871, pl. 30, fig. 3), except for the
most distal elements “…scarcely two specimens of
Ceratodus [⫽Neoceratodus] will be found in which
the caudal termination of the vertebral column is
exactly alike” (Gu¨ nther, 1871:527). The postcaudal
region, formed by a variable number of postcaudal
128 G. ARRATIA ET AL.
cartilages, extends to the end of the body, separating
the rays in dorsal and ventral series (Fig. 21C).
Neoceratodus has fewer ribs (about 30) than Pro-
topterus (33–36) and Lepidosiren (55–58). The cra-
nial rib was identified as the first rib by Gu¨ nther
(1871) because of variation in the presence of incom-
plete, cartilaginous anteriormost neural arches. The
first neural arch of Gu¨ nther’s specimen is a short
structure that Gu¨ nther interpreted as the cranial
arch that has moved forward and fused to the neu-
rocranium. All abdominal arches bear their corre-
sponding ribs. Therefore, no space is left between
the cranial rib and the first vertebral rib, unlike the
conditions observed in Lepidosiren (Fig. 13) and Pro-
topterus (Fig. 19). Because of the variation in num-
ber and size of the anteriormost incomplete neural
arches, Fu¨ rbringer (1904) interpreted the cranial rib
of Neoceratodus as the third one and, consequently,
the first vertebral rib as the fourth. We are uncer-
tain about this interpretation because of the varia-
tion found in Neoceratodus. The bases of the ribs of
Neoceratodus have large remnants of the basiven-
tral arcualia (Fig. 21A), which are still present in
large specimens, with a small arcocentrum fused to
the base of each rib.
Data on the development of dipnoan fins are
scarce; one exception is Semon (1893). Gu¨ nther
(1871, pl. 30, fig. 2) illustrated a continuous series of
dorsal, caudal, and anal fin rays (camptotrichia)
that we have not seen in any living dipnoans. In
addition, the last rays, forming the tip of the tail, are
very long, unlike in the specimens studied by us. In
our specimens, like in Lepidosiren and Protopterus,
the postcaudal cartilaginous elements separate the
dorsal and ventral series of rays. The last rays in
Neoceratodus, unlike in Lepidosiren (compare Fig.
21C and Figs. 6, 7A), are very long and are placed
almost parallel to the axis of the body.
According to Gu¨ nther (1871:530) the fin rays (his
dermoneurals or dermohaemals) are not ossified el-
ements: “they consist entirely of cartilage, in which
numerous spindle-shaped cells are imbedded, many
of these cells being produced at both ends into a very
long process.” A similar statement, that the rays
consist of cartilage that show little or no calcifica-
tion, was mentioned by Ryder (1886). These state-
ments disagree with our observations in cleared and
stained specimens and our serial cross sections in
Lepidosiren and Protopterus, and also with Ge´r-
audie and Meunier’s (1982) observations in P. an-
nectens. The camptotrichia of modern adult dipno-
ans do not stain blue when treated with Alcian blue;
however, they react positively to alizarin red, which
shows that the ray is ossified.
Fossil Dipnoans
Fossil dipnoans usually do not preserve the noto-
chord, the centra, or their associated cartilaginous
elements; therefore, the information on most fossil
dipnoans, independent of age, is scarce and incom-
plete. However, there is information on the vertebral
elements of some Paleozoic lungfishes: e.g., 1) on the
centra of dipnoans (e.g., †Soederberghia groen-
landica) from the Upper Devonian of Greenland,
which have compact centra (Sa¨ ve-So¨ derbergh, 1937;
Jarvik, 1952; Lehman, 1959; Schultze, 1970); 2) on
the neural and haemal arches, spines, ribs, and ‘su-
praneural’ bones of †Rhinodipterus ulrichi and
Fig. 22. A: Part of the verte-
bral column of †Uranolophus wyo-
mingensis (FMNH PF 3874) from
the Lower Devonian of USA.
B: Restoration of an abdominal
vertebra; ep.p, epineural process;
f.sn, foramen for spinal nerve; ha,
haemal arch; na, neural arch; nc,
spinal cord; no, notochord; ns,
neural spine.
129VERTEBRAL COLUMN IN DIPNOANS
†Dipterus cf. D. valenciennesi from the Upper Devo-
nian of Bergisch-Gladbach, Germany (Schultze,
1975a, figs. 2a–c, 3a,b); 3) on the neural arches of
†Uranolophus wyomingensis from the Lower Devo-
nian of Wyoming (Fig. 22A,B) (Denison, 1968a,b;
Campbell and Barwick, 1988, figs. 37, 38); 4) on the
vertebral centra of †Griphognathus sculpta from the
Upper Devonian of Germany (Schultze, 1969); and
5) on the vertebrae of †Griphognathus whitei from
the Upper Devonian of Australia (Rosen et al., 1981,
fig. 34A; Campbell and Barwick, 1988, figs. 34A–C,
35A–F, 36A–C, 1999, figs. 18–20) and on Adelargo
schultzei (Johanson and Ritchie, 2000, figs. 3D,E).
In addition, information is available on well-
preserved material showing the vertebral column
and associated elements of Permian and Triassic
dipnoans such as †Conchopoma (Schultze, 1975b),
†Paraceratodus (Lehman et al., 1959), and others.
Compact centra are missing in †Uranolophus (Fig.
22A,B). Denison (1968a,b) interpreted the remains
of centra in the holotype of †U. wyomingensis as
representing intercentra (⫽haemal arches). He
thought that the pleurocentra (⫽interdorsal) were
not preserved. Such an interpretation is quite differ-
ent from the available information on extant and
fossil dipnoans. The centrum in †Uranolophus is
formed by the broad and well-ossified basal part of
the neural arch and a ventral, broad element that
we interpret as the haemal arch. Our interpretation
of the neural arch of †Uranolophus differs also from
that by Campbell and Barwick (1988, fig. 38), who
reconstructed the neural arch as formed by the fu-
sion of both halves, like the condition present in
†Griphognathus whitei. The abdominal neural
arches of †U. wyomingensis differ from those
present in †G. whitei because they are separated
medially and each half extends dorsally in an elon-
gate neural spine. Each half presents a foramen
(probably for a nerve) at the base of the neural spine;
therefore, each vertebra has two foramina and con-
sequently they are not for the passage of the dorsal
(or supradorsal) ligament, which is a median struc-
Fig. 23. Section of the vertebral column of †Dipterus cf. D. valenciennesi from the Upper Devonian of Germany. A: Specimen MB.
f. uncat. B: Specimen SMNH 6726 (Aand Bslightly modified from Schultze, 1975a). C,D: Diagrammatic lateral views of abdominal
vertebrae showing the relationships between dorsal and ventral elements; a.ri, articular surface for rib; bp.df1-2, basal plate of dorsal
fin 1-2; df1, dorsal fin 1; dra, dorsal radial or dorsal pterygiophore; ?d.‘su’, displaced ‘supraneural’?; ha⫹hs, haemal arch plus haemal
spine forming one element; na, neural arch; no, notochord; ns, neural spine; paph, parapophysis; ri, rib; ‘su,’ ‘supraneural.’
130 G. ARRATIA ET AL.
ture. The neural arches of †Uranolophus present
strong, broad epineural processes posterolaterally
directed like those found in †G. whitei. No parap-
ophyses and ribs have been found in †Uranolophus.
The vertebral column in †Dipterus shows the
same elements (Fig. 23A–D) described above for ex-
tant lungfishes. In the abdominal region, the right
and left sides of the neural arch seem to be fused
distally, forming a broad articular facet for the neu-
ral spine (identified as proximal supraneural by Ahl-
berg and Trewin, 1995). The neural spines are elon-
gated, rod-like, and heavily ossified. The spines
articulate with the ‘supraneural’ bones (identified as
distal supraneural by Ahlberg and Trewin, 1995)
that are comparatively shorter than the neural
spines. Another element is added in the region un-
der the first dorsal fin: an elongate and irregularly
shaped bony plate that supports the fin rays (lepi-
dotrichia) of the first dorsal fin (the basal segments
of these rays were mistakenly interpreted as radials
by Campbell and Barwick, 1988). One dorsal proxi-
mal radial has been observed below the plate. The
lepidotrichia of the second dorsal fin are also sup-
ported by a bony plate. Radials have not been ob-
served associated with the plate of the second dorsal
fin. The sequence of associated elements is unclear
in the caudal region of †Dipterus and †Rhino-
dipterus and we therefore cannot describe it.
In contrast to Recent, Mesozoic, and certain Pa-
leozoic lungfishes, the vertebral columns of the
Paleozoic lungfishes †Griphognathus, †Rhyncho-
dipterus, and †Soederberghia, have ossified, com-
pact, independent centra (Figs. 24A–C, 25A,B). Cen-
tra are unknown in other Paleozoic lungfishes such
as †Dipterus and †Rhinodipterus (Fig. 23A–D); how-
ever, the most anterior vertebrae of some †Rhino-
dipterus specimens form massive, ossified elements.
Small and large specimens do not show centra pos-
Fig. 24. Vertebral centra and fins of †Griphognathus sculpta (slightly modified from Schultze, 1969) from the Upper Devonian of
Germany. A: Posterior part of the body showing caudal centra and the structure of the fins. B: Enlargement of caudal centra
iluustrated in A. C: Enlargement of abdominal centra; af, anal fin; cf, caudal fin; df1-2, dorsal fin 1-2; fu, fulcra; pvf, pelvic fin.
131VERTEBRAL COLUMN IN DIPNOANS
teriorly, and it is most probable that these fishes did
not have chordacentra or ossified centra. The noto-
chord only supported dorsal and ventral bony ele-
ments. No intercentra have been observed. The cen-
tra in †Griphognathus are heavily ossified and have
a small central perforation for the notochord. They
are slightly amphicoelous, with the exception of
some centra with an anteriorly convex face. Accord-
ing to Schultze (1970) and our observations on ver-
tebrae of †Griphognathus whitei, these compact cen-
tra are formed by calcified cartilage and bone (Fig.
26). Numerous cavities of different sizes and shapes
are left in the calcified cartilage and bone. The cen-
tra of †G. whitei (Fig. 25B) are more strongly ossified
than those of †G. sculpta (Figs. 24A–C, 26); this
corresponds to a size difference (centra of †G. whitei
are four times larger in diameter than those of †G.
sculpta). These centra represent ossified holocentra.
Cartilage is replaced by calcified cartilage and the
latter by bone.
In †Griphognathus sculpta and †Soederberghia
groenlandica, both abdominal and caudal centra
show that the neural and haemal arches are autog-
enous because there are no vestiges of the arches
fused with the centra (Fig. 24A–C). The abdominal
vertebrae of †G. whitei also exhibit autogenous neu-
ral arches (Fig. 25B; Campbell and Barwick, 1988,
figs. 34A,b, 35A–C, 1999, figs. 18–20). The neural
arch is a massive structure that expands laterally
and has a broad, short median process. The ventral
face of the arch leaves a small space for the passage
of the spinal cord. The median neural process has a
foramen for the passage of the supradorsal ligament
(Campbell and Barwick, 1988, figs. 34A, 35A–F,
1999, figs. 18A,C, 19A–C,E,F, 20A–C). After compar-
isons with extant lungfishes and other fishes, we
interpret this part as part of the neural arch, not as
a short neural spine as Campbell and Barwick
(1988, 1999) concluded. Consequently, we interpret
the articular surface at the dorsal part of the median
process of the neural arch as for the neural spine,
not as the articular surface for the supraneural bone
as Campbell and Barwick (1988, 1999) stated.
The Late Devonian dipnoans †Scaumenacia and
†Fleurantia (Fig. 27A,B) have a persistent noto-
chord and no vertebral centra. There is a series of
dorsal neural arches plus their spines that extend to
the posterior margin of the second dorsal fin. No
ossified neural arches and spines are found in the
tail. Caudally the size of the dorsal arches and neu-
ral spines decreases strongly. Below the second dor-
sal fin, each neural arch plus its spine (the spine was
identified as supraneural by Cloutier, 1996) articu-
lates distally with a bone that by comparison with
living dipnoans and other sarcopterygians corre-
sponds to the ‘supraneural’ (the ‘supraneural’ was
identified as proximal radial by Cloutier, 1996) and
the distal bone or dorsal radial that supports the
Fig. 25. Abdominal vertebrae in lateral view of †Griphognathus whitei from the Upper Devonian of Gogo, western Australia.
A: Modified from Campbell and Barwick (1988). B: MB. f.7074 (the neural arches are not preserved). Arrow points anteriad; ep.p,
epineural process; na, neural arch; ns, neural spine; paph, parapophysis; vce, vertebral centrum.
132 G. ARRATIA ET AL.
rays (dorsal radial was identified as medial radial by
Cloutier, 1996). The ventral series of bones is incom-
pletely known in †Scaumenacia and †Fleurantia.
Just posterior to the pelvic fins are a few haemal
arches plus their spines. Another series of bones
appears at the base of the caudal fin, but it is un-
clear whether each bone is the interhaemal alone or
the interhaemal plus radial or the radial alone.
The Permo-Carboniferous dipnoan †Sagenodus
shows a morphology similar to that in living forms.
There are no centra, so that the notochord is persis-
tent and the dorsal and ventral series of elements
are identical to those illustrated in Figure 1 (KUVP
84201; Schultze and Chorn, 1997, fig. 39).
A series of ‘supraneural’ bones in the abdominal
region is observed in the Paleozoic forms †Dipterus
and †Rhinodipterus (Fig. 23A–D), the Triassic
Paraceratodus (Fig. 28B), as well as in recent dip-
noans (e.g., Figs. 4A, 10A,B, 13). However, indepen-
dent ’supraneurals’ have not been observed in the
abdominal region of †Scaumenacia, †Fleurantia
(Fig. 27A,B), and †Conchopoma (Fig. 28A).
A pair of cranial ribs is known in many dipnoans,
e.g., †Barwickia (Long, 1993), †Conchopoma (Schul-
tze, 1975b), †Dipterus (Ahlberg and Trewin, 1995,
fig. 9b), †Fleurantia (Graham-Smith and Westoll,
1937), †Rhinodipterus (Schultze, 1975a), †Sageno-
dus (Schultze and Chorn, 1997), †Scaumenacia
(Cloutier, 1996), †Tellerodus sturi (Teller, 1891) and
modern lungfishes (e.g., Gu¨ nther, 1871; Fu¨ rbringer,
1904). Apparently, cranial ribs are absent in marine
Gogo lungfishes †Chirodipterus, †Gogodipterus,
†Griphognathus, and others (Long, 1993). They are
absent in †Gnathorhiza and apparently in †Uranolo-
phus.
The “true” ribs or pleural ribs of †Griphognathus
whitei articulate with very short and broad parap-
ophyses that are fused to the ventrolateral face of
the centra (Fig. 25B). Each parapophysis has an
almost round articular fossa for the rib. The ribs
articulate with irregularly shaped haemal arches
bearing short parapophyses in †Dipterus. These
haemal or ventral arches are built of calcified carti-
lage surrounded by a fine bony ossification (ventral
Fig. 26. Cross section of a verte-
bra of †Griphognathus sculpta from
the Upper Devonian of Bergisch-
Gladbach, Germany (SMNH S2146;
Schultze, 1970). The illustrated sec-
tion corresponds to the lateroven-
tral portion associated to the region
of the ventral arcualia (represented
by an empty space ⫽gr); bon, bone;
cc, calcified cartilage; cc.wL, calci-
fied cartilage with waves of Liese-
gang; l.cb, lacuna of bone cell; l.cc,
lacuna of cartilage cell and the un-
calcified region around it; no.c, ca-
nal for spinal cord; res.c, cavity pro-
duced by reabsorbed tissue.
133VERTEBRAL COLUMN IN DIPNOANS
arcocentra). Parapophyses have not been observed
in †Scaumenacia and †Fleurantia. It is unclear
whether the bases of the ribs of †Scaumenacia,
†Fleurantia, and †Sagenodus are associated with
remnants of the ventral arcualia. However, the de-
scription by Graham-Smith and Westoll (1937)
agrees with the possibility that the base of the rib in
†Fleurantia also includes remnants of the ventral
arcualia, as it does in recent dipnoans. The ribs in
the three genera are long, reaching close to the ven-
tral margin of the body.
In the caudal region of †Dipterus the ventral ele-
ments form elongate bones that correspond to the
haemal arch plus the haemal spine, both fused into
a single element, as described above in Lepidosiren.
Interhaemal bones and radials have not been ob-
served. Furthermore, the information is incomplete
for †Scaumenacia and †Fleurantia. Fortunately,
specimen KUVP 84201 shows that in †Sagenodus,
as in recent dipnoans, a complete ventral series of
elements includes the haemal arches plus haemal
spines, interhaemals, and ventral radials.
Devonian lungfishes have well-developed and in-
dependent unpaired fins, e.g., two dorsal fins, a cau-
dal fin, and an anal fin (Fig. 27A,B). Consequently,
the dorsal, caudal, and anal fins are clearly distin-
guishable from each other, contrary to the condition
found in Lepidosiren, Protopterus, and Neocerato-
dus. The first dorsal fin is reduced in †Scaumenacia
(Fig. 27A). There are no independent dorsal and
anal fins starting with Carboniferous forms such as
†Sagenodus, the Permo-Carboniferous †Concho-
poma (Fig. 28A), etc., but a long fin that may be inter-
preted as a fusion of the dorsal and caudal fins and
another long fin that may correspond to the anal plus
the caudal fins (See Comparison and Discussion).
Fig. 27. Restorations of dipnoans from the Upper Devonian of Miguasha, Quebec, Canada. A:†Scaumenacia curta modified from
Cloutier (1996). B:†Fleurantia denticulata modified from Graham-Smith and Westoll (1937) and Cloutier (1996); af, anal fin; cf, caudal
fin; df1-2, dorsal fin 1-2; dra, dorsal radial or dorsal pterygiophore; ha⫹hs, haemal arch plus haemal spine forming one element;
na⫹ns, neural arch plus neural spine forming one element; pef, pectoral fin; pvf, pelvic fin; ri, rib; ‘su,’ ‘supraneural.’
134 G. ARRATIA ET AL.
The fin rays of †Dipterus, †Rhinodipterus,
†Griphognathus, †Scaumenacia, and †Fleurantia
are distally segmented and finely branched (Figs.
23A, 24A, 27A,B) and are interpreted here as lepi-
dotrichia. The proximal portions of the dorsal fin
rays are represented by unsegmented, long bases
(identified as radials by Campbell and Barwick,
1988) that articulate with the bony plates support-
ing fin rays. They are formed by paired hemilepi-
dotrichia and can be considered as “true” dermal fin
rays. Devonian lungfishes have well-defined caudal
fin rays. For instance, †Fleurantia (Fig. 27B) has
both epaxial and hypaxial caudal rays; however,
†Scaumenacia has well-developed hypaxial rays and
the epaxial ones are minuscule (Fig. 27A). Carbon-
iferous and younger fossil dipnoans (Fig. 28A,B) do
not have “true” caudal fin rays, similar to the condi-
tion shown by extant dipnoans.
Small fringing fulcra are known from the dorsal
fins and anal fin of †Griphognathus sculpta (Fig.
24A; Schultze, 1969), and they also have been ob-
served in the new material of †Dipterus examined
here. An epaxial caudal fulcrum is observed in
†Uranolophus wyomingensis. Basal scutes are asso-
ciated with the base of the second dorsal fin in †U.
wyomingensis; we cannot provide information on
†Griphognathus due to its incomplete preservation.
Basal scutes have been observed in new material of
†Dipterus.
Information on Carboniferous and younger fossil
dipnoans provides a scenario different from that of
Devonian forms. Vertebral elements are well known
from the Permo-Carboniferous genus Conchopoma
(Fig. 28A). The genus †Conchopoma is very interest-
ing because it shows some transitional features. For
instance, most neural arches are separated from the
neural spines from early ontogeny (e.g., FMNH PF
89701, with a body length of ca. 30 mm). Only the
last vertebral segments show a fusion of neural
arches and spines as in modern lungfishes. How-
Fig. 28. Vertebral column and associated elements of certain fossil dipnoans. A: Restoration of †Conchopoma gadiforme from the
Permian of Lebach/Saar, Germany (after Schultze, 1975b). B: Restoration of †Paraceratodus germaini from the Lower Triassic of
Madagascar (restoration done by H.-P. Schultze after the material deposited at the MNHN); dra, dorsal radial or dorsal pterygiophore;
ha⫹hs, haemal arch plus haemal spine forming one element; iha, interhaemal element or infrahaemal element; iha⫹vra?, cartilage
from which the interhaemal and the ventral radial originate?; na, neural arch; na⫹ns, neural arch plus neural spine forming one
element; ns, neural spine; pvf, pelvic fin; ri, rib; ‘su,’ ‘supraneural;’ ‘su’⫹dra?, ‘supraneural’ plus radial?; vra, ventral radial or ventral
pterygiophore.
135VERTEBRAL COLUMN IN DIPNOANS
ever, the smallest specimen of Lepidosiren examined
here has all neural arches unfused with the neural
spines. The fused condition is acquired early in on-
togeny. Another series of elongate rod-like bones
dorsal to the neural spines is preserved in the pos-
terior half of the body. These bones are absent in the
abdominal vertebrae. If the interpretation of Schultze
(1975b) is correct that they are dorsal radials,
because of their association to the dorsal fin rays,
then †Conchopoma lost the series of ‘supraneural’
bones. By comparisons with living and fossil forms
(Fig. 28B) and specimens of †Conchopoma (e.g., MB.
f.7133 and f.7134) that show incomplete sutures, we
suggest that the elements supporting the rays cor-
respond to fused ‘supraneural’ plus radial. The same
is suggested for the ventral series (a fusion of inter-
haemal with radial).
The available information on the vertebral column
and associated elements of the Early Triassic genus
†Paraceratodus from Madagascar shows a similar
pattern of bony elements (Fig. 28B) as those found in
living forms. The abdominal vertebrae (neural
arches, neural spines, ‘supraneurals,’ ribs) and the
caudal vertebrae (neural arches fused to neural
spines, ‘supraneurals,’ dorsal radials, haemal arches
fused with haemal spines, interhaemals, and ven-
tral radials) show the same elements, and sequences
of elements, illustrated above for Lepidosiren, Pro-
topterus, and Neoceratodus.
Fossil dipnoans have usually been restored as if
there is a continuous series of dorsal, caudal, and
ventral rays (e.g., Lehman, 1959, 1966; Schultze,
1975b) following previous illustrations of Pro-
topterus and Neoceratodus (see above) in which a
series of posterior caudal rays was erroneously illus-
trated. According to the new information presented
above for recent dipnoans, we consider the caudal
tip of the tail as rayless in fossil, post-Devonian
dipnoans (Fig. 28A,B).
Other Sarcopterygians
The available information on other sarcoptery-
gians is not very abundant, but still, it documents a
significant diversification among groups. Informa-
tion is not available yet for the onychodonts and
some genera of debated phylogenetic position such
as †Powichthys and the Chinese genera †Youngo-
lepis, †Diabolepis, and †Psarolepis.
Actinistians. The Actinistia are represented to-
day by Latimeria chalumnae. Nevertheless, the
group has a long history, with more than 50 fossil
genera (Cloutier and Forey, 1991), beginning with
the Middle Devonian †Euporosteus eifelianus from
Germany and ending with the Late Cretaceous
†Megalocoelacanthus dobiei from North American
(Schwimmer et al., 1994). In general, the morphol-
ogy of actinistians remained conservative through-
out time, but some interesting changes can be no-
ticed, as, for instance, in the vertebral column and
its associated elements.
The vertebral column and associated elements of
actinistians are known by descriptions and illustra-
tions of the modern Latimeria by Millot and An-
thony (1956, 1958) and the more detailed work by
Andrews (1977), and of some fossil taxa such as the
Early Triassic †Laugia (Stensio¨ , 1932), the Late
Triassic †Diplurus (Schaeffer, 1952), the Late Juras-
sic †Coccoderma and †Lybis (e.g., Lambers, 1992),
and the Early Cretaceous †Axelrodichthys (Maisey,
1986, 1991).
The vertebral column and associated elements in
Latimeria consist of a large notochord surrounded
by its fibrous sheaths and paired rows of cartilagi-
nous elements (remnants of the dorsal and ventral
arcualia) forming almost a continuous series dor-
sally and ventrally (Fig. 29A–C). The notochord does
extend into the posterior tip of the tail. It ends close
to the posterior margin of the supplementary caudal
lobe and posteriorly is surrounded by irregularly
shaped cartilages that support the small, thin, and
usually incompletely ossified rays of the supplemen-
tary caudal lobe (Fig. 29C).
Paired cartilages support the bony neural arches
and spines and the bony haemal arches and spines
along the whole vertebral column (Fig. 29A–C). A
series of dorsal cartilages or intercalaries (⫽pleuro-
centra in tetrapod terminology) occur posterior to
the cartilaginous base of each neural arch. A series
of ventral cartilages, ovoid or rectangular in shape,
with a high degree of variability in shape and size
along the body, is interpreted as being formed by
cartilaginous bases of the haemal arches or, in other
words, remnants of the ventral arcualia (⫽intercen-
tra in tetrapod terminology). The smaller ventral
cartilages probably correspond to ventral intercen-
tra that never ossify (terminology of Andrews, 1977).
In Latimeria, a series of median, unpaired bones
is placed dorsally and ventrally in the caudal region
proximal to the fin rays of epaxial and hypaxial lobes
of the tail. These rod-like bones that support the
rays were identified as epineural and infrahaemal
spines by Andrews (1977). Because of their position,
and by comparison with the situation in other fishes,
we interpret them as a fusion between the ‘supran-
eural’ and radial, and of interhaemal and radial,
respectively. A series of independent ‘supraneural’
bones has not been observed in Latimeria. Ribs are
absent.
The structure of the supplementary or accessory
lobe (Millot and Anthony, 1958, fig. 1; Andrews,
1977, fig. 3) of the tail in Latimeria is difficult to
interpret. The small series of dorsal and ventral
cartilages, remnants of the arcualia, apparently are
the site of insertion of the rays, unlike in any other
fish that we have studied (Fig. 29C), and, therefore,
an autapomorphy. We suppose that these cartilagi-
nous elements are potentially able to form other
bones (as in the abdominal and caudal regions) like
136 G. ARRATIA ET AL.
Fig. 29. Sections of the vertebral column and associated elements of Latimeria chalumnae (modified from Andrews, 1977). A: Most
anterior abdominal region. B: Region between the neural arches 60 and 70. C: Posterior region of body; ha, haemal arch; na1-95, neural
arch 1-95; na⫹ns90, neural arch plus neural spine 90; ns, neural spine; sl.cf, fin rays of accessory caudal lobe; ‘su’⫹ra, ‘supraneural’
plus radial forming one element.
137VERTEBRAL COLUMN IN DIPNOANS
neural spines and radials, but because of the small
space left, these elements never form distal ele-
ments.
According to Millot and Anthony (1958) and An-
drews (1977) the tips of the neural spines are mod-
ified to articulate with epineural spines that support
the bifurcating radials of the caudal fin. (We believe
that the bases of the rays or lepidotrichia, whose
proximal section is deeply sunk in the skin, have
been confused with radials supporting rays.) We
interpret the “epineural spine” as representing ‘su-
praneural’ plus radial. The same situation is seen in
fossil specimens of †Diplurus (Fig. 30), †Laugia, and
†Holophagus.
The caudal fin rays (without those of the supple-
mentary caudal lobe) of Latimeria have long, well-
ossified bases. However, their distal portions are
incompletely ossified and their segmentation is very
irregular, e.g., some rays are segmented, others are
not, and some rays are incompletely segmented
along their length.
The diameter of the notochord in Latimeria seems
to be greater than that in other sarcopterygians
fishes, but at the level of the second dorsal fin it
tapers suddenly, having a smaller diameter. Accord-
ing to the restorations, the notochord was also an
important element in fossil actinistians.
Fringing and basal fulcra and basal scutes are not
associated with fins in Latimeria and in fossil actin-
istians; however, serrated segments or fringing
fulcra-like elements are associated with the most
anterior caudal rays (and also rays of the first dorsal
fin) in †Diplurus (Fig. 30).
Fossil actinistians show a similar vertebral skel-
eton and associated structures as the modern Latim-
eria, but there are some significant differences. For
instance, cartilaginous or ossified intercalaries (or
pleurocentra) have not been found in fossil forms.
Strong ribs occur in †Diplurus (Fig. 30) and a series
of short, delicate ribs is present in †Laugia and
†Holophagus (Fig. 31B). There is no suture or line of
fusion between the neural arch and its spine (the
same in the haemal region). They form a single
element. In the caudal region of †Holophagus, the
distal tips of the neural and haemal spines are
broader than the rest of the spine and leave a lateral
groove where the proximal bifid portion of the ‘su-
praneural’ plus radial inserts. Unlike in Latimeria
(Millot and Anthony, 1958; Bjerring’s figure in Jar-
vik, 1981), the tail of †Laugia, †Diplurus, and †Ho-
lophagus, among others, begins with two or three
dorsal (‘supraneurals’) and ventral (interhaemals)
elements that do not bear fin rays and articulate
proximally with the tip of the neural and haemal
spines, respectively (Figs. 30, 31A,B).
A series of the Middle Pennsylvanian †Rhab-
doderma exiguum (Fig. 32A,B) ranging between ⬃35
mm and 240 mm total length shows a persistent
notochord and no vertebral centra, like the condition
found in Latimeria. The neural arches are large and
they bear long neural spines that become shorter
caudally. In the caudal region, the neural spines of
the specimen of 35 mm total length distally face
other series of bones that we interpret as the ‘supra-
neurals’ plus radials by comparison with other sar-
copterygians. Although the cartilage is not pre-
served, we assume that the distal end of the neural
spine and proximal end of the ‘supraneural’ were
joined by cartilage, as in extant forms. In young
specimens, each dorsal ray is supported by one bony
element (‘supraneural’ plus radial). In the younger
specimens the rays or lepidotrichia are simple and
Fig. 30. Restoration of the actinistian †Diplurus newarki from the Upper Triassic of New Jersey (slightly modified from Schaeffer,
1952); df1-2, dorsal fin 1-2; ha⫹hs, haemal arch plus haemal spine forming one element; iha, , interhaemal element or infrahaemal
element; iha⫹ra?, interhaemal and ventral radial forming one element?; na⫹ns?, neural arch plus neural spine forming one element?;
pef, pectoral fin; pvf, pelvic fin; ri, rib; ‘su,’ ‘supraneural;’ ‘su’⫹ra?, ‘supraneural’ plus radial forming one element?
138 G. ARRATIA ET AL.
unsegmented but they become distally segmented in
older specimens (Fig. 33).
The ventral series of elements of †Rhabdoderma
shows that the ribs are missing and that the haemal
arches are as well-developed as the neural ones and
each is continuous with the haemal spine, forming
one single element (Fig. 32A,B). As for the dorsal
series, there is only one bone between the distal end
of the neural spine and bases of the rays.
Ossified pleural ribs are present in some fossil
actinistians; they can be very short or large. For
instance, large ribs have been described or illus-
trated for †Diplurus (Fig. 30) (Schaeffer, 1952),
†Changxingia (Wang and Liu, 1981), †Axelrodich-
thys (Maisey, 1986, 1991), whereas small ribs are
present in actinistians such as †Laugia (Stensio¨,
1932) and †Holophagus (Fig. 31A,B). Other fossil
actinistians (e.g., †Coccoderma suevicum; Lambers,
Fig. 31. Lateral view of the actinistian †Holophagus penicillata from the Upper Jurassic of Solnhofen, Bavaria (MB. f.7070)
(A), and enlargement of the abdominal region (B). Note the short ribs (indicated by arrows); cf, caudal fin; df1-2, dorsal fin 1-2; pef,
pectoral fin; pvf, pelvic fin; sl.cf, fin rays of accesory caudal lobe; ‘su,’ ‘supraneural.’
139VERTEBRAL COLUMN IN DIPNOANS
Fig. 32. Caudal vertebrae of †Rhabdoderma from the Middle Pennsylvanian of Illinois. A:†Rhabdoderma exiguum, specimen of 35
mm total length (FMNH 6272). B:†Rhabdoderma sp., specimen of about 70 mm total length (FMNH PF 5906); ha⫹hs, haemal arch
plus haemal spine forming one element; iha⫹ra, interhaemal and ventral radial forming one element; na⫹ns, neural arch plus neural
spine forming one element; ‘su’⫹ra, ‘supraneural’ plus radial forming one element.
140 G. ARRATIA ET AL.
1992) do not have ribs. Ribs are lacking in living
Latimeria.
Rhipidistians. The vertebral column and associ-
ated elements are known from several taxa: 1) the
porolepiform †Glyptolepis (Watson and Day, 1916;
Andrews and Westoll, 1970a); 2) the osteolepiforms
†Osteolepis (Moy-Thomas, 1939; Westoll, 1943; An-
drews and Westoll, 1970a), †Thursius (Andrews and
Westoll, 1970a), †Megalichthys (e.g., Agassiz, 1833;
Woodward, 1891; Andrews and Westoll, 1970a),
†Eusthenopteron (e.g., Bryan, 1919; Gregory et al.,
1939; Jarvik, 1952, 1981; Andrews and Westoll,
1970b; Hitchcock, 1995), and †Lohsania (Thomson
and Vaughn, 1968); and 3) the elpistostegalian
†Elpistostege (Schultze and Arsenault, 1985; Schultze,
1996). These taxa reveal an interesting diversifica-
tion of the vertebral column and associated ele-
ments, as shown below. Because both fish and tet-
rapod terminologies have been used in descriptions
of rhipidistian vertebrae, here we use the fish ter-
minology first and in brackets the corresponding
tetrapod terms.
The vertebrae of †Glyptolepis differ from those of
the osteolepiforms in several aspects. The notochord
is the main element supporting the dorsal and ven-
tral arches and their associated structures. In the
abdominal region the dorsal series is formed by the
neural arch with the slightly elongate neural spine.
The neural spine articulates distally with the
‘supraneural.’ The series of ‘supraneurals’ is known
from the first vertebra at least until the second
dorsal fin. The first ‘supraneurals’ are shorter than
the posterior ones, which are elongate, rod-like ele-
ments. Small paired intercalaries [⫽pleurocentra]
are positioned between the neural arches (which
correspond to the ossified interdorsals). The haemal
arches [⫽intercentra] have well-defined parapophy-
ses for articulation with short ribs. In the caudal
region, the haemal arch continues as the haemal
spine. The distal tip of the spine articulates with the
interhaemal.
The structure of the vertebrae of †Osteolepis dif-
fers depending on the body region (Fig. 34A,B). Ver-
tebrae from the anterior part of the trunk have
neural arches composed of two separate halves, a
ventral arcocentrum [⫽intercentrum] forming most
of the centrum and one large dorsal intercalar [⫽
pleurocentrum]. Vertebrae of the middle region of
the trunk have the halves of the neural arches pro-
longed into neural spines that are unfused medially.
The ventral arcocentra are unfused medially as well
as the intercalaries mediodorsally. In contrast, each
caudal vertebra has one elongate neural spine, one
haemal arch with a short haemal spine that articu-
lates with the interhaemal element, and the inter-
calaries are smaller than in the anteriormost verte-
brae.
†Megalichthys has long been known as having
ring-centra (Fig. 34C–E). However, the genus is
characterized by considerable intraspecific and indi-
vidual variation (see Andrews and Westoll, 1970a).
For instance, each abdominal vertebra consists only
of the annular centrum and the neural arch and
spine. No haemal elements and ribs are present.
Nevertheless, considerable variation is found in the
texture of the centra, e.g., the centrum can be
bounded by unfinished bone or by smooth or wrin-
kled periosteal bone. Some individuals may have
half centra. The neural arches of the abdominal
region are associated with their centra only in some
large individuals, where they are fused. ‘Supraneu-
rals’ are unknown in †Megalichthys. The outer sur-
Fig. 33. Dorsal fin rays and their support in †Rhabdoderma exiguum from the Middle Pennsylvanian of Illinois (FMNH UC14389;
about 24 cm length). Arrows point to two rays showing distal segmentation at the distal part of the rays; ‘su’⫹ra, ‘supraneural’ plus
radial forming one element.
141VERTEBRAL COLUMN IN DIPNOANS
face of caudal vertebrae is covered with periosteal
bone. Both neural and haemal arches in the caudal
region are fused to the centrum, but they are still
unfused in some specimens.
According to Jarvik (1981), based on thin sections,
the vertebral column in †Eusthenopteron foordi com-
prises a series of centra that bear neural arches,
neural spines, supraneurals (⫽‘supraneurals’) and
haemal arches, ribs, and haemal spines, depending
on the body region. In addition, there is a series of
small, paired interdorsal elements (Figs. 34F,G,
35C, 36B,C) or intercalaries [⫽pleurocentra],
placed posteriorly to the neural arch. All these ele-
ments are highly variable, as was accurately de-
scribed by Andrews and Westoll (1970b). The cen-
trum is formed by the basal part of the neural arch
and mostly by the haemal or ventral arch that ex-
tends dorsally, enclosing almost the complete noto-
chord. Consequently, the centrum in †Eusthe-
nopteron is of the arcocentral type, fundamentally
formed by the ventral arcocentra [⫽intercentra]. In
contrast to the arcocentral type of centra present in
†Eusthenopteron,†Megalichthys does not have this
type of centrum formation because its arches are
separated from the centrum (in a body segment) in
some specimens. Without ontogenetic series it is
impossible to interpret the fusion of the arches and
centra shown by some specimens and, therefore, the
formation of the centra is incompletely known.
According to Jarvik’s (1981) restoration (Fig. 34G)
the neural arches in †Eusthenopteron are continu-
ous with the neural spines that are sutured medi-
ally. In contrast, both halves are fused into one spine
in large specimens (e.g., FMNH PF 6258). Between
both halves of the spine is the foramen for the su-
pradorsal ligament. The most anterior abdominal
Fig. 34. Abdominal vertebrae of osteolepiform rhipidistians. A,B:†Osteolepis panderi in lateral and posterior views.
C–E:†Megalichthys hibberti in lateral and posterior views, and caudal vertebra in lateral view (E)(A–E modified from Andrews and
Westoll, 1970a). F,G:†Eusthenopteron foordi in lateral and posterior view (slightly modified from Jarvik, 1981); f.vr, foramen for spinal
nerve in intercalar; ha, haemal arch; hs, haemal spine; inc, intercalar; na, neural arch; nc.c, canal for spinal cord; no.c, notochordal
canal; ns, neural spine; paph, parapophysis; pit, pit in the side of the annular centrum; ri, rib; ‘su,’ ‘supraneural.’
142 G. ARRATIA ET AL.
vertebrae bear supraneurals joined to the distal tip
of the neural spines (Fig. 34F,G;) (Jarvik, 1981).
These ‘supraneurals,’ with the exception of the first
two, which are fused to each other (Jarvik, 1981),
are in continuation with the tip of the spine. They
are short, broad elements, unlike the ‘supraneurals’
found in other sarcopterygians. However, similar
‘supraneurals’ are occasionally observed at the dis-
tal portion of the neural spines and we are uncertain
whether †Eusthenopteron have ‘supraneurals’ at all
because they look more like result of a breakage
than real structures. Hitchcock (1995, figs. 3, 7A)
described and figured three large, oval-like elements
that he interpreted as supraneurals. These bones
are even larger than the neural arch plus spine,
unlike similar bones in other fishes. We have been
unable to observe the ‘supraneurals’ described and
illustrated by Jarvik and Hitchcock.
The short ribs present in †Eusthenopteron articu-
late with small parapophyses placed dorsolaterally
on the haemal arches (Fig. 34F,G), unlike in other
sarcopterygians bearing ribs, where the position is
Fig. 35. Vertebrae of the osteolepiform rhipidistian †Eusthenopteron foordi (FMNH 390) from the Upper Devonian of Miguasha,
Quebec, Canada. A: Lateral view. B: Detail of the posterior abdominal vertebrae showing a dorsoposterior process of the haemal arch
(indicated by arrows). C: Anterior caudal region. Arrow points to a rib; ha, haemal arch; ns, neural spine; sc, scale.
143VERTEBRAL COLUMN IN DIPNOANS
Fig. 36. Vertebrae of the osteolepiform rhipidistian †Eusthenopteron foordi (FMNH 390) from the Upper Devonian of Miguasha,
Quebec, Canada. A: Posterior abdominal vertebrae. B: Anterior caudal region. C: Caudal region illustrating the association with the
bases of the dorsal fin 2 and the anal fin; ac.b, accessory bone; bp.af, basal plate of anal fin; bp.df2, basal plate of dorsal fin 2; ha, haemal
arch; hs, haemal spine; inc, intercalar; na, neural arch; ns, neural spine; paph, parapophysis; ra.af, radial bone supporting anal fin;
ra.df2, radial bone supporting dorsal fin 2; ri, rib; r.no, remain of notochord; sc, scale.
144 G. ARRATIA ET AL.
ventrolateral. However, specimen FMNH 390, with
an excellent preservation of the vertebral column,
shows that the first ribs (Figs. 35A,B, 36A) are really a
dorsoposterior process of the haemal arch. The poste-
rior ribs are short and massive bones that articulate
on the dorsolateral face of the haemal arch (Figs. 35C,
36B).
The vertebral column of the elpistostegid fish
†Elpistostege is known from 16 or 17 pieces of ver-
tebrae that were interpreted as neural arches and
intercentra by Schultze (1996:319). The neural
arches have a smooth perichondral surface, whereas
the intercentra (that probably develop from the
basiventral arcualia or basi- plus interventral arc-
ualia) have a rougher internal surface formed by
endochondral bone. Interdorsal elements [⫽pleuro-
centra] were not recognized.
Elasmobranchs
A comparison of the vertebral column of dipnoans
and elasmobranchs is appropriate since similarities
between both have long been stated in the literature
and have even been used for establishing certain
conclusions that our observations do not support.
Among chondrichthyans, Ko¨ lliker (1860) distin-
guished different kinds of centra: 1) centra produced
entirely by the chordal sheath (⫽fibrous sheath),
e.g., in Hexanchus; 2) centra formed by the chordal
sheath and by the outer skeletogenous layer, the
biconcave central cones being strengthened by ex-
ternal thin calcification, e.g., in most elasmo-
branchs; and 3) centra formed entirely by the outer
skeletogenous layer, by the cartilaginous arches,
e.g., in skates.
The vertebral column of selachians has been stud-
ied by Ko¨ lliker (1860), Goette (1878), Hasse (1879–
1885), Klaatsch (1895), Ridewood (1921), Goodrich
(1930), White (1937), Applegate (1967), and Pado-
vani Ferreira and Vooren (1991), among others. The
vertebral column is strictly segmental in origin.
Four paired elements in each segment contribute to
its development. These are the dorsal and ventral
arcualia: the smaller anterior arcualia are the inter-
basalia and the larger posterior ones are the basi-
dorsalia. The bases of the basidorsals and basiven-
trals grow over the elastica externa and the
chordacentra are developed in continuity with them
as complete rings. The vertebral column is a flexible
skeletal cover of notochord, neural canal, and blood
vessels. There are no joints between the centra and
no articular processes between the arches. Unex-
pectedly, the presence of bone tissue in the neural
arches of Scyliorhinus has been reported by
Peignoux-Deville et al. (1982) and Bordat (1987). It
is unclear whether the presence of bone tissue is
found also in other elasmobranchs or is a unique
character of this genus.
Each centrum formed in the fibrous sheath grows
by expansion into typically amphicoelous cartilages
in continuity with the basalia. The elastica externa
may be absorbed and the limits between the chor-
dacentra my be lost. The surface of the cartilage is
usually covered by a thin calcification; in addition,
denser calcifications are deposited in the centra in
large individuals.
In the caudal region the number of the vertebral
segments is double that of the myomeres, e.g., there
are two centra and two sets of dorsal and ventral
arcualia per segment. The duplication of the verte-
bral parts is interpreted as being related to the
lengthening of the caudal segments (Goodrich,
1930). The caudal tip of the column consists of irreg-
ularly placed arcualia. The arcualia are replaced by
blocks of cartilage that tend to fuse to a continuous
rod completely enveloping and replacing the poste-
rior extremity of the notochord, as in dipnoans.
In addition to older publications like those by Ko¨l-
liker (1860) and Gadow and Abbott (1895), more
recently Mookerjee et al. (1954) and Bartsch (1989)
support the view that the centra of dipnoans and
elasmobranchs have similar modes of formation. Ac-
cording to our observations, the centrum formation
of elasmobranchs and dipnoans are different, de-
spite the fact that there is a migration of cartilage
cells from the arcualia (or neural and haemal arches
of larger individuals) into the fibrous sheath in both
(see Figs. 8B, 9A,D).
Most elasmobranchs develop chordacentra. Ex-
tant dipnoans do not form chordacentra. The noto-
chord may become thicker due to the presence of
cartilaginous cells in its matrix, but the fibrous
sheath does not calcify, unlike in most elasmo-
branchs. Therefore, it is incorrect to state that dip-
noans have chordacentra.
Elasmobranchs commonly have basidorsalia (basi-
dorsal and basiventral arcualia) and smaller inter-
basalia (interdorsalia and interventralia). Dipnoans
commonly have only basidorsalia and basiventralia.
Occasionally some specimens show interdorsal
and/or interventral elements. There is no duplica-
tion of vertebral elements in the caudal region of
dipnoans, as occurs in elasmobranchs.
The posterior tip of the vertebral column of elas-
mobranchs is formed by blocks of cartilage that fuse
into a continuous rod that envelops and replaces the
end of the notochord. This is the only point of simi-
larity that we can find between the vertebral column
of dipnoans and elasmobranchs. In dipnoans there is
a continuous series of cartilages (postcaudal carti-
lages) that can become fibrocartilages in large spec-
imens, but they do not tend to fuse with each other
(see Fig. 7A). The notochord ends at the beginning of
the first postcaudal cartilage (Fig. 20A–C). The pos-
terior cartilages that bear neural and haemal arches
(Fig. 20D–F) are only compact blocks of cartilagi-
nous tissue. In conclusion, the similarity is only at
the caudal tip of the vertebral column of young in-
dividuals of elasmobranchs and living dipnoans.
145VERTEBRAL COLUMN IN DIPNOANS
The fin rays or ceratotrichia are known from fossil
and extant elasmobranchs (see, for instance, Dean,
1909; Zangerl, 1973, 1981; Bendix-Almgren, 1975;
Kemp, 1977; Ge´ raudie and Meunier, 1980). Their
structure is reminiscent of actinotrichia. The dorsal
and anal fins supports were described in the 19th
century (e.g., Thacher, 1877; Mivart, 1887; Bridge,
1896). In the majority of species the radial elements
are cartilaginous, rod-like structures and usually
divided into proximal, mesial, and distal segments.
Actinopterygians
Information concerning the actinopterygian verte-
brae and associated elements has been provided in
several articles, but they have never been the sub-
ject of a comprehensive study concerning primitive
and advanced members of the group. Because there
are so many differences between actinopterygian
subgroups, and the information is so complex, we
will deal with this subject in detail in another arti-
cle. Here, we offer general information to contrast it
with that in sarcopterygians.
Vertebral centrum. In general, within acti-
nopterygians there are different types of centrum
formation. The following combination of elements
may participate in the formation of the vertebral
centrum:
1. Cartilaginous basidorsal and basiventral (and
commonly interdorsal and interventral) arcualia
and a persistent notochord (Figs. 37A, 38A–C)
exist throughout the life of the fish. However,
ossified neural and haemal arches can develop
very late in ontogeny, e.g., in some acipenseri-
forms (Grande and Bemis, 1991:81).
Fig. 37. Diagrammatic types of centrum formation in actinopterygians. A: Cartilaginous elements: arcualia (no centra). B: Dorsal
and ventral arcocentra. C: Dorsal and ventral arcocentra and holocentrum including chordacentrum. D: Dorsal and ventral arcocentra
and chordacentrum. E: Dorsal plus ventral arcocentrum and chordacentrum. F: Dorsal and ventral arcocentra, chordacentrum, and
thin autocentrum. G: Dorsal and ventral arcocentra, chordacentrum, and well developed autocentrum. H: Dorsal and ventral
arcocentra plus autocentrum; auc, autocentrum; bd, basidorsal arcuale; bv, basiventral arcuale; chc, chordacentrum; darc, dorsal
arcocentrum; darc⫹auc⫹varc, fused dorsal arcocentra, autocentrum, and ventral arcocentra; darc⫹varc, fused dorsal and ventral
arcocentra; fsh, fibrous sheath of notochord; nc.c, canal for spinal cord; no, notochord; ns, neural spine; varc, ventral arcocentrum.
146 G. ARRATIA ET AL.
2. Dorsal and ventral arcocentra, remnants of car-
tilaginous basidorsal and basiventral arcualia,
and a persistent notochord (Fig. 37B) exist, e.g.,
in primitive actinopterygians such as †Cheirol-
epis (Arratia and Cloutier, 1996), the palaeonis-
ciforms †Mimia (Fig. 39A) and †Moythomasia
(Gardiner, 1984), †Birgeria, and †Boreosomus
(Nielsen, 1949), †saurichthyiforms (Rieppel,
1985), the neopterygian †Semionotus (Olsen and
McCune, 1991), and others.
3. Dorsal and ventral arcocentra, remnants of car-
tilaginous basidorsal and basiventral arcualia,
and an ossified, compact centrum (holocentrum)
with a chordacentrum (Fig. 37C) exist, e.g., in
some Late Paleozoic palaeonisciforms from cen-
tral North America (Schultze and Chorn, 1986,
figs. 2–6) and probably in polypteriforms. (Hemi-
centra, cartilaginous preformed vertebrae, have
been illustrated for a specimen of 21.5 mm length
of Polypterus senegalus by Bartsch et al. [1997].)
4. Dorsal and ventral arcocentra, remnants of car-
tilaginous basidorsal and basiventral arcualia,
and a chordacentrum (Fig. 37D) exist, e.g., in a
variety of primitive actinopterygians such as †Os-
pia (Stensio¨ , 1932) and †Australosomus (Fig.
39B) (Nielsen, 1949); in various neopterygians
such as †Tetragonolepis (Thies, 1991), †caturids,
†pachycormiforms (Arratia and Lambers, 1996),
†Prohalecites (Tintori, 1990), and †Pholidophorus
bechei (Arratia, 1991, 1999). Chordacentra show
variations. They can be only ventral, or only dor-
sal or both ventral and dorsal, or they can be
ring-like chordacentra.
5. Dorsal and ventral arcocentra forming a complete
bony ring and a chordacentrum (Fig. 37E) exist,
e.g., in the Jurassic †Aspidorhynchus and in cer-
tain Late Jurassic ‘pholidophoriforms’ (Arratia,
1999).
6. Dorsal and ventral arcocentra, a chordacentrum,
and an autocentrum (Fig. 37F,G) exist, e.g., in
†Leptolepis coryphaenoides and in more advanced
teleosts (Arratia, 1991, 1997, fig. 89A—D, 1999;
Arratia and Schultze, 1992). The autocentrum is
very thin in †L. coryphaenoides. It becomes thick
and differently ornamented in more advanced te-
leosts, where it may be separated into four por-
tions by remnants of the basidorsal and basiven-
tral arcualia producing the typical cross of Malta
Fig. 38. Vertebral column of Polyodon spathula (PC 190599; 106 mm total length). A: Anteriormost region. B: Posterior abdominal
region. C: Caudal region; a.ry, anal fin ray; bd, basidorsal arcuale; bv, basiventral arcuale; c.ara, cartilaginous anal radial; c.dira,
cartilaginous caudal distal radial; c.dra, cartilaginous dorsal radial; fu, fulcra; hs, haemal spine; id, interdorsal arcuale; iv, interventral
arcuale; na6-41, neural arch 6-41; nc, spinal cord; no, notochord; ?, unknown.
147VERTEBRAL COLUMN IN DIPNOANS
described by Goette (1878), Grassi (1882), and
Scheele (1893) (Fig. 40A), or it may be a compact
piece of bone (Fig. 40C).
7. Dorsal and ventral arcocentra and the autocen-
trum (Fig. 37H) fuse, e.g., in catfishes (Arratia,
1991) (Fig. 40B,C).
The neural arches are independent elements
(from each other) along the vertebral column in ac-
tinopterygians, except at the tip of the caudal region
of living forms where the neural arches form a con-
tinuous cartilaginous plate (e.g., Amia: Fig. 41C)
(Schultze and Arratia, 1986, fig. 14A, 1989, figs.
13A,B; Grande and Bemis, 1998, figs. 80–82) or a
continuous series of large plates (e.g., acipenseri-
forms: Figs. 38A,C, 41A) or small plates (e.g., lepi-
sosteiforms: Fig. 41B) (Schultze and Arratia, 1986,
fig. 6, 1989, fig. 16; polypteriforms: Bartsch and
Gemballa, 1992, fig. 4). In most teleosts the neural
arches of the ural vertebrae are reduced or absent
(see Monod, 1968, figs. 116, 445, 528; Schultze and
Arratia, 1988, figs. 3C, 4, 5A; Arratia, 1991, figs. 1, 5,
8a,b, 11), with the exception of the elopomorphs
that in early ontogeny have a compound cartilag-
inous plate corresponding to the ontogenetic fu-
sion of the basidorsal arcualia belonging to the
preural centrum 1 plus the ural basidorsal arcua-
lia (Schultze and Arratia, 1988, figs. 14A,B, 17A,
16B–D) or may present a special situation, as the
stegural in salmonids (for details, see Arratia and
Schultze, 1992).
The neural spines (Figs. 42A–C, 43A,B) are
formed by dorsal growth and ossification of the basi-
dorsal arcualia. The ossification of the spine differs
among actinopterygians. It may be the result of peri-
chondral (e.g., Amia) or membranous (e.g., clupeo-
morphs, siluriforms) ossification. The neural spines
can be paired or they may fuse medially into a single
element. Most actinopterygians have paired neural
Fig. 39. Restorations of the Late Devonian palaeonisciform †Mimia toombsi (A) (modified from Gardiner, 1984) and of the Triassic
†Australosomus kochi (B) (modified from Nielsen, 1949); af, anal fin; df, dorsal fin; dint, dorsal intercalar; dra, dorsal radial or dorsal
pterygiophore; ep.p, epineural process; f.ra, fused radials; ha, haemal arch; hs, haemal spine; na, neural arch; ns, neural spine; paph,
parapophysis; pef, pectoral fin; pvf, pelvic fin; su, supraneural; vra, ventral radial or ventral pterygiophore.
148 G. ARRATIA ET AL.
spines in the abdominal region (Figs. 37A–G, 38A),
with the exception of advanced teleosts (Fig. 37H).
According to Goodrich (1930), true neural spines
develop through the fusion of the right and left
halves of the neural arches.
Most actinopterygians have elongate neural
spines; however, short neural spines that extend
dorsal to the spinal cord are present in the abdom-
inal region in Acipenseriformes (Fig. 38A,B) and
Cladistia. They are apparently missing in the caudal
region just posterior to the dorsal fin in acipenseri-
forms (Fig. 38B,C). A unique situation is found in
the caudal region where a median broad cartilage is
continuous with the proximal part of the arcualia, or
neural arch. The median, distal portion of this struc-
ture separates from the proximal part, but the single
cartilage is still seen in the caudal region of young
specimens (Fig. 41A). The series of single cartilages
was identified as the supraneural series by Grande
and Bemis (1991). However, the origin and role of
these elements in the caudal region do not corre-
spond to the origin and role of supraneurals in other
actinopterygians. By comparison to other fishes
these elements may correspond to the neural spine
plus the supraneural or the neural spine plus the
radial. Whatever the correct interpretation, it is an
unusual condition among actinopterygians and sar-
copterygians. A similar situation is found in polyp-
teriforms, where it is also unclear whether neural
spines or supraneurals are present. (We will deal in
detail with these problems elsewhere.)
Neural spines are paired in the caudal region in
primitive actinopterygians and median neural
spines in the caudal region are considered a halecos-
tome character (Patterson, 1973). However, median
neural spines occur in the caudal region of †Moytho-
masia (Gardiner, 1984), †Australosomus,†Birgeria
(Nielsen, 1949), and †Haplolepis (Baum and Lund,
1974). Paired neural spines are found in large indi-
viduals of Lepisosteus (Schultze and Arratia, 1986,
fig. 5A,B).
The neural arches of the abdominal region of some
primitive actinopterygians bear robust, broad, short
processes that project posterolaterally. They were
identified as lateral basidorsal processes in †Austra-
losomus,†Boreosomus, and †Glaucolepis by Nielsen
(1949) and as epineural processes in †Mimia by
Gardiner (1984), in the neopterygian incertae sedis
†Prohalecites by Tintori (1990), and in †Pholidopho-
rus bechei by Patterson and Johnson (1995). These
processes are absent in a variety of actinopterygians
such as polypteriforms, acipenseriforms, †saurich-
thyiforms, †pycnodontiforms, amiiforms, lepisoste-
iforms, †aspidorhynchiforms, and others. Primitive
Fig. 40. Cross sections (PC uncat.) at the middle region of teleostean centra. Picro-ponceau staining technique. A: The salmonid
Thymallus thymallus. Note the ‘cross of Malta’ produced by the remnants of the arcualia. B,C: The siluriform Trichomycterus
areolatus. Note the absence of cartilage; centrum and arches are formed of compact bone; auc, autocentrum; ha, haemal arch; na,
neural arch; nc, spinal cord; no.c, notochordal canal.
149VERTEBRAL COLUMN IN DIPNOANS
Fig. 41. Epaxial region of the posterior tip of the body with some hypaxial elements of some actinopterygians. A:Polyodon spathula
(KU 22121). B:Lepisosteus osseus (KU 3677). C:Amia calva (KU 6883); c.ha, cartilaginous haemal arch or ventral arcuale; c.ry, caudal
fin rays; dar, dorsal radial or dorsal pterygiophore; ‘E,’ ‘epural;’ Hy, hypural; id, interdorsal arcuale; nc, spinal cord; no, notochord; ns,
neural spine; sud.c, supradorsal cartilage.
teleosts such as †Leptolepis coryphaenoides,†Thar-
sis, and elopomorphs (Fig. 44A–D) have long, thin
epineural processes, whereas these are separate el-
ements or epineural bones in some advanced forms
(for differences in terminology, see Arratia, 1997,
1999).
Supraneural bones. A series of unpaired supra-
neurals (e.g., Figs. 39A,B, 43A,B, 44A–D) are known
from palaeonisciforms (e.g., †Mimia: Gardiner,
1984; †Glaucolepis and †Birgeria: Nielsen, 1949;
and †Coccolepis: Traquair, 1911), and more ad-
vanced actinopterygians. In most actinopterygians
supraneurals are present from the anteriormost ver-
tebra to the beginning of the pterygiophores of the
dorsal fin, whereas in others they do not reach the
dorsal fin at all. The series of supraneurals is absent
in some teleosts, e.g., the atherinomorphs Basilich-
thys,Odontesthes, and Patagonina.
In all actinopterygians where the ontogeny is
known the supraneural cartilage develops sepa-
rately from the neural spine (Figs. 42B, 43A,B,
44A–D, 45). The ossification of the supraneural car-
tilages commonly progresses from the third or
fourth cartilaginous supraneural posteriorly, and
then the first supraneurals ossify (e.g., Amia calva:
Schultze and Arratia, 1989, fig. 12A; Pomoxis annu-
laris: Mabee, 1988, fig. 4; in Lepisosteus osseus, and
in teleosts such as Elops saurus [Fig. 44A–D], Hi-
odon alosoides,Oncorhynchus mykiss, and others
where we examined ontogenetic series).
Haemal arches. In actinopterygians the parap-
ophyses develop from the ventral arcualia. Their
appearance may progress caudad or rostrad. In most
actinopterygians there is a single series of ventral or
pleural ribs that are associated with the ventral
arcocentra or parapophyses of the abdominal region
and are in the wall of the coelom. However,
Polypterus has two ribs associated with each parap-
ophysis.
Caudal fin and rays. A “true” caudal fin with
rays formed independently from those of the dorsal
and anal fins is found in actinopterygians. According
to the relationships of caudal fin rays and caudal
endoskeleton, the caudal fin of actinopterygians may
be heterocercal (e.g., lepisosteiforms) or homocercal
(e.g., teleosts). The situation is again controversial
for acipenseriforms and polypteriforms. For in-
stance, most authors claim that the polypteriform
caudal fin includes both dorsal (epichordal) and ven-
tral (hypochordal) lobes, and it has been considered
a “disguised heterocercal” by Goodrich (1930), an
“abbreviated homocercal” by Jollie (1962), a “diphyc-
ercal type” by Romer (1970), a “heterocercal type” by
Jarvik (1981), and more recently a “shortened het-
erocercal caudal fin confluent with the dorsal fin” by
Bartsch and Gemballa (1992). According to Schmal-
hausen (1912, 1913) and Sewertzoff (1924) the pre-
sumed epichordal lobe is the posterior part of the
specialized dorsal fin, an interpretation that we
share. If this is correct, then the tail of polypteri-
forms lacks dorsal caudal fin rays and includes only
the hypaxial caudal rays.
Early in ontogeny, the dorsal-caudal-anal finfold
is supported by actinotrichia that are slender rods of
collagen (e.g., Becerra et al., 1983) or a kind of col-
lagen, elastoidin (e.g., Baudelot, 1873; Ge´ raudie and
Meunier, 1980; Ge´ raudie and Landis, 1982). The
function of the actinotrichia is taken over by the
lepidotrichia and by the fulcra in actinopterygians.
The lepidotrichia are formed either by cellular or
acellular bone. Each lepidotrichium is composed of
two bony hemicylinders (Fig. 16C–E) that surround
an intrasegmental region containing blood vessels,
bundles of nerves, and connective tissue (Schultze
and Arratia, 1989). The two bony hemicylinders or
hemilepidotrichia may be slightly separated from
each other (Fig. 16C,D), or may independently be
close together (e.g., in most teleosts; Fig. 16E).
COMPARISON AND DISCUSSION
The notochord
Independent of the modifications presented by the
vertebral elements in elasmobranchs and teleo-
stomes, the notochord is always present either as 1)
a persistent element during the whole life of the
individual of some fish groups (e.g., dipnoans, actin-
istians, acipenseriforms) or 2) as a complete, well-
developed element in early ontogeny, but as a con-
stricted and reduced element in late ontogeny in
other fish groups (e.g., elopomorphs, osteoglosso-
morphs, and in more advanced teleosts).
The notochord has long been interpreted as the
point of origin of the formation of the vertebral col-
umn. It is clearly shown in Figure 42A–C that the
notochord and its fibrous sheath direct the forma-
tion of elements of the centrum (e.g., chordacen-
trum, autocentrum) by changes in its appearance
that include the presence of invaginations that are
regularly present and in changes in the density and
chemical composition of the sheaths and notochord.
A persistent notochord is known from ‘Agnatha’
and Gnathostomata. A relationship between the no-
tochord and the basal plate in the adult basicranium
may be a synapomorphy of osteichthyans (Schaeffer,
1968). A notochordal canal is present in the basal
plate of the palaeoniscoids, crossopterygians, and
Devonian dipnoans. Although the notochord was
persistent in adult placoderms, there was no rela-
tionship to the basal plate. In acanthodians the ven-
tral occipital ossification lies in the posterior part of
the basal plate. This ossification has a dorsal groove
that must be the lower part of the notochordal canal.
The notochordal canal passes at least halfway along
the ventral occipital ossification (Miles, 1973:88).
The relationship of the notochord to the basal plate
is unknown in acanthodians.
Among living sarcopterygians and actinoptery-
gians, a notochord extending to the posterior end of
the tail is found in actinopterygians such as acipens-
151VERTEBRAL COLUMN IN DIPNOANS
Figure 42
eriforms (Fig. 41A) and lepisosteiforms (Fig. 41B).
In other actinopterygians such as Amia calva and
elopomorphs, osteoglossomorphs, and more ad-
vanced teleosts, the notochord does not reach the
caudal margin of the tail.
The posterior end of the notochord surrounded by
a series of cartilages is the condition found in the
actinistian Latimeria chalumnae (Fig. 29C) and in
acipenseriforms (Fig. 41A) in which the dorsal and
ventral arcualia join posterior to the notochord or
around its posterior end. The posterior end of the
notochord is followed by a series of postcaudal car-
tilages in dipnoans (Figs. 7A, 14A) and elasmo-
branchs or by only one small cartilage, the opisthural
cartilage, in teleosts (see Arratia and Schultze,
1992, figs. 9A, 13A, 21). In contrast to the postcaudal
cartilages in dipnoans, the teleostean opisthural
cartilage lacks both neural and haemal arches.
The Gadowian Theory
Gadow and Abbott (1895) proposed a theory of the
composition of fish vertebrae and Gadow (1896) ex-
plicitly applied the theory to tetrapods. The theory
was accepted for fishes (despite criticisms by Hay,
1897) and tetrapods, until Williams (1959) pre-
sented contrary evidence for tetrapods. The theory
stated succinctly:
“The key to the solution of the composition of the
vertebral column is given [Fig. 46A] by the
metameric repetition of the four pairs of symmetri-
cally arranged cartilaginous elements the origin of
which we have traced in Fishes, namely–
One pair of basidorsalia.
One pair of basiventralia (with its lateral out-
growths ⫽ribs or pleurapophyses and haemal ven-
tral outgrowths ⫽haemal arches, chevrons, wedge-
bones, haemapophyses).
One pair of interdorsalia.
One pair of interventralia.
The first of these four pairs is always present and
forms the neural arch. Of the other three pairs any
one may be suppressed, sometimes even two in the
same sklerotome” (Gadow, 1896:50).
As shown in Figure 46A, the posterior limit of a
body segment runs posterior to the basidorsal and
basiventral arcualia. We have not observed this
scheme in living fishes and we will present else-
where a detailed description of our findings in acti-
nopterygians.
Interdorsal and interventral arcualia (Fig. 46B,C)
are variably present in those sarcopterygians and
actinopterygians that have these elements (Figs.
21B, 29B,C, 39B; e.g., Neoceratodus,Latimeria, ami-
iforms, ‘pholidophoriforms’). The intercalaries are
unknown in early ontogeny of teleosts above the
phylogenetic level of †Leptolepis coryphaenoides,
where only basidorsal and basiventral arcualia are
present (Fig. 46D).
The Gadowian theory has had a strong influence
on theoretical interpretations of the formation of the
centra. For instance, starting from the assumption
that a centrum must include the four arcualia, the
centrum of †Eusthenopteron has been interpreted as
being developed from the basidorsal (neural arch),
the interdorsal (intercalar or pleurocentrum), and
basiventral and interventral arcualia (both of which
are supposed to form the single haemal or ventral
arch) (Jarvik, 1952, 1981; Thomson and Vaughn,
1968; Andrews and Westoll, 1970a). A similar ap-
proach was followed to interpret the formation of the
vertebral centrum in other rhipidistians (e.g., Rack-
off, 1976; Schaeffer, 1968; Andrews and Westoll,
1970b). Following this scheme, the pleurocentra and
intercentra of rhipidistians have been postulated to
be homologous to the pleurocentra and intercentra
of labyrinthodont amphibians by Jarvik (1952), Pan-
chen (1967, 1977), Andrews and Westoll (1970b),
and others.
The idea that the four arcualia should be consis-
tently present has also influenced interpretations of
how the centrum of living fishes should be formed,
independently of what information is provided by
early ontogenetic stages. For instance, the presence
of interdorsal and/or interventral elements in dipno-
ans has been assumed because of some ‘swellings’ in
the posterior part of the neural or of the haemal
arches (e.g., Mookerjee et al., 1954). These ‘swell-
ings’ are supposed to be interdorsal and interventral
elements that become fused with the basidorsal and
basiventral arcualia. However, detailed studies of
early ontogeny show that these approaches are
wrong; we suggest checking a series of actual growth
and avoiding ontogenetic inferences based on shape
and size of structures.
Unfortunately, the theory that apparently holds
for some fishes, specifically for some of those studied
by Gadow (e.g., elasmobranchs such as Scyllium,
Acanthias, and probably Centrophorus, and primi-
tive actinopterygians such as Acipenser, Polyodon),
was applied to vertebrates as a whole. It was said
that Myxine possesses cartilaginous elements re-
stricted to the tail in the shape of dorsalia and that
Petromyzon possesses dorsalia and ventralia (Ga-
dow and Abbott, 1895); in reality, the Myxini lack
arcualia (Janvier, 1981). Our preliminary studies in
young Amia and teleosts do not support the theory.
Fig. 42. Portion of the vertebral column and dorsal and anal
radials of the perciform Stizostedion vitreum. (PC 130499; 11 mm
total length; c. and st.). A: Anterior abdominal region. Note that
cartilaginous supraneurals are absent. B: Posterior abdominal
and anterior caudal regions with cartilaginous supraneurals and
radials. C: Enlargement of two basidorsal arcualia forming the
neural arches and spines, under phase contrast. Arrows point to
obliterations of the notochord; c.dra1, cartilaginous dorsal radial
1; cl, cleithrum; c.su, cartilaginous supraneural; d.li, dorsal liga-
ment; ha1, haemal arch 1; hs, haemal spine; na, neural arch; ns,
neural spine; paph, parapophysis.
153VERTEBRAL COLUMN IN DIPNOANS
Gadow (1896) distinguished four types of vertebrae
in tetrapods. However, the arcualia theory is not
applicable to tetrapods (Williams, 1959).
Centrum Formation
General information is first required on the ver-
tebral column of various vertebrates before the quite
diverse composition of the vertebral centra in fishes
can be discussed.
There are present different vertebral centra in
different groups of Gnathostomata. For instance:
Chondrichthyans commonly have cartilaginous neu-
ral and haemal arches and cartilaginous chordacen-
tra, which are the result of migration of cartilage
cells into the fibrous sheath of the notochord. Placo-
derms commonly lack centra (but see Schultze,
1975a, pl. 1, fig. 5a,b). Ossified neural and haemal
arches and spines are known (Fig. 47) (Ørvig, 1960,
figs. 4B, 5B, pls. 26, 27; Stensio¨ , 1963, pls. 1, 32). The
arthrodire †Jagorina has complete vertebral centra
(Stensio¨ , 1959, figs. 61–65). Each centrum is contin-
uous with a short, broad neural spine. Among the
teleostomes the acanthodians have a vertebral col-
umn consisting of a persistent notochord, neural and
haemal arches, and a few radials in the tail. For
sarcopterygians and actinopterygians, see descrip-
tions above.
Types of centra. As mentioned above, Gadow
and Abbott (1895:190) distinguished two types of
centra formation in fishes: archcentrum and chorda-
centrum. According to these authors, the archcentra
Fig. 43. Development of the supraneural cartilages (separated from the neural spines) and cartilaginous radials of the perciform
Stizostedion vitreum (PC 130499). A: Specimen of 11 mm total length. B: Specimen of 15 mm total length; c.ara1-4, cartilaginous anal
radial 1-4; c.dra1-5, cartilaginous dorsal radial 1-5; c.su2-17, cartilaginous supraneural 2-17; d.ry, dorsal ray; ha1, haemal arch 1; hs,
haemal spine; na1-2, cartilaginous neural arch 1-2; no, notochord; paph, parapophysis.
154 G. ARRATIA ET AL.
are directly dependent on the existence of arcualia;
they occur in “Ganoids,” Teleostei, Amphibia, and
Amniota. In contrast, chordacentra are of their own
origin (autocentrum), independent of the arches.
They occur in all elasmobranchs and potentially also
in Holocephali and Dipnoi.
We disagree with the interpretation by Gadow and
Abbottt (1895) that the chordacentrum is an autocen-
Fig. 44. Anterior vertebrae and anterior part of dorsal fin of Elops saurus (TCWC 0782.1). A,B: Specimen of 41 mm total length.
C,D: Specimen of 58 mm total length; auc⫹chc, autocentrum surrounding the chordacentrum; chc, chordacentrum; dra, dorsal radial or
dorsal pterygiophore; ep.p, epineural process; na, neural arch; ns, neural spine; ri, rib; su1-33, supraneural 1-33; v1-30, vertebra 1-30.
155VERTEBRAL COLUMN IN DIPNOANS
trum formed by itself, independent of the arches. The
formation of the chondrichthyan chordacentrum is de-
pendent on the existence of arches because the carti-
lage cells that form the centrum migrate from the
arches into the fibrous sheath of the notochord. As
shown above, dipnoans do not have chordacentra.
Schultze and Arratia (1988, 1989), Arratia (1991),
Arratia and Schultze (1992), and herein recognize
different types of centra in fishes, e.g., arcocentrum,
chordacentrum, and autocentrum (see above).
Arcocentra are present in placoderms (Fig. 47),
acanthodians, sarcopterygians (Figs. 4A–C, 13,
17A–C, 21, 23A–D, 34A–G), and actinopterygians
(Figs. 37B–H, 39A,B). They are absent in chondrich-
thyans. The arcocentra develop from dorsal and ven-
tral arcualia. Therefore, they constitute the dorsal
or neural arch and the ventral or haemal arch that
sit dorsally and ventrally, respectively, on the noto-
chord. Dorsal and/or ventral arcocentra of the cau-
dal region are laterally expanded in some fishes
without contacting each other (e.g., in large dipno-
ans, Fig. 4A,C, and in numerous actinopterygians,
e.g., in †pachycormiforms).
Arcocentra can form the vertebral centrum itself
and, consequently, the centrum is of the arcocentral
type. For instance, the notochord is completely en-
closed mainly by the ventral arcocentra and partly
by the dorsal arcocentra in †Eusthenopteron (Figs.
34F,G, 35A,B, 36A,B). Another arcocentral type is
found in several actinopterygians (e.g., Jurassic as-
pidorhynchiforms and Late Jurassic ‘pholido-
phoriforms’) where the centra are formed by the
lateral growth of both dorsal and ventral arcocentra
that fuse with each other, enclosing the chordacen-
tra[um] (Fig. 37E).
The presence of ossified neural and haemal arches
(or arcocentra) appears homologous in placoderms
and teleostomes, with loss in chondrichthyans. The
development of the arches into arcocentra producing
most of the centrum or the complete centrum is
interpreted here as independently acquired in some
rhipidistians (e.g., †Eusthenopteron) and in some
actinopterygians. Therefore, they are not homolo-
gous.
Compact centra appear independently in some
placoderms, actinopterygians, and sarcopterygians.
They are not present in acanthodians. Compact, os-
sified centra are known from some arthrodires
within †placoderms, some Paleozoic dipnoans, some
†rhipidistians, and some †palaeonisciforms.
A vertebral column formed by a persistent noto-
chord without vertebral centra is the primitive pat-
Fig. 45. Neural spines, last supraneurals, and first dorsal radials of Elops saurus (TCWC 0781.1; specimen of 58 mm total length).
Note that the last supraneurals are positioned between dorsal radials 3 and 4; dra1, dorsal radial or dorsal pterygiophore 1; d.ry, dorsal
fin ray; ns, neural spine; su, supraneural.
156 G. ARRATIA ET AL.
tern for all gnathostomes. The formation of centra is
acquired independently in some lineages of placo-
derms, some primitive and most advanced acti-
nopterygians, and some dipnoans and rhipidistians.
The chordacentrum of chondrichthyans is a spe-
cialization of the group. A chordacentrum, as a di-
rect mineralization and/or calcification of the fibrous
sheath of the notochord, is apparently a novelty
found only in actinopterygians. However, because
the information is missing for many primitive acti-
nopterygians we cannot establish where the chorda-
centrum arose first. An autocentrum, as a direct
ossification formed outside the elastica externa, is a
novelty found in advanced actinopterygians, e.g., in
teleosts such as †Leptolepis coryphaenoides and
more advanced forms (Arratia, 1999).
Neural Spines, Supraneurals, and Radials
Neural spines. Appearance and ossification of
neural spines differ among recent fishes, judging
from the ontogeny where it is known.
Neural spines, as products of dorsal growth of the
basidorsal arcualia (either as perichondral ossifica-
tion or as membranous ossification at the distal end
of the basidorsal arcualia) are present in placoderms
and teleostomes (Figs. 10A–D, 11A,B, 13, 21A,B,
27A,B, 28A,B, 30, 37A–G, 43A,B, 47). The neural
spines of the abdominal region of most adult fishes
are in continuation with the neural arches or arco-
centra. However, the neural spines of the abdominal
region are separate elements in most adult dipno-
ans, including some Paleozoic forms such as
†Dipterus (Fig. 23A–D) and †Conchopoma (Fig.
28A), the Triassic †Paraceratodus (Fig. 28B) and
recent Lepidosiren,Protopterus, and Neoceratodus,
(Figs. 4A, 17A, 21A). The neural spines of the ab-
dominal region are in continuation with their arches
in the Paleozoic dipnoans †Uranolophus and
†Griphognathus (Figs. 22A,B, 25A), †Scaumenacia
(Fig. 27A), and †Fleurantia (Fig. 27B). We interpret
the presence of neural arch plus spine as the prim-
itive character state and the separation of the neu-
ral arch from its spine as an apomorphic character
state by comparison with the condition of the neu-
ral spine plus the neural arch in fishes other than
dipnoans (e.g., placoderms, chondrichthyans, and
most teleostomes).
†Scaumenacia (Fig. 27A) and †Fleurantia (Fig.
27B) are unique among dipnoans by the presence
of a neural arch plus long neural spine that almost
reaches the dorsal margin of the body. It is
unclear, by comparison with other dipnoans,
whether the long spine also includes the ‘supran-
eural.’
Fig. 46. Gadow’s diagram for
the formation of a vertebra (en-
closed in a rectangle of heavy
lines) from four arcualia (A). The
region between slashed lines
would correspond to an original
body segment (based on Gadow
and Abbott, 1885, and Gadow,
1886). B–F: Diagram illustrating
variation of the formation of the
vertebrae from the four arcualia:
basidorsal (BD), basiventral (BV),
interdorsal (ID), and interventral
(IV).
157VERTEBRAL COLUMN IN DIPNOANS
Most dipnoans are characterized by vertebrae
with one single, unpaired neural spine (Fig. 1A–C).
In contrast, †Uranolophus has both halves of each
neural spine unfused (Fig. 22A,B). The single neural
spine is elongated and narrow in most dipnoans. It
represents a continuation of the neural arch as a
detached spine.
Most actinopterygians have both left and right
spines as continuations of both halves of the neural
arch in the abdominal region. However, advanced
teleosts (e.g., ostariophysans and euteleosts; see Ar-
ratia, 1999: character 81) have both halves fused
medially starting early in ontogeny. The spinal cord
is protected dorsally by one or two large supradorsal
cartilages that form a bridge over the cord in all
fishes with separate halves of the neural arch and
spine. The supradorsal cartilage ossifies as part of
the dorsal part of the canal formed by the neural
arch in fishes with fused halves.
The neural spines of the caudal region are fused to
their arches in all adult actinopterygians, with the
exception of polypteriforms and acipenseriforms.
There is a controversy concerning the identification
of the elements above the neural arch in polypteri-
forms and acipenseriforms that we plan to present
elsewhere.
Supraneurals, ‘supraneurals,’ and radials.
Supraneural or ‘supraneural’ bones are unknown in
placoderms, chondrichthyans, and acanthodians.
Cartilaginous supraneurals are present in chon-
drichthyans and through most of the life of acipens-
eriforms; nevertheless, they can partially ossify in
late ontogeny of the latter group.
‘Supraneural’ bones as a continuous series along
the vertebral column are found in recent dipnoans
(e.g., see Fig. 1) and in some fossil dipnoans,
where they are probably fused with the neural
spines (Fig. 27A,B). Among actinopterygians the
situation is very diverse. For instance, a continu-
ous series of supraneural bones was restored in
†Birgeria groenlandica by Nielsen (1949, fig. 77).
The presence of a continuous series of supraneu-
rals is unclear for both polypteriforms and aci-
penseriforms.
The elements extending until the end of the tail in
Polypterus bichir were labeled as neural spines by
Jarvik (1981, fig. 238); in contrast, they were labeled
as supraneurals by Bartsch and Gemballa (1992,
figs. 3B,C, 7) in their illustrations of P. senegalus.
According to Gardiner (1984), supraneurals are
missing in Polypterus; the only element present is
the neural spine (after Gardiner, 1984).
Fig. 47. Anterior vertebrae of the placoderm †Erromenosteus diensti from the Upper Devonian of Bad Wildungen, Germany (MB.
f.73135); Ca1⫹2, Ca6⫹7, compound vertebrae, each consisting of arcual elements belonging to two adjacent trunk metameres;
Da9⫹10, compound dorsal arcual element consisting of the dorsal arcual elements of the trunk metameres 9 and 10; Da17-19, dorsal
arcual elements of trunk metameres 17 to 19; md, median dorsal plate; Va15-19, ventral arcual of trunk metameres 15 and 19.
158 G. ARRATIA ET AL.
In acipenseriforms, Woodward (1895), Goodrich
(1930), MacAlpin (1947), and Jollie (1962), among
others, named these elements neural spines. How-
ever, Grande and Bemis (1991:30) preferred to in-
terpret them as supraneurals “following Patterson
(1973) and Patterson and Rosen (1977) because
these elements in acipenseriforms do not form part
of the neural arch (basidorsals) as do true neural
spines.…Also these elements are all median in aci-
penseriforms; and true median neural spines appear
to be derived for halecostomes (amiiforms plus te-
leosts). But see above for Moythomasia,Australoso-
mus,Birgeria, and others with only one median
neural spine. The figure of Neoceratodus in Rosen et
al. (1981, fig. 57) with an autogenous median neural
element labeled neural spine was labeled that way
in error, and the element is actually a supraneural
(Patterson, pers. comm.).” (For contrary evidence,
see descriptions of dipnoan neural spines and ‘supra-
neurals’ above and Figs. 1–6, 10, 11, 17A–C, 21A,B.)
However, our observations of young specimens of
Polyodon show that the so-called supraneural is still
joined to one of the halves of neural arches 7 and 8
(Fig. 38A), whereas all other so-called supraneurals
are separated, with the exception of the elements in
the caudal region. The neural arch and the so-called
supraneural form incomplete, single cartilages or
large single cartilages in the caudal region (Figs.
38C, 41A). The separation of the single cartilage into
the basidorsal cartilage and the so-called supraneu-
ral progresses anteriorly in the caudal region, and
apparently the process of formation of new arcualia
is a permanent process in the caudal tip of the body.
Starting from the assumption of a hypothetical
series of supraneurals extending along the entire
vertebral column, Boreske (1974) considered the
epurals in Amia as relics of an originally continuous
row of supraneurals and thus as equivalents of the
few anterior supraneurals observed in Amia and
chondrosteans. Bartsch (1988) considered this the
correct interpretation, but neither of these authors
has demonstrated the presence of a continuous se-
ries of supraneurals in primitive bony fishes, includ-
ing acanthodians and primitive actinopterygians
and sarcopterygians. Our present results do not sup-
port this theoretical interpretation.
In most actinopterygians the series of supraneu-
ral bones is short. It extends from the first vertebrae
to the first dorsal radials (e.g., †Mimia, Amia, †Ca-
turus, †Eurycormus, †Leptolepis, †Anaethalion,
Elops, Hiodon) or it can be shorter (e.g., Denticeps),
or even absent (e.g., Basilichthys, Odontesthes).
Homology. Disagreements exist over the homol-
ogy of neural spines, supraneurals, and radials. For
instance: 1) Supraneural bones in fishes were pro-
posed as the homologs of median neural spines by
Goodrich (1930) and Eaton (1945). 2) Teleostean
supraneural bones were proposed as serial homologs
of radials by Smith and Bailey (1961). In contrast,
Mabee (1988) proposed that supraneural bones in
fishes are not serial homologs of median neural
spines, or of radials. Goodrich (1930) based his prop-
osition on a survey of data from a variety of teleos-
tomes and tetrapods. In contrast, Smith and Bailey’s
(1961) proposition was mainly based on certain te-
leosts. Mabee’s (1988) conclusions were based on
data that were mainly provided by ontogenetic se-
ries of advanced teleosts (perciforms) and a review of
the literature concerning some teleostomes (e.g., ac-
anthodians, some actinopterygians, and †Eusthe-
nopteron). In addition, she tested the hypothesis of
serial homology between median neural spines and
supraneurals in a phylogeny taken from Rosen et al.
(1981) and Patterson (1973) (Mabee, 1988, fig. 1).
We believe that the disagreement lies in the erro-
neous conclusion that the same structure develops
in the same manner in different fishes. However, as
we have shown above, the vertebral column in fishes
develops differently in different fish groups. We will
now analyze the hypotheses proposed by previous
authors in light of the new and broader ontogenetic
information that we present above, and we will test
the results in a phylogeny.
Early ontogenetic stages of the living dipnoans
Lepidosiren, Protopterus, and Neoceratodus (Figs.
10A–D, 11A,B, 21A–C) show that the ‘supraneural’
bone in these fishes develops as continuation of the
neural spine. Both neural spine and ‘supraneural’
originate in early ontogeny from one unique carti-
lage that later separates into independent elements.
Furthermore, this is also the origin of the radial.
Therefore, neural spine, ‘supraneural,’ and radial
are linear homologs from their origin. In contrast,
supraneural and radial develop from independent
cartilages (Figs. 42B, 43A,B, 44A–D) (e.g.,
Polypterus: Bartsch and Gembala, 1992; Amia:
Schultze and Arratia, 1989; Grande and Bemis,
1998; teleosts: Mabee, 1988) in actinopterygians
where early ontogenetic stages have been studied
(acipenseriforms seem to be an exception). (In
Polypterus the situation is curious. A continuous
series of cartilages interpreted as supraneurals ex-
tends along the body in a larva of 17.1 mm total
length, but each one becomes articulated with the
neural arch in specimens of 21 mm on [Bartsch and
Gembala, 1992, figs. 2C,B, 3A–C].)
The neural spine may develop as a continuation of
the dorsal arcuale (e.g., Amia, Elops) or it may result
as a dorsal membranous ossification of the dorsal
arch (e.g., catfish Diplomystes, perciform Percich-
thys). Additionally, the neural spines, supraneurals,
and radials may have different directions of appear-
ance and a different direction of ossification (e.g.,
Mabee, 1988; Schultze and Arratia, 1989; herein).
Our results show that both Goodrich (1930) and
Mabee (1988) are partially correct in their conclu-
sions of the origin of supraneural bones. Mabee’s
(1988) conclusion is correct for actinopterygians and
Goodrich’s (1930) for sarcopterygians. Conse-
quently, our results support once more (see also
159VERTEBRAL COLUMN IN DIPNOANS
Arratia and Schultze, 1990, 1991) the idea that
fishes may reach similar terminal stages by using
different ontogenetic processes and/or sequences. It
is a serious error to base conclusions on the assump-
tion that teleostomes have the same development of
structures that have been previously assumed to be
homologous.
We now test our observations and results in an
accepted phylogeny (see Fig. 48). According to our
observations and review of the literature, supraneu-
ral bones are absent in placoderms and unknown in
acanthodians. Chondrichthyans do not have supra-
neural bones, but have supraneural cartilages.
Among teleostomes, supraneural bones are known
in actinopterygians and ‘supraneural’ bones are
known in sarcopterygians; however, the information
is missing or is very incomplete for most primitive
actinopterygians (e.g., †Dialipina, †Cheirolepis, and
most palaeonisciforms) as well as for most primitive
sarcopterygians (e.g., †Psarolepis, †Onychodus,
†Youngolepis, †Powichthys). The information pro-
vided by members of both lineages, Sarcopterygii
and Actinopterygii, shows that a series of ‘supran-
eural’ bones is known in fossil and extant dipnoans
(see for instance Figs. 4A,B, 10A–C, 13, 17A,B,
23A–D) and in the rhipidistian †Glyptolepis. The
‘supraneural’ is probably fused with the radial in
certain dipnoans (Fig. 28A,B) and in actinistians
(Figs. 29C, 30). Small elements articulated with the
neural spines of the most anterior vertebrae have
been identified as ‘supraneurals’ in the rhipidistian
†Eusthenopteron (Fig. 34F–G). A series of supraneu-
rals is known in most actinopterygians; its absence
in some of the most advanced teleosts is interpreted
here as a secondary loss.
The so-called supraneural bones form in different
ways in both lineages: The ‘supraneurals’ develop
from one single cartilage that also forms the neural
spine and the radial in sarcopterygians. The three
elements articulate with each other. In contrast, the
supraneural develops independently from both the
neural spine and the radial in actinopterygians.
Primitively, the three series (radials – supraneurals
– neural spines) develop from three different series
of cartilages in actinopterygians.
Because we do not know the supraneural condi-
tion in the most primitive members of both lineages
(Fig. 48) and, in addition, the condition is unknown
Fig. 48. Gnathostome phylogeny and distribution of neural spines (NS), so-called supraneural (SU), and radial (RA) bones. The
series formed by neural spine / supraneural bone / radial originates from different cartilages; the series formed by neural spine 3
supraneural bone 3and radial originates from the segmentation of one cartilage into three elements.
160 G. ARRATIA ET AL.
in acanthodians, it is more parsimonious to inter-
pret the supraneurals of actinopterygians and of
sarcopterygians as independently acquired in both
lineages. Consequently, the supraneural bones of
actinopterygians are not homologs with the ‘supra-
neural’ bones of sarcopterygians.
Neural Arches and Epineural Processes
The epineural process is a lateral or posterolat-
eral process of the neural arch. Short and broad
epineural processes are known in some fossil dip-
noanssuchas†Uranolophus and †Griphognathus
(Figs. 22A,B, 25A), but they are absent in other
fossil and living dipnoans (for distribution of this
character, see Fig. 49). In addition, they are un-
known in other sarcopterygians studied here. Be-
cause short, broad processes are found in certain
dipnoans that are not closely related to each other,
the presence of this structure can be considered an
autapomorphy of these forms (e.g., †Uranolophus,
†Griphognathus).
A process of the neural arch, also termed epineu-
ral process, is known from some actinopterygians
such as the Triassic †Prohalecites, some Jurassic
‘pholidophoriforms,’ and teleosts above the level of
†Leptolepis coryphaenoides.
The so-called epineural processes of †Prohalecites
and some ‘pholidophoriforms’ are short, dorsolateral
processes of the neural arches of the abdominal ver-
Fig. 49. Dipnoan phylogeny and distribution of certain features of the vertebral column and associated elements (phylogeny after
Schultze and Marshall, 1993).
161VERTEBRAL COLUMN IN DIPNOANS
tebrae. In contrast, those of †Leptolepis corypha-
enoides and more advanced teleosts are long, fine
processes, which are wrongly named in the litera-
ture epineural bones without being independent
bones. Nevertheless, in some teleosts fine elongate
bones are articulated or attached to the neural arch;
they are correctly named the epineural bones. For
differences in terminology see Arratia (1997:127–
128, 1999:301).
According to the distribution of the so-called
epineural processes in some sarcopterygians and in
advanced actinopterygians, the one present in some
fossil dipnoans and those found in certain unrelated
actinopterygians are not homologous. It is assumed
here that they have arisen independently in these
forms because there is no evidence that epineural
processes were present in basal sarcopterygians,
basal actinopterygians, and in their sister-group,
the acanthodians.
Ribs
In polypteriforms there are two distinct kinds of
ribs, dorsal ribs in the horizontal septum, and ven-
tral or pleural ribs in the wall of the coelom. This
pattern gave the basis for the identification of the
two types of ribs (e.g., Devillers, 1954).
The homologization of ribs using a comparative
topographical criterion fails because in actinoptery-
gians and chondrichthyans the ribs vary in position
and there is variation from anterior to posterior
direction in the same fish. Emelianov (1935, 1936)
set a developmental criterion to compare and homol-
ogize ribs, and in addition he also considered the
relationship of the cartilaginous anlagen to the hor-
izontal and transverse septa. Furthermore, ribs in
some actinopterygians such as Polypterus and te-
leosts develop centrifugally from cartilages close to
the vertebra, whereas their dorsal ribs develop cen-
tripetally from cartilaginous anlage beneath the lat-
eral line. Following such criteria, the ribs of the
dipnoans and many actinopterygians are ventral
ribs.
The relation between the base of the rib and the
notochord or the centrum varies in teleostomes. In
certain primitive Devonian dipnoans such as
†Dipterus (Fig. 23A–D), the pleural ribs articulate
with parapophyses that lie ventrolateral to the per-
sistent notochord. In others such as †Griphognathus
(Fig. 25A,B) each parapophysis is fused to the ven-
trolateral side of the compact vertebral centrum and
each parapophysis bears an articular surface inter-
preted as for articulation with the ribs (Campbell
and Barwick, 1988); however, ribs have not been
reported from †G. whitei. In younger dipnoans (Figs.
27A,B, 28A,B), parapophyses are not observed and
the ribs are attached to the persistent notochord. We
assume that the condition in these fishes was simi-
lar to that found in extant dipnoans in which the
base of the rib (unlike in other teleostomes) includes
remnants of the ventral arcuale that may ossify into
a small ventral arcocentrum. This is interpreted
here as a synapomorphy shared by dipnoans such as
†Scaumenacia, †Fleurantia, †Sagenodus, †Concho-
poma, †Paraceratodus, and extant dipnoans.
Among sarcopterygians outside the dipnoans
there are some differences. For instance, the ribs
may be comparatively short in some actinistians and
absent in others, including Latimeria. Short and
massive ribs are supposed to be present in †Eusthe-
nopteron; they articulate with parapophyses placed
dorsally in the haemal arch of posterior abdominal
vertebrae (Figs. 34F,G, 36B) (Andrews and Westoll,
1970b; Jarvik, 1981). However, there is variation of
this pattern (see Figs. 35B, 36A) and the so-called
ribs may be elongate dorsoposterior processes of the
haemal arch. Independently of what the condition is
in †Eusthenopteron, the short and massive ribs, dor-
sally placed in parapophyses of the haemal arches of
the abdominal vertebrae, do not resemble the ribs
known in other sarcopterygians.
A pair of cranial ribs is known from some fossil
and living dipnoans. They are apparently absent in
†Uranolophus and other dipnoans such as †Gripho-
gnathus and †Gnathorhiza. The feature could be
considered a synapomorphy of dipnoans above the
level of †Uranolophus with several losses at differ-
ent phylogenetic levels (for distribution of this char-
acter, see Fig. 49).
Fin Rays and Fins
The camptotrichia of unpaired fins of living dip-
noans are thin, weak structures incompletely ossi-
fied and with different degrees of development
among the different genera, as noted by Gu¨ nther
(1871), Fu¨ rbringer (1904), Goodrich (1904), and oth-
ers. Among the three living genera, Lepidosiren has
the shortest, thinnest, and less developed rays. They
show a reduced and irregular segmentation (Fig.
15B,C), and the number of rays between radials is
commonly lower (ca. 1–4) than in Protopterus (ca.
4–5) and Neoceratodus (ca. 4–6). The structure of
the camptotrichia resembles the rays present in the
Permian dipnoan †Conchopoma gadiforme and the
Early Triassic dipnoan †Paraceratodus germaini
(compare Figs. 3, 7A, and 28A,B). In contrast, Devo-
nian dipnoans such as †Uranolophus, †Dipterus,
†Griphognathus, †Scaumenacia, and †Fleurantia
have well-ossified segmented and branched lepi-
dotrichia (compare Figs. 3, 7A, 28A,B, and 24A).
The Devonian dipnoans have independent dorsal,
caudal, and anal fins (e.g., Figs. 24A, 27A,B). In
contrast, Late Paleozoic (e.g., †Sagenodus) and
younger forms (Fig. 28A,B), as well as living dipno-
ans, have a dorsal and a ventral series of rays that
are separated by the postcaudal cartilages (in living
forms). Consequently, there does not exist a true
caudal fin in these forms and externally there is no
landmark to indicate a limit between dorsal and
162 G. ARRATIA ET AL.
caudal rays. Then the question still remains: Is
there a caudal fin in living dipnoans? Anatomically
there is no evidence of a true caudal fin, but a long
series of dorsal rays that phylogenetically corre-
sponds to dorsal plus caudal rays and a ventral
series that corresponds to anal plus caudal rays.
The fins and rays of dipnoans show some well-
defined evolutionary trends: Middle Devonian forms
have independent dorsal, caudal, and anal fins pos-
sessing lepidotrichia. Late Devonian forms still have
the three fins and the epaxial rays of the caudal fin
are short. Younger forms, e.g., from Carboniferous
to Recent, have lost the three independent fins and
there is only one series of dorsal and one series of
ventral rays that in living forms are called campto-
trichia to differentiate them from other rays.
Acanthodians, early actinopterygians, and sarcop-
terygians have independent dorsal, caudal, and anal
fins, the presence of which corresponds to the plesi-
omorphic condition. The condition found in living
adult dipnoans (no independent fins) could be inter-
preted as a retention of the early ontogenetic condi-
tion, but it represents a new stage where dorsal and
anal fins reach close to the end of the body. They are
separated by the postcaudal cartilages (gephyrocer-
cal according to Abel, 1911). The caudal fin proper
has been lost. Consequently, it can be interpreted as
a unique feature that is probably shared by various
dipnoans such as †Sagenodus, †Conchopoma,
†Paraceratodus, Neoceratodus, Protopterus, and
Lepidosiren (phylogenetic arrangement follows that
of Schultze and Marshall, 1993). In parallel, the
presence of camptotrichia should probably be a char-
acter shared by all of these forms. The presence of
lepidotrichia in different sarcopterygians and in ac-
tinopterygians would be a feature shared by sarcop-
terygians and actinopterygians, but not by ac-
anthodians with dermotrichia.
Apparently, the presence of actinotrichia is the
primitive condition in the evolution of fin rays (pro-
posed first by Goodrich, 1904), at least for chondrich-
thyans and teleostomes. According to Ge´ raudie and
Meunier (1980) the distal unossified portions of the
dermotrichia of acanthodians might be actinotrichia
and the fine structure of the chondrichthyan cerato-
trichia is reminiscent of actinotrichia.
Pedomorphosis and Dipnoans
Pedomorphosis, “the retention of ancestral juve-
nile characters by later ontogenetic stages of descen-
dants” (Gould, 1977:484), is important in the evolu-
tion of lungfishes (Bemis, 1980, 1982, 1984).
Previous authors such as Moy-Thomas and Miles
(1971), Gardiner (1973), and Smith (1977) consid-
ered that certain features of dipnoans exhibit pedo-
morphic tendencies but did not analyze the evidence
and the implications for Dipnoi as a whole. “In cases
of paedomorphosis, an understanding of the com-
plete ancestral ontogeny is essential.” (Bemis, 1984:
294). Obviously, this understanding for dipnoans is
severely affected by the fact that most dipnoans are
fossil (with a long history beginning in the Devoni-
an), and they are represented by only three living
genera.
Information on early ontogenetic stages of fossil
dipnoans is scarce and unsatisfactory. For instance,
Westoll (1936), Forster-Cooper (1937), and White
(1965) described certain aspects of growth in
†Dipterus. Lund (1970) reported juvenile †Sageno-
dus, and Schultze (1977) described juveniles of
†Megapleuron. Certainly, the information about the
early ontogeny of fossil dipnoans is minimal and
most of the results can be inferred by comparison of
large fossil dipnoans with young and adult living
dipnoans.
Bemis (1984) documented several features that he
considered evidence of pedomorphosis in lungfishes.
The features are: 1) loss of a heterocercal tail; 2)
fusion of median fins; 3) reduction of fin rays; 4) loss
of cosmine; 5) change in scale shape; 6) reductions in
ossification; and 7) increase in cell size. We will
analyze features 1–3 and 6, which are related to the
subject of this contribution, and test them in the
most recent phylogeny of lungfishes (Schultze and
Marshall, 1993).
Loss of a heterocercal tail. Changes in the
shape of the median fins is one of the most common
examples of change during dipnoan evolution (noted
first by Dollo, 1895). The external shape of the tail is
heterocercal in Devonian genera such as †Uranolo-
phus, †Dipterus, and †Griphognathus (Figs. 24A, 49:
Nodes A–D) or reduced heterocercal in †Scaumena-
cia and †Fleurantia (Figs. 27A,B, 49: Nodes E, F). In
contrast, Late Paleozoic forms such as the Carbon-
iferous genus †Sagenodus (Schultze and Chorn,
1997: fig. 19), the Permian †Conchopoma (Fig. 28A)
and the Triassic †Paraceratodus have a gephyrocer-
cal tail, a condition also shown by the living genera
Neoceratodus, Protopterus, and Lepidosiren (Figs. 3,
6, 7A, 15B, 49: Nodes G–K). Abel (1911:112, table)
distinguished fins after function, external shape,
and internal morphology. He used the term gephy-
rocercal for the above-described tail; the caudal fin
proper has been lost in his interpretation.
According to Bemis (1984), evidence from the on-
togeny of the tail supports the pedomorphic inter-
pretations. Early in the ontogenetic development of
gnathostomes, the notochord is straight and the tail
is protocercal. Recent actinopterygians, where the
development is known, show that early in ontogeny
the caudal end of the notochord bends dorsally. Dor-
sal bending of the notochord and formation of the
hypocordal lobe does not occur during ontogeny of
living lungfishes (see above), and therefore the tail
retains the protocercal shape in adults (Bemis,
1984). Unfortunately, early ontogeny of one of the
most primitive actinopterygians, †Cheirolepis,isun-
known and in †Dialipina the tail is triphycercal and
there is no bending. The evidence provided for
163VERTEBRAL COLUMN IN DIPNOANS
adults varies; the tail is heterocercal in †Cheirolepis,
whereas it is triphycercal in †Dialipina (Schultze, in
press). In primitive Actinistia, the notochord does
bend and the tail is heterocercal, e.g., †Miguashaia
(Schultze, 1975a; Cloutier, 1996) and †Gavinia
(Long, 1999).
According to the distribution of the heterocercal
tail (and its loss) in the cladogram (Fig. 49), the
presence of a straight notochord in the adult tail and
loss of a caudal fin proper in dipnoans is a synapo-
morphic character shared by †Sagenodus and more
advanced dipnoans (see Fig. 49: Node G), not a re-
tention of an ancestral juvenile character.
Fusion of median fins. The position and number
of median fins are variable in dipnoans. Devonian
dipnoans have two dorsal fins of variable size, one
anal fin, and a caudal fin (e.g., Figs. 24A, 27A,B, 49:
Nodes A–F). However, one dorsal and one ventral
median fin are present in later forms (e.g., Figs. 3,
28A,B, 49: Nodes G–K). Comparison of the primitive
actinopterygian †Dialipina and actinistians shows
that the presence of two dorsal and one anal fins is
plesiomorphic. Primitive dipnoans show the plesi-
omorphic condition in contrast to Late Paleozoic and
younger lungfishes.
The presence of the “continuous” median fin in
dipnoans has been interpreted as a result of loss of
the caudal fin proper (Abel, 1911), or of a fusion of
originally separate fins (e.g., Goodrich, 1930), or as
the retention of the embryonic condition (dorsal–
caudal–anal fin fold) (Bemis, 1984). According to
Goodrich (1930) the fusion can have two interpreta-
tions: posterior growth of the dorsal and anal fins or
forward growth of the caudal fin. We have not seen
either type of fusion during the ontogeny of living
dipnoans. Furthermore, due to the separation of the
rays by the notochord into a dorsal and a ventral
series of rays (which are missing at the posterior
tip), we are uncertain whether living dipnoans have
caudal rays. The structure of the caudal region of
adult living dipnoans is different from the structure
of the dorso–caudal–anal finfold in early ontogeny of
living actinopterygians and actinistians. Therefore,
the presence of a dorsal and a ventral series of rays
separated completely by the notochord (and inter-
preted by others as a fusion of dorsal, caudal, and
anal fins) in dipnoans cannot be considered a pedo-
morphic feature. The distribution of this feature in
the cladogram shows that this is an apomorphic
character shared by †Sagenodus and younger dipno-
ans (Fig. 49: Nodes G–K).
Reduction of fin rays. Living dipnoans have
fine, unossified fin rays or camptotrichia compared
to fossil dipnoans with lepidotrichia. Both halves of
a pair of camptotrichia are broadly separated from
each other (Fig. 16A,B) and each half shows a soft
external surface, with the exception of some rays in
large specimens showing incomplete segmentation
(Fig. 15B,C). The inner structure of camptotrichia
differs from that of lepidotrichia (Fig. 16C–E).
Devonian dipnoans have lepidotrichia (e.g., Figs.
23A,B, 24A, 27A,B, 49: Nodes A–F). However,
younger forms such as †Sagenodus, †Conchopoma,
and †Paraceratodus (e.g., Fig. 28A,B) show rays that
resemble camptotrichia of living dipnoans. Among
the living forms there is a trend to diminish the
number of rays of the median fins, from numerous
(and longer) in Neoceratodus to less numerous (and
shorter) in Lepidosiren.
Bemis (1984) suggested that fin rays of the Dipnoi
are homologous to those of other teleostomes but
that those of modern dipnoans show pedomorphic
features. The differences in structure of campto-
trichia in living dipnoans from that of actinotrichia
cannot be considered pedomorphic, but a novelty in
the evolution of more advanced forms (Fig. 49:
Nodes G–K).
Reduction in ossification. We agree with Bemis
(1984:301) that dipnoans exhibit phylogenetic de-
creases in the ossification of endochondral and ex-
oskeletal bones of the skull, pectoral girdle, snout,
visceral skeleton, and ossification of vertebral ele-
ments. As mentioned above, some Paleozoic dipno-
ans show ossified vertebral centra (e.g., Figs. 24,
25A,B, 49: Node D1), whereas in others only well-
ossified dorsal and ventral arcocentra are present
(Figs. 22A,B, 49: Node A). In contrast, younger fossil
dipnoans do not have ossified centra but a persistent
notochord in adult individuals (e.g., Figs. 4A, 7A,
17A, 27A,B, 28A,B, 49, Nodes E–K). The cladogram
(Fig. 49) shows that the retention of the persistent
notochord and absence of ossified centra are features
in dipnoans from Late Paleozoic to Recent.
CONCLUSIONS
1. A vertebral column formed by a persistent no-
tochord without vertebral centra is the primi-
tive pattern for all vertebrates.
2. The formation of centra is acquired indepen-
dently in different lineages such as placoderms,
some primitive and most advanced actinoptery-
gians, and some dipnoans and rhipidistians and
thus is not homologous among vertebrate
groups.
— The chordacentrum of chondrichthyans is a
specialization of that group.
— Some Paleozoic dipnoans have a compact
centrum or holocentrum, whereas others
have only a persistent notochord like living
dipnoans. An autocentrum is not present in
living dipnoans. A compact centrum or holo-
centrum represents an apomorphic charac-
ter shared by some Devonian lungfishes such
as †Soederbergia and †Griphognathus.
— A persistent notochord represents the prim-
itive condition in lungfishes (e.g., †Uranolo-
phus).
— An autocentrum, as a direct ossification
formed outside the elastica externa, is a nov-
164 G. ARRATIA ET AL.
elty found in advanced actinopterygians,
e.g., †Leptolepis coryphaenoides and more
advanced teleosts (Arratia, 1999).
— A chordacentrum, as a direct mineralization
and/or calcification of the fibrous sheath of
the notochord, is also a novelty found only in
actinopterygians; however, it is not possible
to establish where the chordacentrum arose
first because of lack of information concern-
ing many primitive actinopterygians.
3. Living dipnoans have only basidorsal and ba-
siventral arcualia early in ontogeny. Interdorsal
and interventral arcualia are irregularly present
in Neoceratodus and, occasionally, in some indi-
viduals of Protopterus. Basidorsal, basiventral,
interdorsal, and interventral arcualia are
present in the actinistian Latimeria, some rhi-
pidistians, and different actinopterygians.
4. In Paleozoic dipnoans as well as in most teleos-
tomes, the pleural rib articulates with a parap-
ophysis or ventral arcocentrum that develops
from a basiventral arcuale. The base of each rib
attaches to a cavity on the lateroventral wall of
the notochord in living dipnoans, unlike in other
teleostomes. The base of the pleural rib of living
adult dipnoans may include remnants of the
basiventral arcuale surrounded by a small arco-
centrum.
5. Ontogenetic studies of different major taxa of
fishes show once more that the assumption that
members of one group show similar ontogenetic
origin and sequence of development of struc-
tures as members of another is misleading. For
instance, the origin of the neural spine, so-called
supraneural, and dorsal radial differs between
sarcopterygians and actinopterygians.
— The neural spine, ‘supraneural,’ and dorsal
radial result as dorsal growth and distal dif-
ferentiation of one median cartilage (origi-
nated from the basidorsal arcuale) into two
(neural spine and ‘supraneural’ in the ante-
rior part of the vertebral column) and three
(neural spine, ‘supraneural,’ and dorsal
radial in the middle caudal region of the
vertebral column) bones in living dipnoans.
In contrast, the neural spine originates from
distal growth of the basidorsal arcuale and
the supraneural and dorsal radial from dif-
ferent cartilage series in actinopterygians.
— The ‘supraneural’ of sarcopterygians and the
supraneural of actinopterygians are consid-
ered nonhomologous because of their differ-
ent origin and formation.
— The haemal spine, interhaemal bone, and
ventral radial of one median cartilage (orig-
inated from the basiventral arcuale) devel-
ops into three bones (haemal spine, interh-
aemal, and ventral radial) in living
dipnoans.
6. The presence of ossified neural spines and arco-
centra characterizes placoderms and teleos-
tomes and represents the primitive condition for
gnathostomes. In contrast, calcified neural
spines and chordacentra are found in chondrich-
thyans. Because of their distribution, these fea-
tures are considered apomorphic for chondrich-
thyans among Gnathostomata.
7. Independent dorsal, caudal, and anal fins are
primitive for acanthodians, early actinoptery-
gians, and sarcopterygians (as well as for placo-
derms and chondrichthyans).
8. Living dipnoans—as well as Carboniferous and
younger fossil dipnoans—do not have indepen-
dent unpaired fins, and a caudal fin proper is
lost. The series of dorsal rays may be inter-
preted as phylogenetically formed of dorsal plus
caudal rays and the ventral series of anal plus
caudal rays. Ontogenetic information does not
clarify the origin of both series of rays.
9. Actinotrichia are present in the early ontogeny
of living dipnoans, of other teleostomes, and also
of chondrichthyans.
10. Recent dipnoans have camptotrichia in contrast
to Devonian dipnoans with lepidotrichia. Lepi-
dotrichia are found in primitive sarcopterygians
and actinopterygians, but not in acanthodians
(with dermotrichia). Lepidotrichia are a feature
shared by primitive sarcopterygians and acti-
nopterygians. Camptotrichia probably were al-
ready present in post-Devonian dipnoans.
Camptotrichia and actinotrichia have a differ-
ent structure. Therefore, camptotrichia cannot
be considered a pedomorphic feature. We hy-
pothesize that they are a novelty in the evolu-
tion of dipnoans and a feature shared by living
and probably post-Devonian dipnoans.
ACKNOWLEDGMENTS
For gift of specimens we thank especially A. Kemp
(Brisbane, Queensland, Australia), K.W.S. Camp-
bell (The Australian National University, Cam-
berra, Australia), and Mr. Peter Bru¨ hn (Essen, Ger-
many). For loans of specimens we thank M.
Arsenault (Parc de Miguasha, Que´ bec, Canada), H.
Bjerring, E. Jarvik, S. Kullander (Naturhistoriska
Rikmuseet, Stockholm, Sweden), C.H. von Daniels
(Bundesanstalt fu¨ r Geowissenschaften und Rohstoffe,
Hannover, Germany), W. Eschmeyer and D. Catania
(California Academy of Sciences, San Francisco, Cal-
ifornia), W.L. Fink and D. Nelson (University of
Michigan, Ann Arbor), L. Grande (Field Museum of
Natural History, Chicago, Illinois), H. Jahnke (In-
stitut und Museum fu¨ r Geologie und Pala¨ ontologie,
Georg-August Universita¨t,Go¨ttingen, Germany), M.
Louette and G. Teugels (Muse´ e Royal de l’Afrique
Centrale, Tervuren, Belgium), H.-J. Paepke (Insti-
tut fu¨ r Systematische Zoologie, Museum fu¨r
Naturkunde, Berlin, Germany), R. Rosenblatt (Uni-
165VERTEBRAL COLUMN IN DIPNOANS
versity of California, La Jolla, California), W. Saul
(Academy of Natural Sciences of Philadelphia,
Pennsylvania), J.D. Stewart (Los Angeles County
Museum, Los Angeles, California), A. Tintori (Uni-
versita` delgi Studi di Milano, Italy), G. Viohl (Jura
Museum, Eichsta¨ tt, Germany), P. Wellnhofer (Bay-
erische Staatssammlung fu¨ r Pala¨ ontologie und his-
torische Geologie, Mu¨ nchen, Germany), E.O. Wiley
and K. Shaw (Natural History Museum, The Uni-
versity of Kansas, Lawrence). P. Bartsch (Museum
fu¨ r Naturkunde, Berlin, Germany) shared with us
his specimen of Protopterus dolloi. Mr. J.-P. Mendau
(Museum fu¨ r Naturkunde, Berlin, Germany) pre-
pared the final illustrations based on the original
drawings by G. Arratia; Mrs. E. Siebert prepared
the cover illustration based on Figure 21B and Mrs.
W. Harre prepared the photographs. Mr. Roberto
Bustamante (Argentina) was crucial in collecting
living Lepidosiren paradoxa of different sizes.
LITERATURE CITED
Abel O. 1911. Palaeobiologie. Stuttgart: E. Schweizerbart’sche
Verlag (E. Na¨gele).
Agassiz L. 1833–1844. Recherches sur les Poissons Fossiles. Neu-
chaˆtel and Soleure: Petitpierre.
Ahlberg PE, Trewin NH. 1995. The postcranial skeleton of the
Middle Devonian lungfish Dipterus valenciennesi. Trans R Soc
Edinb Earth Sci 85:159–175.
Andrews SM. 1977. The axial skeleton of the coelacanth, Latim-
eria. In: Andrews SM, Miles RS, Walker AD, editors. Problems
in vertebrate evolution. London: Academic Press. p 271–288.
Andrews SM, Westoll TS. 1970a. The postcranial skeleton of
rhipidistian fishes excluding Eusthenopteron. Trans R Soc Ed-
inb 68:391–489.
Andrews SM, Westoll TS. 1970b. The postcranial skeleton of
Eusthenopteron foordi Whiteaves. Trans R Soc Edinb 68:307–
329.
Applegate SP. 1967. A survey of shark hard parts. In: Gilbert PW,
Mathewson RF, Rall DP, editors. Sharks, skates and rays.
Baltimore: Johns Hopkins University Press. p 37–69.
Arratia G. 1991. The caudal skeleton of Jurassic teleosts; a phy-
logenetic analysis. In: Chang M-m, Liu H, Zhang G, editors.
Early vertebrates and related problems in evolutionary biology.
Beijing: Science Press. p 249–282.
Arratia G. 1997. Basal teleosts and teleostean phylogeny. Palaeo
Ichthyol 7:5–168.
Arratia G. 1999. The monophyly of Teleostei and stem-group
teleosts. Consensus and disagreements. In: Arratia G, Schultze
H-P, editors. Mesozoic fishes — systematics and fossil record.
Mu¨nchen: Verlag Dr. Pfeil. p 265–334.
Arratia G, Cloutier R. 1996. Reassessment of the morphology of
Cheirolepis canadensis (Actinopterygii). In: Schultze H-P, Clou-
tier R, editors. Devonian fishes and plants of Miguasha, Que-
bec, Canada. Mu¨nchen: Verlag Dr. Pfeil. p 165–197.
Arratia G, Lambers P. 1996. The caudal skeleton of pachycormi-
forms: parallel evolution? In: Arratia G, Viohl G, editors. Me-
sozoic fishes — systematics and fossil record. Mu¨nchen: Verlag
Dr. Pfeil. p 191–218.
Arratia G, Schultze H-P. 1990. The urohyal: development and
homology within osteichthyans. J Morphol 203:247–282.
Arratia G, Schultze H-P. 1991. Palatoquadrate and its ossifica-
tions: development and homology within osteichthyans. J Mor-
phol 208:1–81.
Arratia G, Schultze H-P. 1992. Reevaluation of the caudal skel-
eton of certain actinopterygian fishes. III. Salmonidae. Homolo-
gization of caudal skeletal structures. J Morphol 214:187–249.
Ax P. 1987. The phylogenetic system. The systematization of
organisms on the basis of their phylogenesis. New York: John
Wiley & Sons.
Bartsch P. 1988. Funktionelle Morphologie und Evolution des
Axialskelettes und der Caudalis urspu¨nglicher Knochenfische.
Palaeontographica A 204:117–226.
Bartsch P. 1989. Some remarks on the axial skeleton in Pro-
topterus (Pisces, Dipnoi). Gegenbaurs Morphol Jb 135:419–426.
Bartsch P, Gemballa S. 1992. On the anatomy and development
of the vertebral column and pterygiophores in Polypterus
senegalus Cuvier, 1829 (“Pisces”, Polypteriformes). Zool Jb
Anat 122:497–529.
Bartsch P, Gemballa S, Piotrowski T. 1997. The embryonic and
larval development of Polypterus senegalus Cuvier, 1829: its
staging with reference to external and skeletal features, behav-
iour and locomotory habits. Acta Zool (Stockh) 78:309–328.
Baudelot ME. 1873. Observations sur la structure et le de´veloppe-
ment des nageoires des poissons osseux. Arch Zool Exp Ge´n
2:87–244.
Baum KA, Lund R. 1974. Vertebral centra in Haplolepis (Haplo-
lepidae, Paleonisciformes) from the Allegheny group, Pennsyl-
vanian. J Paleontol 48:199–200.
Becerra J, Montes GS, Bexiga SRR, Junqueira LCU. 1983. Struc-
ture of the tail fin in teleosts. Cell Tiss Res 230:127–137.
Bemis WE. 1980. Pedomorphosis and the evolution of lungfishes
(Dipnoi). Am Zool 20:757.
Bemis WE. 1982. Studies on the functional and evolutionary
morphology of lepidosirenid lungfish. PhD Diss. Univ Calif,
Berkeley.
Bemis WE. 1984. Pedomorphosis and the evolution of the Dipnoi.
Paleobiology 10:293–307.
Bendix-Almgren SE. 1975. The paired fins and shoulder girdle in
Cladoselache, their morphology and phyletic significance. Coll
Intern CNRS, Paris, Proble`mes actuels de Pale´ ontologie (E
´vo-
lution des Verte´bre´ s):111–123.
Bordat C. 1987. E
´tude ultrastructurale de l’os des verte`bres du
Se´lacien Scyliorhinus canicula L. Can J Zool 65:1435–1444.
Boreske JR. 1974. A review of the North American fossil amiid
fishes. Bull Mus Comp Zool Harv Uni 146:1–87.
Bouvet J. 1974. Diffe´rentiation et ultrastructure du squelette
distal de la nageoire pectorale chez la truite indige`ne (Salmo
trutta fario L.) I. Diffe´rentiation et ultrastructure des acti-
notriches. Arch Anat Microsc Morphol Exp 63:79–96.
Bridge TW. 1896. The mesial fins of Ganoids and Teleosts. Zool J
Linn Soc 25:530–602.
Bryan WL. 1919. On the structure of Eusthenopteron. Bull Buff
Soc Nat Sci 13:1–59.
Campbell KSW, Barwick RE. 1988. Uranolophus: a reappraisal of
a primitive dipnoan. In: Jell PA, editor. Dipnoan and Carbon-
iferous fish studies. Assoc Australas Paleontol Mem 7:87–144.
Campbell KSW, Barwick RE. 1999. Dipnoans fishes from the Late
Devonian Gogo Formation of Western Australia. Rep Aust Mus
Suppl 57:107–138.
Cloutier R. 1996. Dipnoi (Akinetia: Sarcopterygii). In: Schultze
H-P, Cloutier R, editors. Devonian fishes and plants of
Miguasha, Quebec, Canada. Mu¨nchen: Verlag Dr. Pfeil. p 198 –
226.
Cloutier R, Forey P. 1991. Diversity of extinct and living actinis-
tian fishes (Sarcopterygii). In: Musick JA, Bruton MN, Balon
EK, editors. The biology of Latimeria chalumnae and evolution
of coelacanths. Environ Biol Fishes 32:59–74.
Cuvier G. 1829. Le Re`gne Animal distribue´ d’apre`s son organisa-
tion, pour servir de base a` l’histoire naturelle des animaux, et
d’introduction a` l’anatomie compare´ e. Paris: Hippolyte Tilliard.
Dalquest WW, Kocurko MJ, Grimes JV. 1989. Aspects of the
postcranial skeleton of the Lower Permian lungfish, Gnatho-
rhiza. J Paleontol 63:919–930.
Dean B. 1909. Studies on fossil fishes (Sharks, Chimaeroids, and
Arthrodires). Am Mus Nat Hist Mem 9:211–287.
Denison RH. 1968a. Early Devonian lungfishes from Wyoming,
Utah and Idaho. Field Geol 17:353–413.
Denison RH. 1968b. The evolutionary significance of the earliest
known lungfish, Uranolophus. In: Ørvig T, editor. Current
166 G. ARRATIA ET AL.
problems of lower vertebrate phylogeny. Nobel Symposium 4.
Stockholm: Almqvist & Wiksell. p 247–257.
Devillers C. 1954. Les coˆtes. In: Grasse´ PP, editor. Traite´de
Zoologie, vol. 12. Paris: Masson. p 673–697.
Dollo L. 1895. Sur la phyloge´nie des Dipneustes. Bull Soc Belge
Ge´ol Pale´ ontol Hydrol 9:79–128.
Eaton TH. 1945. Skeletal support of the median fins of fishes. J
Morphol 76:193–212.
Emelianov SW. 1935. Die Morphologie der Fischrippen. Zool Jb
(Anat) 60:132–262.
Emelianov SW. 1936. Die Morphologie der Tetrapodenrippen.
Zool Jb (Anat) 62:173–274.
Forster-Cooper C. 1937. The middle Devonian fish fauna of
Achanarras. Trans R Soc Edinb 59:223–239.
Franc¸ois Y. 1966. Structure et de´ veloppement de la verte`bre de
Salmo et des Te´le´oste´ ens. Arch Zool Exper Ge´n 107:283–328.
Fu¨rbringer K. 1904. Beitra¨ ge zur Morphologie des Skeletes der
Dipnoer, nebst Bemerkungen u¨ber Pleuracanthiden, Holoceph-
alen und Squaliden. Denkschr Med Nat Ges Jena 4:423–510.
Gadow H. 1896. On the evolution of the vertebral column of
Amphibia and Amniota. Philos Trans R Soc Lond B 187:1–57.
Gadow H, Abbott EC. 1895. IV. On the evolution of the vertebral
column of fishes. Philos Trans R Soc Lond B 166:163–221.
Gardiner BG. 1973. Interrelationships of teleostomes. In: Green-
wood PH, Miles RS, Patterson C, editors. Interrelationships of
fishes. Zool J Linn Soc Suppl 1:105–135.
Gardiner BG. 1984. The relationships of the palaeoniscid fishes, a
review based on new specimens of Mimia and Moythomasia
from the Upper Devonian of Western Australia. Bull Brit Mus
(Nat Hist) Geol 37:173–428.
Gegenbaur C. 1867. U
¨ber die Entwickelung der Wirbelsa¨ule des
Lepidosteus mit vergleichend-anatomischen Bemerkungen.
Jena Z Med Nat 3:359–419.
Ge´raudie J. 1984. Fine structural comparative peculiarities of the
developing dipnoan dermal skeleton in the fins of Neoceratodus
larvae. Anat Rec 209:115–123.
Ge´raudie J, Landis WJ. 1982. The fine structure of the developing
pelvic fin dermal skeleton in the trout Salmo gairdneri.AmJ
Anat 163:141–156.
Ge´raudie J, Meunier F-J. 1980. Elastoidin actinotrichia in coela-
canth fins: a comparison with teleosts. Tissue Cell 12:637–645.
Ge´raudie J, Meunier F-J. 1982. Comparative fine structure of the
osteichthyan dermotrichia. Anat Rec 202:325–328.
Goette A. 1878. Beitra¨ge zur vergleichenden Morphologie des
Skeletsystems der Wirbelthiere. Arch Mikrosk Anat 15:442–
641.
Goodrich ES. 1904. On the dermal fin-rays of fishes — living and
extinct. Q J Microsc Sci 47:465–522.
Goodrich ES. 1909. Vertebrata Craniata. I. Cyclostomes and
fishes. In: Lankaster R, editor. A treatise on zoology. London:
Adam & Charles Black.
Goodrich ES. 1930. Studies on the structure and development of
vertebrates, vol. 2. New York: Dover.
Gould SJ. 1977. Ontogeny and phylogeny. Cambridge: Harvard
Univ Press.
Graham-Smith W, Westoll TS. 1937. On a new long-headed dip-
noan fish from the Upper Devonian of Scaumenac Bay, P.Q.,
Canada. Trans R Soc Edinb 59:241–246.
Grande L, Bemis WE. 1991. Osteology and phylogenetic relation-
ships of fossil and Recent paddlefishes (Polyodontidae) with
comments on the interrelationships of Acipenseriformes. J Vert
Paleontol 11 Suppl Mem 1:1–121.
Grande L, Bemis WE. 1998. A comprehensive phylogenetic study
of amiid fishes (Amiidae) based on comparative skeletal anat-
omy. An empirical search for interconnected patterns of natural
history. J Vert Paleontol 18 Suppl Mem 4:1–690.
Grassi B. 1882. Beitra¨ge zur na¨ heren Kenntniss der Entwicklung
der Wirbelsa¨ule der Teleostier. Morphol Jb 8:457– 473.
Gregory WK, Rockwell H, Evans FG. 1939. Structure of the
vertebral column in Eusthenopteron foordi Whiteaves. J Pale-
ontol 13:126–129.
Gu¨nther A. 1871. Description of Ceratodus, a genus of Ganoid
fishes, recently discovered in Rivers of Queensland, Australia.
Philos Trans R Soc Lond 166:511–571.
Hasse JC. 1879-1885. Das natu¨rliche System der Elasmo-
branchier auf Grundlage des Baues und der Entwicklung ihrer
Wirbelsa¨ule. Jena.
Hasse JC. 1893. Die Entwicklung der Wirbelsa¨ule der Dipnoi. Z
wiss Zool 55:533–542.
Hay OP. 1895. On the structure and development of the vertebral
column of Amia. Field Columbian Ms Publ 5 Zool Ser 1:54.
Hay OP. 1897. Dr. Gadow and Miss Abbottt on the vertebral
column of fishes. Zool Bull 1:131–141.
Heyler D. 1962. Les Acanthodiens et le proble`me de
l’aphe´tohyoı¨die. Coll Intern Prob Pale´ ontol 104:39–47.
Hitchcock EC. 1995. A functional interpretation of the anterior-
most vertebrae and skull of Eusthenopteron. Bull Mus Nat Hist
nat, Paris, 4
e
se´r 17, Sec C 1-4:269 –285.
Janvier P. 1981. The phylogeny of the Craniata, with particular
reference to the significance of fossil “agnathans.” J Vert Pale-
ontol 1:121–139.
Jarvik E. 1952. On the fish-like tail in the Ichthyostegid Stego-
cephalians with descriptions of a new Stegocephalian and a new
Crossopterygian from the Upper Devonian of East Greenland.
Meddr Grønland 111:1–90.
Jarvik E. 1981. Basic structure and evolution of vertebrates. New
York: Academic Press.
Johanson Z, Ritchie A. 2000. A new Late Famennian lungfish
from New South Wales, Australia, and its bearing on
Australian-Asian terrane relations. Alcheringia 24:99–118.
Jollie M. 1962. Chordate morphology. New York: Reinhold.
Kamel AH. 1953. On the development of the vertebral column in
Gambusia affinis. Bull Fac Sc 32:95–115.
Kemp NE. 1977. Banding pattern and fibrillogenesis of cerato-
trichia in shark fins. J Morphol 154:187–203.
Kemp A. 1999. Ontogeny of the skull of the Australian lungfish,
Neoceratodus forsteri (Osteichthyes: Dipnoi). J Zool 248:97–
137.
Kerr JG. 1909. Normal plates of the development of Lepidosiren
paradoxa and Protopterus annectens. In: Keibel F, editor. Nor-
mentafeln zur Entwicklungsgeschichte der Wirbeltiere. Part
10:1–31. Jena: Verlag Gustav Fischer.
Klaatsch H. 1895. Beitra¨ge zur vergleichenden Anatomie der
Wirbelsa¨ule. II. U
¨ber die Bildung knorpeliger Wirbelko¨rper bei
Fischen. Morphol Jb 20:143–186.
Ko¨lliker A. 1860. U
¨ber das Ende der Wirbelsa¨ule der Ganoiden
und einiger Teleostier. Leipzig: Wilhelm Engelmann.
Laerm J. 1979. The origin of the rhipidistian vertebrae. J Pale-
ontol 53:175–186.
Lambers P. 1992. On the ichthyofauna of the Solnhofen Litho-
graphic Limestone (Upper Jurassic, Germany). PhD Thesis.
Univ Gro¨ningen, Holland.
Lehman J-P. 1959. Les Dipneustes du De´vonian supe´ rieur du
Groenland. Meddr Grønland 164:1–58.
Lehman J-P 1966. Actinopterygii. In: Piveteau J, editor. Traite´de
Pale´ontologie, vol. 4. Paris: Masson et Cie. p 1–242.
Lehman J-P, Chateau C, Laurin M, Nauche M. 1959. Pale´ontolo-