Content uploaded by Donald J Stewart
Author content
All content in this area was uploaded by Donald J Stewart on Aug 10, 2015
Content may be subject to copyright.
BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research
libraries, and research funders in the common goal of maximizing access to critical research.
Re-description of Arapaima agassizii (Valenciennes), a Rare Fish from Brazil
(Osteoglossomorpha: Osteoglossidae)
Author(s): Donald J. Stewart
Source: Copeia, 2013(1):38-51. 2013.
Published By: The American Society of Ichthyologists and Herpetologists
DOI: http://dx.doi.org/10.1643/CI-12-013
URL: http://www.bioone.org/doi/full/10.1643/CI-12-013
BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and
environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published
by nonprofit societies, associations, museums, institutions, and presses.
Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of
BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use.
Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial
inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.
Re-description of Arapaima agassizii (Valenciennes), a Rare Fish from Brazil
(Osteoglossomorpha: Osteoglossidae)
Donald J. Stewart
1
The bony-tongue fish genus Arapaima Mu¨ller has been considered monotypic since 1868, with A. gigas being the only
recognized species. Review of species-level taxonomy of Arapaima has revealed that Arapaima agassizii Valenciennes (in
Cuvier and Valenciennes, 1847) should be considered a valid species. The holotype was destroyed in World War II, but
the species can be recognized based on the original description, which included detailed osteological illustrations. At
least nine characters distinguish it from all other Arapaima: 1) dentary teeth 44 (counted on one ramus only, vs. 21–37 in
other Arapaima); 2) maxillary teeth 43 (vs. 21–38 in other Arapaima); 3) orbit diameter 1.5%standard length (SL, vs. 1.5–
2.8, relatively larger in all other Arapaima at similar SL); 4) interorbital width 4.1%SL (vs. 5.3–6.5 in other Arapaima); 5)
parietals with pronounced posterior projections that are pointed and curve slightly toward midline (vs. absent in other
Arapaima); 6) caudal fin widely separated from dorsal and anal fins by long caudal peduncle, 9.7%SL (vs. much shorter
peduncle, 3.2–5.5 in others); 7) anal fin with only 26 rays (vs. 30–40 in others), with distinctly shorter basal length than
dorsal-fin base; 8) dorsal and anal fins extremely low in profile; dorsal-fin base divided by longest dorsal ray about 7 (vs.
3.1–5.5 in others); and 9) first pectoral-fin ray with proximal tip similar in form to subsequent pectoral-fin rays (vs. first
pectoral-fin ray noticeably enlarged relative to subsequent rays). Arapaima agassizii still is known only from the
holotype, which was collected in 1817–20 somewhere in lowlands of the Brazilian Amazon. It thus is important to locate
this taxon to determine its distribution and conservation status.
A
RAPAIMA gigas (Schinz, in Cuvier, 1822) is consid-
ered to be one of the most distinctive, readily
recognized fishes in tropical South America (Ferraris,
2003). That species also is the only freshwater fish in South
America that is listed as an endangered species in CITES,
Appendix II, but the IUCN (2011) considers it to be ‘‘Data
Deficient.’’ Arapaima are threatened, in part, because of their
large size (.3 m total length, see Stone, 2007), fine eating
quality, and relative ease of capture (Castello and Stewart,
2010; Castello et al., 2011). Because of this special situation,
my students and I have been studying the ecology and
conservation status of Arapaima in both Guyana (Watson,
2011) and the central Amazon (e.g., Castello et al., 2011), and
those field observations raised questions about taxonomic
status of previously described Arapaima from those two
drainage basins. Review of the early literature (e.g., Cuvier,
1816, 1822; Spix and Agassiz,1829; Schomburgk, 1841; Cuvier
and Valenciennes, 1847; Gu¨nther, 1868) and holotypes for
two species of Arapaima in Muse´um national d’Histoire
naturelle, Paris, indicated that all four species of Arapaima
recognized by Valenciennes (in Cuvier and Valenciennes,
1847) should be considered valid. Those inferences contrast
sharply with the uncritical synonymy of Gu¨nther (1868), who
offered neither analysis nor rationale. There is no evidence
that Gu¨nther examined type materials in Paris or Munich.
Here I reconsider the taxonomic status of A. agassizii, which
was described by Valenciennes on the basis of the osteological
illustrations and text in Spix and Agassiz (1829). Those
illustrations allow for a more expansive morphological
analysis of A. agassizii, while comparable details are not
available for the otherspecies. For that reason, re-descriptions
of A. gigas, A. mapae, and A. arapaima will be presented
elsewhere, but data needed for a clear differential diagnosis of
A. agassizii are included here.
Recent observations on morphological variation in refer-
ence populations of Arapaima from Mamiraua´ Reserve,
Brazil, and the Rupununi River basin, Guyana, provide a
comparative framework for evaluating those morphometric
and meristic features of A. agassizii that appear to be
diagnostic for the species. Examination of the holotypes of
A. gigas and A. mapae (both dried and stuffed mounts in
MNHN, Paris) revealed several external characters that can
distinguish those nominal species from A. agassizii. Finally,
studies on evolutionary relationships of osteoglossomorph
fishes (e.g., Taverne, 1977; Li and Wilson, 1996; Hilton,
2003) as well as my own observations on several Arapaima
skeletons provide a foundation for evaluating osteological
characteristics of A. agassizii. The holotype and only known
specimen of A. agassizii was lost when most of the fish
collection of the Zoologische Staatssammlung Mu¨ nchen,
Germany, was destroyed during World War II (Kottelat,
1988, and pers. comm.). Fortunately, that loss is compen-
sated by finely detailed osteological drawings made under
the direction of L. Agassiz (in Spix and Agassiz, 1829; for an
English translation, see Pethiyagoda and Kottelat, 1998).
I have examined nearly every specimen of Arapaima in
museums of the world (i.e., France, England, United States,
Brazil, Guyana, Ecuador, and Peru´ ) and have failed to locate
a second specimen of A. agassizii. Recognition of this
interesting fish potentially could facilitate its rediscovery
in the field. Ultimately, that could lead to efforts to conserve
remaining populations, which potentially could be threat-
ened. Specimens of Arapaima preserved in museums are so
few in number, compared to the vast geographic range of
the genus, that there is insufficient basis for judging
conservation status of this species (i.e., presently, negative
data are not meaningful). The same can be said for A. gigas
and A. mapae, both of which are still known only by their
holotypes. It is urgent to look more closely at variation
among remaining populations of Arapaima because overex-
ploitation and illegal harvests are a continuing problem
(Castello and Stewart, 2010). For widely distributed, non-
migratory fishes like Cichla (e.g., Kullander and Ferreira,
2006) and Arapaima, it is reasonable to expect that the
Amazon basin harbors additional diversity awaiting discov-
ery (e.g., Lundberg et al., 2000; Piorski et al., 2008).
1
Department of Environmental and Forest Biology, College of Environmental Science and Forestry, State University of New York, 1 Forestry
Drive, Syracuse, New York 13210; E-mail: djstewart@esf.edu.
Submitted: 28 January 2012. Accepted: 24 September 2012. Associate Editor: R. E. Reis.
F2013 by the American Society of Ichthyologists and Herpetologists DOI: 10.1643/CI-12-013
Copeia 2013, No. 1, 38–51
MATERIALS AND METHODS
Comparative materials.—Analyses of the taxonomy of Ara-
paima involve meristic, morphometric, osteological, molec-
ular, and/or live-color observations on a wide size range of
individuals from Mamiraua´ Reserve in central Brazil (e.g., C.
Arantes, unpubl. data, morphometrics for 28 individuals)
and additional observations of individuals from Guyana
(e.g., Watson, 2011, and unpubl. data, including morpho-
metrics for 45 individuals). Because of logistical constraints,
most large individuals from Mamiraua´ and Guyana were
measured in the field and not retained. Other non-type
museum specimens studied were from about 33 localities in
the Amazon basin (n575, from between Bele´m, Brazil, and
Pucallpa, Peru´ ). Further analyses of those data will be
published elsewhere, but they provide an initial perspective
on morphological variation among populations of Ara-
paima, and selected data are integrated into the primary
diagnosis of A. agassizii by the phrase ‘‘distinguished from
all other.’’ Institutional abbreviations are as listed at http://
www.asih.org/node/204, and the Centre for the Study of
Biological Diversity at the University of Guyana, George-
town, is ‘CSBD/UG.’ Anatomical abbreviations for the
figures mostly follow Hilton (2003) and Taverne (1977).
Type materials.—Type specimens examined for A. gigas and
A. mapae are as follows: status of the holotype for A.
arapaima is also discussed.
Arapaima gigas: holotype (also 5holotype of Vastres cuvieri,
a junior synomym): MHNH a-8837, 203 cm standard length
(SL), dried and stuffed specimen, Brazil, Para´ State, ‘‘Villa de
Santare´m,’’ 2.41693uS, 54.69361uW, A. R. Ferreira, between
1783 and 1787 (Ferreira, 1903; Bleher, 2006:360). An
attached label gives date of 1808, which should be when
this specimen was moved from Lisbon, Portugal, to Paris.
Arapaima gigas is still known only from its holotype.
Arapaima mapae: holotype: MHNH a-8836, 203 cm SL, dried
and stuffed specimen, ‘‘lac Mapa, sur les confins des
nouvelles frontie`res de la Guyane franc¸aise,’’ purchased by
C. Pradier, a French naval officer. An attached label gives
date of 1837. During most of the 1800s, the area between
the Araguari and Oyapock (Oiapoque, in Portuguese) rivers
was disputed by Brazil and France, and in that period, there
was the village of Mapa and, 9.5 km to the east, lac Mapa
(2.03677uN, 50.71047uW). After Brazil annexed the region in
1900, that area was incorporated into what is now Amapa´
State; the village and lake apparently were renamed Amapa´
and Lago do Amapa´ (also sometimes called Lago Grande).
Arapaima mapae is still known only from its holotype.
Arapaima arapaima (Valenciennes, in Cuvier and Valenci-
ennes, 1847): holotype: BMNH 2009.1.19.1, 246 cm total
length, dried and stuffed specimen, British Guiana, Sir
Robert H. Schomburgk (from his expeditions of 1835–1839).
This specimen was mentioned in Schomburgk (1841, quoted
by Jardin, ‘‘The specimen, the skin of which I brought with
me, and which is now in the possession of the British
Museum, measured, when taken, eight feet one inch . . . .’’)
and by Valenciennes (in Cuvier and Valenciennes, 1847,
within the description of A. arapaima, ‘‘M. Schomburgk took
an eight-foot long individual of it in the Rupununi.’’). This
large specimen was on public display at BMNH from about
1905 to 1979 and was moved to storage after that.
Unfortunately, a recent, extensive search has failed to locate
it (O. Crimmen, BMNH, pers. comm.). Ferraris (2003) noted
that there were ‘‘no types known’’ for A. arapaima. It seems
there was at least a known holotype; only recently it was
catalogued as a missing specimen. At BMNH, there also is a
partial dry skeleton: BMNH 2006.10.12.2, 170 cm SL, British
Guiana, Sir Robert H. Schomburgk (label has no date). Given
uncertainty about the date that this specimen was collected
and/or received at BMNH, its possible status as a type is also
uncertain (i.e., if it were received after 1841, it could not be a
type). The skeleton was not mentioned in Schomburgk
(1841) or Valenciennes (in Cuvier and Valenciennes, 1847),
and the vertebral count reported by Schomburgk is notably
lower than that of the skeleton. If the holotype is not found,
a neotype from the Rupununi River basin in Guyana will be
designated in a subsequent paper that re-describes this
species; that particular specimen is listed below under ‘non-
type voucher specimens’ and, for various comparisons, it is
referenced here as a ‘topotype.’
Non-type voucher specimens.—Arapaima arapaima (Valenci-
ennes, in Cuvier and Valenciennes, 1847): topotype, CSBD/
UG 1667, 104 cm SL, alcohol specimen, Guyana, Rupununi
River basin, Grass Pond, ,4.5 km SW from Rewa village,
going upstream along Rewa River, 3.86769uN, 58.76678uW,
L. C. Watson and D. J. Stewart, 12 August 2006.
Arapaima sp. incertae sedis: all of following from Mamiraua´
Sustainable Development Reserve, near Comunidade Jaraua´,
Amazonas State, Brazil, approx. 65.000uW, 2.833uS, C. Arantes
(see Castello, 2008:fig. 1): INPA 26580, approx. 160 cm SL (tail
was missing), dry skull, Parana´doJaraua´, close to Comuni-
dade Jaraua´, 13 November 2006; INPA 26581, 126 cm SL,
complete dry skeleton, Lago Samau´ ma, 14 November 2006;
INPA 26582, 5, 61.9–73.5 cm SL, 4 alcohol specimens and 1
complete dry skeleton, Ressaca do Curuc¸a´, 17 November
2006; INPA 26583, 3, 86.6–112 cm SL, 2 alcohol specimens
and 1 complete dry skeleton, Lago Samaumeirinha do Jaraqui,
25 November 2006; INPA 26584, 96.8 cm SL, alcohol
specimen, Lago Cedrinho, 24 November 2006.
Morphometric and meristic characters.—Measurements for A.
agassizii (other than those for the basibranchial toothplate)
were taken from digital images of the illustrations in Spix
and Agassiz (1829), so some could only be approximated
because of perspectives in the drawings. Standard length was
measured from tip of upper jaw to mid-point between bases
of posterior-most caudal-fin rays, because there is no
hypural plate illustrated. Caudal peduncle length is from
posterior end of anal-fin base to mid-point of caudal-fin
base. Head length is from tip of upper jaw to tip of opercular
flap, which I assumed to be at posterior margin of cleithrum
(the typical condition in Arapaima; but nonetheless, that
would only be an approximation); post-orbital distance was
measured from posterior margin of bony orbit to that same
point. Tooth counts for the dentary are for a single ramus; so
total number of teeth on the lower jaw is double the values
reported. All tooth counts include gaps where openings in
the skin indicated a missing tooth. Measurements other
than those from the drawings were made with digital
calipers to nearest mm, or for long measurements on large
fishes, to nearest 0.5–1.0 cm using a tape measure.
Spix and Agassiz (1829) gave the length of the type as
‘‘more than 3 feet’’ but never gave the exact length. To
provide approximate scale bars on the figures, I assumed a
total length of 1.0 m (539.4 in). Uncertainty in total length
Stewart—Arapaima agassizii re-description 39
might bias morphometric proportions where there is strong
allometry (e.g., orbit diameter), but relatively isometric
proportional relations would be affected less (e.g., interor-
bital width and basibranchial toothplate length). Improving
on what was possible in this context will require finding
another specimen.
A difficulty in evaluating a fish taxon known from a single
specimen is in developing statistical support for the
conclusion that various morphometric features may be
sufficiently different that they could be considered diagnos-
tic. Further uncertainties arise because many characters have
notable allometric changes with fish length. The solution
for this problem was to obtain measurement data for a wide
size range of Arapaima from reference populations, and in
this case, comparisons focus on a sample from Mamiraua´
Reserve, Brazil. Comparisons with a reference sample from
Guyana give similar results. Taxonomic status of the
Mamiraua´ and Guyanese populations is a complex problem
involving characteristics A. arapaima and beyond the scope
of this analysis. Individual characters were evaluated using
bivariate plots of body proportions versus SL, and the 95%
confidence ellipse was computed for values of the reference
population (using software package PAST Ver. 2.13; Hammer
et al., 2001). If the corresponding value for the holotype of
A. agassizii fell well outside that confidence ellipse, then it
could be inferred that it was different from the reference
population. Likewise, if values for specimens of the other
three nominal species fell within (or near) the 95%
confidence ellipse or outside in the opposite direction, it
might be inferred that they also were different from A.
agassizii. It is not possible, however, to compute a statistical
probability that two individual type specimens are different;
only indirect inferences can be made about such differences.
A second approach to this problem is to compute a linear
regression of each character versus SL for the reference
population. Then for a case like the holotype of A. agassizii
that was not included in the regression, it is possible to
compute an outlier statistic for the null hypothesis that the
holotype conforms to the fitted regression (i.e., what is the
probability that the holotype belongs to the reference
population, see Cook and Weisberg, 1982; analysis available
in software package Statistix 7 for Windows; http://www.
statistix.com). These two approaches are conceptually at
odds, because the first assumes no dependent variable while
the latter assumes dependent and independent variables. I
ran the analyses both ways and, for the four characters
presented graphically below and some of the other charac-
ters included in the diagnosis, results were consistent in
showing A. agassizii to be significantly different.
Arapaima agassizii (Valenciennes, in Cuvier and
Valenciennes, 1847)
Figures 1B–D, 3–8 (parts)
Holotype.—Vastres agassizii Valenciennes, in Cuvier and
Valenciennes, 1847, based on the description and illustra-
tions of Spix and Agassiz (1829:31–40, pl. 16, anatomy pl. B,
skeleton, and C, scales only). As noted by Valenciennes, the
color illustration in plate 16 (labeled Sudis pirarucu) of that
same work (Fig. 1B) appears to be a composite drawing
including some details from the original illustration pub-
lished by Cuvier (1816) for what is now Arapaima gigas
(Fig. 1A, see comments below). The type specimen of V.
agassizii was collected between 1817 and 1820 somewhere in
the Brazilian Amazon (Spix and Agassiz, 1929). Spix traveled
from the mouth of the Amazon to Tabatinga, Brazil, some
3,000 km upstream near the Peruvian border; he also
ascended the Rio Negro to Barcelos (Tiefenbacher, 1983).
The type was deposited in the Zoologische Staatssammlung
Mu¨nchen, Munich, Germany, but was destroyed as noted
above; there apparently was no catalogue number.
The name Sudis pirarucu is not available because it was first
published in the synonymy of Sudis gigas; that name first
appeared in Cuvier (1829, March) as a name only, and then
in Spix and Agassiz (1829, June) along with the description
of what also was considered S. gigas by those authors (see
Eschmeyer and Fricke, 2011). So S. pirarucu is a nomen nudum
attributed to Cuvier (1829).
Diagnosis.—Distinguished from all other Arapaima by follow-
ing nine characters or character complexes (morphometric
proportions 5%SL): 1) dentary teeth 44 (counted on single
ramus only, vs. A. mapae 37, A. gigas 30, A. arapaima 22–37,
and Arapaima sp. from Mamiraua´ 21–36, Fig. 2B); 2) maxil-
lary teeth 43 (vs. A. mapae 38, A. gigas 26, A. arapaima 21–36,
and Arapaima sp. from Mamiraua´ 22–35); 3) orbit diameter
1.5, relatively small compared to similar-sized individuals (vs.
A. mapae 1.5, A. gigas 1.7, A. arapaima 1.7–2.8, and Arapaima
sp. from Mamiraua´ 1.6–2.8, Fig. 2C); 4) interorbital width 4.1,
relatively narrow (vs. A. mapae 5.3, A. gigas 6.2, A. arapaima
5.6–6.5, and Arapaima sp. from Mamiraua´ 5.8–6.3, Fig. 2D);
5) parietals with pronounced posterior projections that are
pointed and curve slightly toward the midline (Figs. 3A, 4D;
vs. exposed posterior margin of skull roof in other Arapaima
straight to moderately concave, Fig. 4A); those projections
have rugose striations like on much of skull roof; 6) caudal fin
widely separated from dorsal and anal fins by relatively long
caudal peduncle, 9.7 (Figs. 1C, 5; vs. A. mapae 3.3, A. gigas 3.2,
A. arapaima 4.2–5.5, and Arapaima sp. from Mamiraua´ 3.9–
5.3, with posterior tips of dorsal and anal fins in other
Arapaima overlapping caudal-fin base when depressed);
caudal-peduncle length divided by peduncle depth 2.4 (vs.
A. mapae 0.81, A. gigas 0.60, A. arapaima 0.65–0.87, and
Arapaima sp. from Mamiraua´ 0.65–0.83); 7) anal fin with only
26 rays, with distinctly shorter basal length than that of
dorsal fin (Fig. 1C; vs. anal fin with 30–40 rays in other
Arapaima, and just slightly shorter in basal length than
dorsal-fin base; note that original illustrations of A. gigas
[Fig. 1A] and Sudis pirarucu [Fig. 1B] have anal fins incorrectly
drawn, see comments below); 8) dorsal and anal fins
extremely low in profile (Fig. 6); dorsal-fin base divided by
longest dorsal-fin ray about 7 (Fig. 1C; vs. 3.1–5.5 in other
Arapaima; higher value is for A. gigas); longest dorsal-fin ray in
anterior third of fin (vs. in center or posterior half of fin in
other Arapaima); 9) first pectoral-fin ray not (or only slightly)
enlarged in diameter and with proximal tip similar in form to
second and subsequent pectoral-fin rays (Fig. 7C; vs. rela-
tively enlarged in diameter with proximal tip extended
medially relative to subsequent pectoral-fin rays, Fig. 7E).
Further distinguished from A. gigas by having single row of
small teeth on dentary (vs. 2–2.5 rows of enlarged teeth in A.
gigas).
Characters 10–13 cannot be evaluated in holotypes of A.
gigas and A. mapae, but clearly distinguish A. agassizii from
other Arapaima that have been illustrated in the literature
(cited above) or studied by me. 10) An extremely long
anterior basibranchial toothplate (i.e., the bony tongue
[Figs. 2A, 8] 18; vs. A. arapaima 9.3–10.4, Arapaima sp. from
40 Copeia 2013, No. 1
Mamiraua´ 9.6–10.5, typically isometric at ,10 in other
Arapaima; also see Martinelli and Petrere, 1999; that tooth-
plate is missing from type of A. gigas, and it was not possible
to measure in type of A. mapae); extending posteriorly to
ceratobranchial of fourth gill arch, expanded posteriorly
such that lateral margins of toothplate are distinctly
concave (Fig. 8A; vs. not extending beyond hypobranchial
of third gill arch in other Arapaima, and typically slightly
constricted at posterior tip, with lateral margins straight to
slightly convex, Fig. 8C); a relatively large, rounded tooth-
plate is situated farther posterior, over bases of fourth and
fifth ceratobranchials (Fig. 8A; vs. in other Arapaima, a
separate toothplate between bases of third and fourth arches
and two or more tiny tooth patches associated with bases of
fifth ceratobranchials). 11) Postcleithrum absent or fused to
the cleithrum (Fig. 7A–C; vs. well-developed, separate
element in other Arapaima, Fig. 7D). 12) Caudal fin with
ten branched rays, of which five rays each articulate with a
single neural spine of the posterior vertebral centra and five
rays each articulate with a single haemal spine of those same
five posterior centra (Fig. 5; see Comments on diagnostic
features, below). 13) Neural and haemal spines just anterior
to those supporting caudal-fin rays much shorter than first
spines that do support caudal-fin rays (Fig. 5; vs. spines just
anterior to those supporting caudal-fin rays similar in length
or only slightly shorter, Fig. 5C).
Description.—Morphometric (as %SL) and meristic characters
not presented elsewhere are as follows: predorsal-fin length
66; dorsal-fin base length 30; longest dorsal-fin ray 4.3;
preanal-fin length 68; anal-fin base length 22; longest anal-
fin ray 4.4; pectoral-fin length 12; caudal-peduncle depth,
approximately 4.0; caudal-peduncle length 9.7; head length
21; head depth at occiput 11; head depth between eyes 6.4;
snout length 4.1; preorbital distance 2.1; fourth infraorbital
length 10; postorbital length 16; dorsal-fin rays ii,34, 36 total;
anal-fin rays i,25, 26 total; gill rakers (upper limb, first arch)
15; gill rakers (lower limb, first arch) more than 23 (some
hidden in the illustration); total premaxillary teeth 36.
Translation of original description.—Words in brackets are my
additions or, in some cases, reflect uncertainty about the
precise translation from the French of Valenciennes (in
Cuvier and Valenciennes, 1847). ‘‘Compared to the two
previous [species, i.e., V. cuvieri 5gigas and V. mapae], it
appears to me to resemble them by the general proportions
of the body. The head also equals one fifth of the total
length; but what distinguishes it, is that the third and fourth
suborbitals, covering the cheek more completely, didn’t
[have indications of] indentations posteriorly; that the pits
to the mucous cavities of these two suborbitals are equal
between them, more round and bigger than those of the
previous [species]. The skull is proportionally a lot wider.
Fig. 1. (A) Cuvier’s (1816) illustration of Sudis ge´ ant; (B) illustration of Sudis pirarucu (5Sudis gigas) from Spix and Agassiz (1829), based on Spix’s
field observations and, perhaps, of the same specimen that is the holotype for A. agassizii; (C) osteological illustration (supervised by Agassiz) of
holotype for A. agassizii from the latter work; and (D) reconstruction of the holotype of A. agassizii based on superficial features of an Arapaima
photographed by me in Guyana and modified to fit the skeleton. Scale bars 525 cm.
Stewart—Arapaima agassizii re-description 41
The temporal cavities are more remote and larger; the
parietal and the occipital smaller, from which it results that
the bony surface, formed by combination of the frontal and
parietal, is longer. The transverse furrows, that are ahead of
the interparietal depression, appear bigger to me and more
numerous. The vomerine and [para-] sphenoidal teeth trace
a lot wider band [initially], but get narrower at the extremity
of the pterygoids, because these are a lot wider. The lingual
bone forms a large oval plate, a bit narrower at the insertion
of the first and second gill arches, but widens again below
the big [para-] sphenoidal plate.
All this description is made based on the figure of the
skeleton that M. Agassiz drew with great care, and which he
gave a description as detailed as rigorous. To the characters
that I have just picked for shapes of some parts of the skull, I
will add again that the opercle appears wider and rounder to
me; that the preopercle has cavities more equal among
them, and finally, that the striations of all these bones tend
to be more parallel, which recalls Vastres Cuvieri (sic) [5A.
gigas] a lot more than our second species [5A. mapae]. An
external character that distinguishes the new species readily
from the two that I previously established, I chose the much
larger extent of the dorsal and the more considerable length
of the pectoral; it is about one third of the length of body,
the caudal not included [i.e., dorsal-fin base length]; the
anal is, on its side, a lot shorter; the pectoral only has one
ninth of the total length; it is one eleventh in the other
species. Here are the numbers:
Branchiostegals 10; Dorsal 36; Anal 26; Caudal 10;
Pectoral 12.
M. Agassiz counted eighty vertebrae in the backbone, of
which thirty-eight are abdominal and carry some ribs. The
anterior vertebrae of this trunk don’t resemble in any way
those of the erythrins; because the Webberian [sic] ossicles,
if they existed, would not have escaped notice by the
meticulous observer who described this skeleton, and the
drawing shows that there is no trace of them.
After having established the characters of the species, I
refer my readers back to the detailed description of M.
Agassiz, which appears quite useless for me to reproduce
here in whole.
To judge some colors by the painting that Spix gave to this
Vastre`s, one must believe that the back of the fish is reddish
brown, and the stomach is white more or less smudged with
gray; top of the head is brown; fins are reddish.
We can also speak of specific characters of the scales,
because M. Agassiz took the care to illustrate two of them:
they have on their naked portion more elevated scabrosites
and therefore deeper [rivulations] than do those of our
other two species [see Fig. 4E, F]. The skeleton kept in the
Museum of Munich is more than three feet long; but M.
Martius observed that fish attains a length of more than
five feet, and that it is not rare to find some individuals
who weigh several hundred kilograms. The lingual bone of
the skeleton of Munich is seven inches long and an inch
and a half wide. Among the bony plates kept for a long
time in the King’s Cabinet, we find some that perfectly
correspond, by their shape, to the one from the Munich
skeleton that appeared in the work quoted: it has eight
inches of length, like that from Vastres Mapae (sic) of six
Fig. 2. Morphometric features of A. agassizii, holotype (asterisks), compared to a population sample of Arapaima sp. from Mamiraua´ Reserve, Brazil
(dots), and comparable values for A. gigas (G), A. mapae (M), and a topotype of A. arapaima (A). For basibranchial toothplate length, observations
were not available for A. gigas or A. mapae, and value for A. arapaima is average for several topotypes that were dissected or skeletonized. Shaded
areas represent the 95%confidence ellipses for the bivariate relationships of the Mamiraua´ sample. P-values represent the probabilities that the
holotype of A. agassizii belongs to the Mamiraua´ population based on an outlier analysis relative to a linear regression of SL versus each
morphometric feature.
42 Copeia 2013, No. 1
and a half feet long. There is therefore room to believe that
it came from an individual at least as big [end of original
description].’’
Comments on original description.—In the foregoing original
description, Valenciennes commented on the lack of
‘indentations’ on the posterior margins of the third and
fourth infraorbitals. The fourth infraorbital, in particular,
has a notable emargination at its postero-dorsal corner
associated with a large sensory cavity that is present in all
Arapaima. That is shown clearly in Agassiz’s drawing of the
complete skeleton (Fig. 1C) and absent from a close-up view
of the same bone (not shown here). The view where it is
absent is perhaps a mistake by the artist. Spix and Agassiz’s
Fig. 3. Arapaima agassizii, holotype: (A) dorsal view of cranium and infraorbital bones; (B) lateral view of cranium, left suspensorium, and jaws with
infraorbital bones removed; and (C) ventral view of cranium, right suspensorium, and right half of jaws; inset is enlargement of maxillary teeth from B
(anterior is to left for all views). Scale bar 55 cm. Abbreviations are: bop, basioccipital process of the parasphenoid; c, vertebral centrum; d, dentary;
dpl, dermopalatine; dsp, dermosphenotic; ecp, ectopterygoid; enp, entopterygoid; fr, frontal; h, hyomandibula; io, infraorbital; iop, interopercle; mpt,
metapterygoid; mx, maxilla; n, nasal; op, opercle; pa, parietal; pas, parasphenoid; pmx, premaxilla; pop, preopercle; pto, pterotic; q, quadrate; soc,
supraoccipital; sop, subopercle; spop, suprapreopercle (among osteoglossids, a novel ossification shared by all Arapaima, forming posterior margin
of lateralis sensory cavity at dorsal tip of preopercle); sym, symplectic; and vo, vomer.
Stewart—Arapaima agassizii re-description 43
(1829) illustration shows ten branchiostegals (Fig. 8B), but
their text says 11. In Arapaima, branchiostegal counts
sometimes differ by one on different sides of the same fish
and range from 10 to 12. In the original descriptions of A.
gigas and A. mapae, Valenciennes (in Cuvier and Valenci-
ennes, 1847) reported 16 branchiostegals for both species,
but my observations on those holotypes revealed that he
was wrong (i.e., I counted ten for A. gigas and 11–12 for A.
mapae). Agassiz appears to have correctly identified and
illustrated the first vertebral centrum that is closely joined to
the back of the skull (Hilton et al., 2007); so, vertebral
counts given above should be correct, with the caveat that
the total count does not include ural centra. The last
abdominal centrum is shown as lacking a rib (Fig. 1C), and
that condition also can be found in many individuals of
Arapaima that I have observed.
The scales are described as being more rugose (vs. less so in
A. gigas and A. mapae), with the exposed surface covered by
ridges and grooves (Fig. 4E, F). That may be a useful
diagnostic character for this taxon, but roughness of scales
Fig. 4. (A) Posterior part of cranium, dorsal view, for Arapaima sp. from
Mamiraua´ Reserve, Brazil (INPA 26583; 110 cm SL); arrow indicates
groove where nape scales insert; (B) nape scales, ventral view, from a
122 cm SL Arapaima arapaima from Guyana (specimen not pre-
served); (C) anterior lateral-line scale from same Guyanese specimen;
(D) posterior part of cranium for holotype of A. agassizii (enlarged from
Fig. 3A); and (E, F) two apparent nape scales from holotype of A.
agassizii (from Spix and Agassiz, 1829; in original publication, there was
no scale bar to indicate sizes of those scales). Scale bars for A, B, and C
52 cm. Abbreviations are: epo, epioccipital; es, extrascapular; pa,
parietal; and soc, supraoccipital.
Fig. 5. Skeletal elements for caudal fin and posterior end of dorsal and
anal fins for A. agassizii, holotype (enlarged from Fig. 1C, scale bar 5
5 cm). Insets are: (A) proximal end of a dorsal-fin ray with distal end of
associated proximal radial (pterygiophore) and medial radial; (B)
penultimate vertebral centrum (pu2) with associated neural spine;
and (C) posterior vertebrae, neural spines, haemal spines, and
associated procurrent caudal-fin rays for an Arapaima from Mamiraua´
Reserve, Brazil (INPA 26583; 110 cm SL); scale bar 52cm.
Abbreviations are: c, vertebral centrum; hs, haemal spine; mrd, medial
radial of dorsal-fin pterygiophore; nspu, neural spine of preural
centrum; phy, parhypural; prd/pra, proximal radial of dorsal- or anal-
fin pterygiophore; and pu, preural centrum.
44 Copeia 2013, No. 1
in Arapaima varies on different areas of the fish. So,
evaluation of scale rugosity as a character must await the
discovery of another specimen of A. agassizii. Apart from
surface rugosity, the illustrated scales are unusual in other
ways and present an interesting puzzle. The scale shown in
Figure 4E appears to have a lateralis sensory pore, but it is
distinctly different from a typical Arapaima lateral-line scale
(Fig. 4C); it has a narrow, truncated anterior field and the
pore opens anterior to the rugose exposed surface of the
scale. In a typical Arapaima, the rugose posterior field
typically is about a third of the scale surface, and openings
of the lateral-line system are in the posterior half of the scale
(Fig. 4C). Scales along the posterior margin of the head in
Arapaima typically insert into a bony groove, which laterally
runs beneath the posterior margin of the parietal (Fig. 4A).
Those nape scales have the anterior field truncated (Fig. 4B).
The scale shown in Figure 4F also has a truncated anterior
field, lacks a lateralis pore, and only half of the posterior
field is rugose, indicating that it was partially covered. In a
typical Arapaima, the most lateral nape scale is variously
irregular in shape, often narrower than the more median
nape scales, and has a large lateralis pore about in the center
of the scale (i.e., dorsal to the distal tip of the extrascapular
bone, Fig. 4A, B). It is also common for the median nape
scales to overlap the lateral scales. I infer that both scales
illustrated by Spix and Agassiz (1829) are nape scales. That is
reasonable considering that the specimen was skeletonized
in the field and that nape scales are strongly attached to the
back of the skull. Still it is not clear precisely how those
scales fit against the skull.
Fig. 6. Comparisons of fin ray height profiles for: (A) dorsal fins and (B)
anal fins of A. agassizii (holotype, circles) and Arapaima sp. from
Mamiraua´ Reserve, Brazil (triangles, INPA 26583; 110 cm SL).
Measurements are for total length of bony elements taken from
skeletons, and so are slightly longer than measurements that would be
possible for intact specimens.
Fig. 7. Pectoral girdle of A. agassizii, holotype: (A) mesial view, (B) lateral view with bones expanded, and (C) lateral view of articulated girdle with
pectoral fin attached. Anterior is to left in B and C, to right in A (scale bar 55 cm). Insets from a specimen of Arapaima from Mamiraua´ Reserve, Brazil
(INPA 26583; 110 cm SL) are: (D) mesial and lateral views of postcleithrum and dorsal tip of cleithrum (left side), scale bar 52 cm; and (E) dorsal view
of proximal tips for right pectoral-fin rays and radials, showing enlarged first pectoral-fin ray, scale bar 51 cm. Abbreviations are: cl, cleithrum; co,
coracoid; mcor, mesocoracoid; pcf, pectoral fin; pcl, postcleithrum; pt, posttemporal; rad, radials of pectoral fin; sca, scapula; and scl, supracleithrum.
Stewart—Arapaima agassizii re-description 45
Many other characters noted in the original description by
Valenciennes need further study and data on variation before
they can be confirmed as diagnostic. He appears to be wrong
about pectoral-fin length; I found that A. agassizii has a
pectoral fin slightly shorter than both A. gigas and A. mapae.
Another feature that may prove to be diagnostic is position of
the sensory cavity on the pterotic (Fig. 3A). In other Arapaima
that I have seen, that cavity lies on the lateral margin of the
bone, whereas in A. agassizii, it is more medial and distinctly
separated from the lateral margin. Another interesting feature
is that the parasphenoid where it joins the vomer is as wide as
the anterior margin of the vomer (Fig. 3C). In contrast, the
types of A. gigas, A. mapae,andallotherArapaima that I have
seen all have the anterior extension of the parasphenoid where
it joins the vomer notably narrower than the anterior
margin of the vomer (e.g., Taverne, 1977:fig. 127). This
latter feature could be diagnostic for A. agassizii,but
possible variation needs to be considered when more
specimens are available. Finally, the fourth epibranchial
was illustrated by Agassiz as slender, resembling epibran-
chials of gill arches 1–3 (Fig. 8A; vs. fourth epibranchial
forming a broad, often nearly rectangular plate in Ara-
paima; Taverne, 1977:fig. 136). Again, this character needs
to be verified when another specimen becomes available.
Comments on diagnostic features.—Several morphological
features listed in the diagnosis above warrant further
discussion.
Fig. 8. Gill arches and hyoid apparatus of A. agassizii, holotype: (A) dorsal view, and (B) lateral view. Anterior is to left, scale bar 510 cm. Inset at
bottom: (C) basibranchial toothplates and proximal ends of hyoid apparatus and gill arches for an Arapaima from Mamiraua´ Reserve, Brazil (INPA
26583; 110 cm SL), scale bar 52 cm. Abbreviations are: abbtp, anterior basibranchial toothplate; br, branchiostegal; cb, ceratobranchial; cha,
anterior ceratohyal; chp, posterior ceratohyal; eb, epibranchial; gr, gill rakers; hb, hypobranchial; hbd, hypobranchial denticles; iph,
infrapharyngobranchial; lptp, lower pharyngeal toothplate; pbbtp, posterior basibranchial toothplate; and uh, urohyal.
46 Copeia 2013, No. 1
Parietal projections: Arapaima typically have four scales
(rarely three or five) across the nape that insert tightly into a
groove along the relatively straight posterior margin of the
parietals (Fig. 4A, B). Cranial elements ventral to those nape
scales lack the rugose striations that cover much of the skull
roof. In other Arapaima, those scales cover the median
supraocciptial, and that is bracketed by the epioccipitals.
The epioccipitals form distally rounded posterior projections
(Fig. 4A; Taverne, 1977:fig. 126) that serve as attachments
for ligaments from epaxial muscles. A medial expansion of
the posttemporal is also just below the nape scales and
dorsal to the epioccipitals. In A. agassizii, the epioccipitals
are not visible dorsally, perhaps because they are hidden
ventral to the parietal projections. Because nape scales
normally insert into a groove below the posterior margin of
the parietal, it is plausible for the surface of the parietal to
grow posteriorly to partially cover the nape scales.
Elongation of caudal peduncle: other species of Arapaima
and Heterotis are similar to each other in having relatively
long dorsal- and anal-fin bases that extend posteriorly
almost to the caudal-fin base. The fore-shortened anal-fin
base in A. agassizii might be explained, in part, if some anal-
fin rays had been lost from the posterior end, but profile of
that fin (Fig. 6B) tapers gradually to a final ray that is similar
in length to that of an Arapaima from Mamiraua´. That sug-
gests no posterior rays were lost. Also, the caudal peduncle is
unusually long even if it is measured from posterior end of
the dorsal-fin base.
Form of dorsal fin: Arapaima typically have a dorsal-fin
profile that is relatively highest in the center of the fin
(Fig. 6A). The dorsal profile in A. agassizii is comparatively
level, while the longest rays are in the anterior third of the
fin. It seems unlikely that unusual form of the dorsal fin in
this species can be explained by the fin being damaged,
becauseitishighlyunlikelythatdamagewouldbe
relatively uniform along the entire fin, and broken fin rays
probably would not retain branches distally. The dorsal
fin also is unusual in having only two unbranched rays
anteriorly, whereas other Arapaima that I have seen
typically have the first 5–7 anterior rays unbranched. If
the fin were badly damaged and reconstructed by the artist,
it seems he might have followed the sketch of the whole
fish that was based on field observations of Spix (Fig. 1B),
and that shows longest rays in the center of the fin.
Resolving these uncertainties must await discovery of
another specimen.
Reduction of first pectoral-fin ray: all other Arapaima have
the first pectoral-fin ray enlarged relative to the other
pectoral rays and extended mesially beyond the pectoral
radials to articulate with the scapula (Fig. 7E, Taverne,
1977:fig. 137). In two separate drawings of the pectoral fin
for A. agassizii, the first ray is shown to be similar in
diameter to the second ray and not extending medial to the
other rays (Figs. 1C, 7C). In contrast, those drawings each
show precisely 12 pectoral-fin rays, which is typical of most
Arapaima when counts include the first, unbranched
element. Agassiz’s text description of the pectoral fin differs
from what can be seen in the drawing in indicating that the
first ray ‘‘is much stronger at the base and rougher than the
other rays’’ (translation in Pethiyagoda and Kottelat, 1998),
so it remains to be seen if this character will be retained as
diagnostic.
Hypertrophy of basibranchial tooth plate: Spix and Agassiz
(1829) stated that the ‘‘os hyoideum’’ was ‘‘7 inches’’ in ‘‘the
more than 3 ft long specimen’’. Inspection of the gill arch
drawing (Fig. 8A) indicates that the basibranchial tooth
plate extends farther posterior than a typical Arapaima and,
perhaps, also extends farther anterior. As noted above, the
tooth patch on the anterior extension of the parasphenoid
where it articulates with the vomer is unusually wide
(Fig. 3C). Hypertrophy of tooth patches in that region on
the roof of the buccal chamber would complement an
unusually long basibranchial tooth plate. Precise homology
of basibranchial toothplates in A. agassizii versus other
Arapaima remains to be determined when another specimen
is collected, but it seems possible that the elongate tooth-
plate derives from fusion of anterior and posterior basibran-
chial toothplates (Fig. 8A). Spix and Agassiz (1829) state that
the toothplate in A. agassizii is a fusion of four elements. If
so, there is an enlarged fifth toothplate (lptp, Fig. 8A)
posterior to that, perhaps developed from the small tooth
patches seen on the fifth arch in other Arapaima. In all
Arapaima that I have dissected, there is a cartilaginous
fourth basibranchial below a separate, rounded fourth
toothplate (pbbtp, Fig. 8C), and that basibranchial has a
slender posterior extension that projects along the midline
beyond the base of the fifth gill arch. Spix and Agassiz (1829)
describe the fourth basibranchial in A. agassizii as ‘‘ . . . very
level and almost quadrangular. It is loosely joined with the
third ...’’Sotheir observations on form of the fourth
basibranchial differ notably from what I have observed for
other Arapaima.
Loss or fusion of postcleithrum: in other Arapaima, the
postcleithrum is a well-developed, thin plate that is tightly
joined to the dorsal end of the cleithrum; its anterior margin
fits into a vertical groove along the anterior margin of the
cleithrum (Fig. 7D; Taverne, 1977:fig. 137). If the postclei-
thrum is removed, the dorsal tip of the cleithrum is distinctly
asymmetrical. If this configuration were present in A. agassizii,
it would be clearly evident in one or more of the threepectoral
girdle drawings (Fig. 7A–C). It is unlikely that an artist
worrying about drawing precisely 12 pectoral-fin rays would
somehow overlook the postcleithrum three times. I infer that
this bone was absent or, perhaps, completely fused to the
cleithrum. Either way, the dorsal tip of the cleithrum has an
unusual, diagnostic configuration.
Caudal fin morphology: perhaps the center of the caudal fin
(including ural centra, hypural plates, and median fin rays)
in A. agassizii was damaged when the fish was young or was
lost in preparation or during shipping from South America.
The probabilities of damage or loss, however, are lessened by
the likely presence of ligaments, thick skin, and dense scales
similar to those covering the caudal-fin base in other
Arapaima. Uncertainty about the missing elements can only
be resolved when another specimen of this taxon is found,
but regardless of any lost bones, those elements illustrated
by Agassiz represent an unusual caudal morphology (Fig. 5).
In other Arapaima, it is common for about four neural and
5–6 haemal spines of posterior centra to support caudal-fin
rays, involving centra pu1, pu2, and two or three vertebrae
anterior to those (e.g., Hilton, 2003:fig. 36). However, this
typical condition in Arapaima usually involves each neural
and haemal spine supporting two–three procurrent caudal-
fin rays (Fig. 5C). Total caudal-fin rays in Arapaima,
including procurrent rays, is often as high as 30–32, with
Stewart—Arapaima agassizii re-description 47
18–22 branched rays (pers. obs., including A. arapaima
topotypes from Guyana, Arapaima sp. from Mamiraua´ and
many other areas). Valenciennes (in Cuvier and Valenci-
ennes, 1847) reported a total of 17 caudal-fin rays for A. gigas
and 14 for A. mapae. My observations on the holotype of A.
gigas suggested that there are about three unbranched
procurrent rays dorsally and ventrally, and about 11
branched rays (i.e., total count 17), but much of the fin is
too thickly covered with skin and scales to be confident
about that total ray count. In most Arapaima, it is difficult to
see the smallest, anterior procurrent rays without an x-ray or
dissection of the fin base. The caudal fin in the holotype of
A. mapae is indeed small, as noted by Valenciennes, but now
it is too damaged to re-confirm the count reported by
Valenciennes.
DISCUSSION
Agassiz’s skeletal illustrations.—Relative length of the basi-
branchial toothplate given in the text of Spix and Agassiz’s
(1829) paper is a clear diagnostic feature for this species (vs.
other Arapaima where toothplate data are available; Fig. 2A),
but most other aspects of the species diagnosis depend on
the illustrations that were supervised by Agassiz (in Spix and
Agassiz, 1929). So it is important to carefully consider the
accuracy of those figures. Spix began the monograph on
Brazilian fishes and supervised preparation of the color
illustrations (i.e., Fig. 1B), but Agassiz was hired to complete
that study after Spix died in 1826 (Kottelat, 1988). The
following is my translation of Valenciennes’ (in Cuvier and
Valenciennes, 1847) opinion of the figures in Spix and
Agassiz (1829).
‘‘A third species of Vastre`s is incontestably the one that M.
Spix had figured in his compilation of the ichthyological
plates under the name of Sudis pirarucu. I don’t retain this
name for two reasons: first, it is that the name of Pirarucu is
the name of all species of Vastre`s of the Amazon. Secondly,
the figure of Spix is poor, inexact, because it is made based
on a poorly mounted skeleton. Without any other scales
than those of the base of the dorsal and anal fins, the figure
has been composed according to the small engraving of our
first species, presented by M. Cuvier in his ‘Regne animal’.
These complements of various documents, borrowed from
different species, were unfortunately only too frequent in all
branches of zoology: they render all scientific critique very
difficult, because most often the author, to hide his
falsification, doesn’t copy accurately the parts that he adds.
In the present case, the species of Sudis pirarucu would be
quite impossible to determine, if M. Agassiz had not taken
the care to make a drawing of the whole skeleton and the
different bones of the head and the hyoid apparatus of this
beautiful fish. Deceived by a first determination, he believed
that Vastres pirarucu presented by Spix, was merely Sudis
gigas; as it is very different from the latter, I believe it is my
duty to dedicate to my scholarly friend the species that he
first made known.’’
A careful reading of this commentary by Valenciennes
reveals that he was highly critical of the whole-fish
illustration (pl. 16) of Spix. Compared to the two dried
specimens that Valenciennes had available, the figures of
both Cuvier and Spix (Fig. 1A, B) have the anal-fin origin
much too far posterior. So it appears that Spix copied
Cuvier’s mistake for that detail. In Valenciennes’ detailed
description of V. cuvieri (5A. gigas), he described position of
the dorsal-fin origin but made no mention of the anal-fin
origin and, furthermore, did not illustrate the whole body of
the fish. So Cuvier’s (1816) mistake was left uncorrected. To
be fair to Spix, however, details of bones on the head, scales,
paired fins, and color all would be from his field notes
because those same details were not evident in Cuvier’s
(1816) rather imprecise drawing (compare Fig. 1A, B).
It was an unfortunate choice of words by Valenciennes,
however, to say ‘‘poorly mounted skeleton’’ in the middle of
his criticism of the ‘‘falsification’’ of Spix. A casual reader
might then wonder about the veracity of the skeletal
drawings. In marked contrast, Valenciennes was sufficiently
impressed by the osteological drawings of the young Agassiz
that he named the species after his ‘‘scholarly friend.’’ In the
original description by Valenciennes (quoted above), he
notes that the skeleton was drawn with ‘‘great care’’ and the
description was ‘‘rigorous.’’ I concur with that assessment,
and as discussed further below, there is nothing in the
drawings of Agassiz that would indicate a ‘‘poorly mounted
skeleton.’’
It is not difficult to find evidence that the drawings
supervised by Agassiz were detailed and rigorous. When I
enlarged a digital image of the skull in lateral view, I was
surprised to see that each individual tooth in the upper jaw
had been counter-shaded to give it shape (Fig. 3B, inset).
The same is true of each gill raker in the gill-arch drawing
(Fig. 8A, B). That is the work of an exceptionally meticu-
lous, professional artist. Most Arapaima have about 12 gill
rakers on the upper limb of the first arch and 25–30 on the
lower limb, with surprisingly little variation. Allowing for a
few anterior rakers that would be hidden under the large
basibranchial tooth plate, the number of rakers illustrated
by Agassiz is what I would expect. The number of teeth on
the premaxilla is also similar to what occurs in nearly all
Arapaima that I have studied, but maxillary and mandibular
counts are significantly higher than for a typical Arapaima
of this size (and so diagnostic for this species). Counts for
vertebrae, pleural ribs, and supraneurals also are all similar
to what one finds in most Arapaima. The dorsal view of the
skull is carefully counter-shaded as if a light were shining
from right to left. So each of the posterior projections on
the parietals has shading on the left margin, and the
supraocciptial is correctly shaded to indicate that it is in a
more ventral plane than the surface of the skull. In two
separate drawings, the pectoral fin is shown to have exactly
12 rays, which is typical of nearly all Arapaima.Soitisclear
that Agassiz and his artist were precise about meristic
counts and many other anatomical details. At that time,
there were no published illustrations of Arapaima osteolo-
gy, so those details had to come from the specimen in
hand.
Position of the anal fin deserves special mention because
that has been misrepresented in previous illustrations
(Fig. 1A, B), as noted above. Agassiz shows the first anal-
fin pterygiophore (proximal radial) inserted between the
first and second haemal spines (Fig. 1C). That is precisely
what I have seen in several Arapaima skeletons that I have
prepared or studied and also matches what has been
published (Taverne, 1977). Close inspection of the posterior
end of the anal fin reveals that lengths of the anal-fin rays
taper gradually toward the last, shortest ray (Fig. 6B). If the
fin had been damaged with about ten posterior rays lost, the
last ray would be relatively longer, and the posterior third of
the fin would be abruptly truncated. I infer that the anal fin
was intact and correctly positioned, not ‘‘poorly mounted.’’
48 Copeia 2013, No. 1
The first pterygiophore (or less often, the first two) of the
dorsal fin in Arapaima typically inserts between the neural
spines of the last and penultimate abdominal centra. In A.
agassizii it inserts one centrum farther forward, and two
pterygiophores are thus inserted (Fig. 1C). So this position
of the dorsal-fin origin is slightly different from the average
Arapaima, but still quite reasonable. I have seen that
identical configuration in one specimen of Arapaima from
Mamiraua´ Reserve, Brazil. Another indication of careful
attention to detail is the enlargement of a dorsal-fin ray with
its proximal and medial radials (Fig. 5A), and further, the
dorsal-fin rays are shown to have medial radials linking their
bases along the entire dorsal fin. An interesting contrast to
those fine details is the absence of medial radials along the
anal-fin base. Taverne (1977) indicated that medial radials
should be present along both the dorsal and anal fins of
Arapaima but were not developed on the posterior-most rays
of each fin. If A. agassizii does lack medial radials along the
anal fin, that could be an additional diagnostic character,
and it would be a reasonable correlate of an anal fin reduced
in form and function.
Some might wonder if the caudal fin illustrated by Agassiz
was damaged (or missing) and, perhaps, reconstructed by
the artist to complete the illustration. Two features of the
drawings argue against that scenario. First, the illustration
shows the pelvic fins as dotted lines to indicate that the
artist added them to the illustration because the pelvic fins,
in fact, were missing. If the artist drew the missing pelvic
fins as broken lines to indicate their absence, why would he
construct a missing caudal fin using solid lines, and why
would he not attempt to reconstruct the median and basal
elements of that fin? Also, such a flaw in the skeleton should
have been mentioned in the text by Agassiz just as he
mentioned the missing pelvic fins. Instead, Agassiz’s text
description of the caudal fin is similar to the illustration
(‘‘The rays of the caudal fin are joined together on a wide
truncated apical platform of the spiny processes of the
lowest vertebrae, which diverge like spokes of a wheel at the
upper end of the tail’’; translated from Latin, in Pethiyagoda
and Kottelat, 1998). Second and even more telling, the
drawings include a close-up sketch showing centrum pu2
with its associated neural spine (Fig. 5B). That close-up
shows a blade-like neural spine that could only have been
drawn by reference to the actual bone (compare to Fig. 5C).
So that is clear evidence that caudal elements were present
at least to centrum pu2.
In summary, the illustrations of Agassiz appear to be
carefully executed, although different views are done at
different scales and seem to emphasize different features. So
a structure that is vague at large scale may be more clearly
shown in another close-up view (e.g., posterior parietal
projections are shown in two close-ups of the skull but are
not evident in the coarser-scale, whole-fish drawing). There
is no evidence in the drawings that the skeleton was
damaged, except for the missing pelvic fins and, perhaps,
the median caudal-fin elements. Specialized features, how-
ever, were illustrated side-by-side with well-executed typical
features of Arapaima in the skull, jaws, gill arches, pectoral
girdle, and median fins. I infer that the illustrations are
accurate because, overall, they are anatomically plausible,
and to argue otherwise would require invoking ad hoc
hypotheses to refute many of the main diagnostic charac-
ters. The anatomical and taxonomic observations of Agassiz
and Valenciennes, thus, should be recognized as valid until
someone can present a credible falsification of their results.
Gu¨nther (1868) clearly made no effort to do that, but his
opinion has stood as conventional wisdom for 144 years.
Conservation status.—So far, I have been unable to locate a
specimen of A. agassizii, but that is not surprising given the
scarcity of Arapaima in museums. Most of the largest fish
collections in the world only have a few specimens of
Arapaima, and those usually include aquarium specimens
with no locality data. There are vast areas of the Amazon
basin for which there are no preserved materials of
Arapaima to be studied. It is my hope that this paper will
stimulate field biologists working in Brazil to locate any
residual population, and in general, to bring more speci-
mens of Arapaima to museums. To facilitate that process, I
have re-constructed what an intact specimen of A. agassizii
might look like (Fig. 1D). The strongly tapered posterior
third of the body and short anal fin in this species should
be features allowing easy field identification. Given the
extent to which Arapaima have been heavily exploited
during the past 150 years, especially around population
centers (Eigenmann and Allen, 1942; Goulding, 1980;
Neves, 1995; Castello and Stewart, 2010; Castello et al.,
2011), it is possible that populations of this species also
have been depleted. Since A. agassizii has not been
collected in the past 190 years (inference based on
assessment of museum holdings) and the type locality is
unknown, it is clearly ‘Data Deficient.’
ACKNOWLEDGMENTS
This research was supported, in part, by National Geograph-
ic Society and SUNY College of Environmental Science and
Forestry. I thank C. Watson and fishers from the community
of Rewa, Rupununi District, Guyana, for assistance with
related field efforts; Iwokrama International Centre for Rain
Forest Conservation and Development also assisted in
Guyana. Unpublished morphometric and meristic data from
Mamiraua´, Brazil, were provided by C. Arantes, Instituto de
Desenvolvimento Sustenta´vel Mamiraua´. The manuscript
benefited from constructive comments by L. Castello, C.
Arantes, and C. de Santana. I thank M. Hall for help with
French to English translations. Librarians at Cornell Uni-
versity, MCZ, and AMNH provided access to rare books;
figures from Spix and Agassiz (1829) are all based on my
photos from the AMNH copy of that monograph, and
Figure 1A is my photo from Cornell University’s copy of
Cuvier (1816); M. Kottelat also helped with early literature.
For assistance during museum visits, I thank: P. Pruvost
(MNHN); O. Crimmen (BMNH); K. Hartel (MCZ); B. Brown
(AMNH); M. Rogers (FMNH); J. Williams (USNM); L. Rapp
Py-Daniel (INPA); M. de Pinna (MZUSP); P. Buckup (MNRJ);
W. Wosiacki (MPEG); R. Barriga (EPN); H. Ortega (MNHJP);
C. Bernard (CSBD/UG); and many others. In Guyana, the
Environmental Protection Agency provided permit
no. 060706 BR 055; Brazilian (IBAMA-Manaus) research
permit to C. Arantes was #02005.0021/06-67.
LITERATURE CITED
Bleher, H. 2006. Bleher’s Discus, Vol. 1. Aquapress Bleher,
Pavia, Italy.
Castello, L. 2008. Lateral migration of Arapaima gigas in
floodplains of the Amazon. Ecology of Freshwater Fish
17:38–46.
Stewart—Arapaima agassizii re-description 49
Castello, L., and D. J. Stewart. 2010. Assessing CITES non-
detriment findings procedures for Arapaima in Brazil.
Journal of Applied Ichthyology 26:49–56.
Castello, L., D. J. Stewart, and C. C. Arantes. 2011.
Modeling population dynamics and conservation of
arapaima in the Amazon. Reviews in Fish Biology and
Fisheries 21:623–640.
Cook, R. D., and S. Weisberg. 1982. Residuals and Influence
in Regression. Chapman and Hall, New York.
Cuvier, G. 1816. Le Re`gne Animal distribue´ d’apre`s son
organisation pour servir de base a` l’histoire naturelle des
animaux et d’introduction a` l’anatomie compare´e. Les
reptiles, les poissons, les mollusques et les anne´lides.
Edition 1. V. 2:i–xviii +1–532 [pls. 9–10, in v. 4].
Cuvier, G. 1822. Das Thierreich eingetheilt nach dem Bau
der Thiere als Grundlage ihrer Naturgeschichte und der
vergleichenden Anatomie. Mit vielen Zusa¨ tzenversehen
von H. R. Schinz. Cotta, Stuttgart and Tu¨ bingen,
Germany.
Cuvier, G. 1829. Le Re`gne Animal, distribue´ d’apre`s son
organisation, pour servir de base a` l’histoire naturelle des
animaux et d’introduction a` l’anatomie compare´e. Edi-
tion 2. V. 2:i–xv +1–406.
Cuvier, G., and A. Valenciennes. 1847. Histoire naturelle
des poissons. Tome dix-neuvie`me. Suite du livre dix-
neuvie`me. Brochets ou Lucioı
¨des. Livre vingtie` me. De
quelques familles de Malacopte´rygiens, interme´diaires
entre les Brochets et les Clupes. P. Bertrand, Paris V.
19:i–xix +1–544 +6 pp., pls. 554–590 [not 520–556].
Eigenmann, C. H., and W. R. Allen. 1942. Fishes of
Western South America: I. The Intercordilleran and
Amazonian Lowlands of Peru; II. The High Pampas of
Peru, Bolivia and Northern Chile, with a Revision of the
Peruvian Gymnotidae, and of the Genus Orestias. The
University of Kentucky, Lexington, Kentucky.
Eschmeyer, W. N., and R. Fricke (eds.). 2011. Catalog of
Fishes. California Academy of Sciences. Electronic version
accessed. 30 November 2011. http://research.calacademy.
org/research/ichthyology/catalog/fishcatmain.asp
Ferraris, C. J., Jr. 2003. Family Arapaimatidae (Bonyton-
gues), p. 31. In: Check List of the Freshwater Fishes of
South and Central America. R. E. Reis, S. O. Kullander,
and C. J. Ferraris, Jr. (eds.). EDIPUCRS, Porto Alegre,
Brazil.
Ferreira, A. R. 1903. Memoria sobre o peixe Pira´-Urucu´de
que ja se remetteram, dous da Villa de Santarem para o
Real Gabinete de Historia Natural e agora se remettem
mais cinco desta Villa de Barcellos, os quaes va˜o incluidos
nos cinco caizoens, que constituem parte da sexta remessa
do Rio Negro. Archivos do Museu Nacional do Rio de
Janeiro 12:155–158 (Posthumous publication of original
manuscript drafted by Ferreira about 1787, submitted and
anotated by A. de Miranda-Ribeiro).
Goulding, M. 1980. The Fishes and the Forest: Explorations
in Amazonian Natural History. University of California
Press, Los Angeles, California.
Gu¨nther, A. 1868. Catalogue of the Fishes of the British
Museum. Catalogue of the Physostomi, containing the
Families Heteropygii, Cyprinidae, Gonorhynchidae, Hyo-
dontidae, Osteoglossidae, Clupeidae, Chirocentridae, Ale-
pocephalidae, Notopteridae, Halosauridae, in the Collec-
tion of the British Museum. V. 7:i–xx +1–512.
Hammer, Ø., D. A. T. Harper, and P. D. Ryan. 2001.
PAST: paleontological statistics software package for
education and data analysis. Palaeontologia Electro-
nica 4(1):1–9. http://palaeo-electronica.org/2001_1/past/
issue1_01.htm
Hilton,E.J.2003. Comparative osteology and phylogenet-
ic systematics of fossil and living bony-tongue fishes
(Actinopterygii, Teleostei, Osteoglossomorpha). Zoologi-
cal Journal of the Linnean Society 137:1–100.
Hilton, E. J., R. Britz, G. D. Johnson, and P. L. Forey.
2007. Clarification of the occipito-vertebral region of
Arapaima gigas (Osteoglossomorpha: Osteoglossidae)
through developmental osteology. Copeia 2007:218–
224.
IUCN. 2011. World Conservation Monitoring Centre 1996,
Arapaima gigas.In: IUCN Red List of Threatened Species.
Version 2011.2. www.iucnredlist.org. Downloaded on 11
November 2011.
Kottelat, M. 1988. Authorship, dates of publication, status
and types of Spix and Agassiz’s Brazilian fishes. Spixiana
11:69–93.
Kullander, S. O., and E. J. G. Ferreira. 2006. A review of the
South American genus Cichla, with descriptions of nine
new species (Teleostei: Cichlidae). Ichthyological Explo-
rations of Freshwaters 17:289–398.
Li, G.-Q., and M. V. H. Wilson. 1996. Phylogeny of
Osteoglossomorpha, p. 163–174. In: Interrelationships of
Fishes. M. L. J. Stiassny, L. R. Parenti, and G. D. Johnson
(eds.). Academic Press, San Diego, California.
Lundberg,J.G.,M.Kottelat,G.R.Smith,M.L.J.
Stiassny, and A. C. Gill. 2000. So many fishes, so little
time: an overview of recent ichthyological discovery in
continental waters. Annals of the Missouri Botanical
Garden 87:26–62.
Martinelli, N. M. C., and M. Petrere, Jr. 1999. Morpho-
metric relationships and indirect determination of the
length frequency structure of the pirarucu Arapaima gigas
(Cuvier), in Brazilian Amazonia. Fisheries Management
and Ecology 5:233–240.
Neves, A. M. B. 1995. Conhecimento atual sobre o pirarucu
Arapaima gigas (Cuvier 1817). Boletim di Museu Paraense
Emı´lio Goeldi, Se´rie Zoologia 11:33–56.
Pethiyagoda, R., and M. Kottelat (eds.). 1998. Fishes of
Brazil: an aid to the study of J. B. Spix and L. Agassiz
(1829–31), Selecta genera et species piscium Brasiliensium,
including an English translation of the text by V. L.
Wirasinha and reproduction of all illustrations. WHT
Publications, Colombo, Sri Lanka.
Piorski, N. M., A. Sanches, L. F. Carvalho-Costa, T.
Hatanaka,M.Carrillo-Avila,P.D.Freitas,andP.M.
Galetti, Jr. 2008. Contribution of conservation genetics
in assessing neotropical freshwater fish biodiversity.
Brazilian Journal of Biology 68(4, suppl.):1039–1050.
Schomburgk, R. H., Esq. 1841. The natural history of fishes
of Guiana. Part I:[1–16], 17–263, pls. 1–30. In: The
Naturalists’ Library, Volume 3. W. Jardine (ed.). W. H.
Lizars, Edinburgh, Scotland (Schomburgk wrote the
introduction, and Jardine wrote the fish descriptions from
Schomburgk’s field notes).
Spix, J. B. V., and L. Agassiz. 1829–31. Selecta genera et
species piscium quos in itinere per Brasiliam annos
MDCCCXVII–MDCCCXX jussu et auspiciis Maximiliani
Josephi I . . . colleget et pingendso curavit Dr. J. B. de Spix
. . . Monachii. Selecta Piscium Brasiliam. Part 1:i–xvi +i–ii
+1–82, pls. 1–48, Part 2:83–138, pls. 49–101 [Part 1,
published June 1829; Part 2, January 1831].
50 Copeia 2013, No. 1
Stone, R. 2007. The last of the leviathans. Science 316:1684–1688.
Taverne, L. 1977. Oste´ologie, phyloge´ne`se et syste´ matique
des te´le´ oste´ens fossiles et actuels du super-ordre des
oste´oglossomorphes, premie`re partie. Oste´ologie des
genres Hiodon,Eohiodon,Lycoptera,Osteoglossum,Sclero-
pages,Heterotis et Arapaima. Acade´mie Royale de Belgique,
Me´moires de la Classe des Sciences 42:1–235.
Tiefenbacher, L. 1983. Die Brasilienexpedition von J. B. v.
Spix and C. F. Ph. v. Martius in den Jahren 1817 bis 1820,
ein abriss. Spixiana Supplement 9:35–42.
Watson, L. C. 2011. Ecology and conservation of Arapaima
in Guyana: recovery of a giant fish in distress. Unpubl.
Ph.D. diss., SUNY College of Environmental Science and
Forestry, Syracuse, New York.
Stewart—Arapaima agassizii re-description 51
