Content uploaded by Michael Wink
Author content
All content in this area was uploaded by Michael Wink on Jun 19, 2015
Content may be subject to copyright.
Journal of Natural History
Vol. 46, Nos. 35–36, September 2012, 2179–2258
A critical revision of the churchill snoutfish, genus Petrocephalus
Marcusen, 1854 (Actinopterygii: Teleostei: Mormyridae), from southern
and eastern Africa, with the recognition of Petrocephalus tanensis, and
the description of five new species
Bernd Kramera*, Roger Billsb,PaulSkelton
band Michael Winkc
aZoological Institute, University of Regensburg, D-93040 Regensburg, Germany; bSouth African
Institute for Aquatic Biodiversity, Private Bag 1015, Grahamstown 6140, South Africa; cInstitut
für Pharmazie und Molekulare Biotechnologie, Universität Heidelberg, D-69120 Heidelberg,
Germany
(Received 13 September 2011; final version received 28 June 2012)
We morphologically and genetically studied the southern African electric fish
Petrocephalus catostoma, or churchill, and its six nominal species, five of which
by synonymization (three valid subspecies). We reinstate the synonymized species,
and recognize Petrocephalus tanensis (Whitehead and Greenwood, 1959) from the
Tana River in Kenya, also using electric organ discharges. The Okavango delta
(Botswana) is inhabited by Petrocephalus okavangensis sp. nov. and Petrocephalus
magnitrunci sp. nov., and the Namibian Cunene River by Petrocephalus magnoculis
sp. nov. We recognize Petrocephalus petersi sp. nov. fortheLowerZambeziRiver
(Mozambique), and Petrocephalus longicapitis sp. nov. for the Upper Zambezi River
(Namibia). The Lufubu River in Northern Zambia is inhabited by Petrocephalus
longianalis sp. nov. For the southern churchill, Petrocephalus wesselsi Kramer
and Van der Bank, 2000, we confirm intraspecific and interspecific differentia-
tion. Sequence data from mitochondrial DNA confirm differentiation of two new
western and two eastern species, forming mutual sister groups.
Keywords: systematics; morphometrics; electric organ discharges; molecular
genetics; allopatric speciation
Introduction
The African snoutfish genus Petrocephalus Marcusen, 1854 is defined on characteris-
tic skeletal features (Taverne 1969), certain characters of external morphology, such
as a pair of narrowly spaced nostrils the posterior one of which is closely apposed
to the eye (Bigorne and Paugy 1991), and molecular DNA studies (Lavoué et al.
2000; Sullivan et al. 2000). About 25 species are distributed throughout the more
tropical regions of Africa, two in southern Africa. The type locality for the widely
distributed Petrocephalus catostoma (Günther, 1866), or churchill, is the Rovuma
River that arises in the highlands east of Lake Malawi (Livingstone Mountains),
whence it flows eastward into the Indian Ocean (Figure 1, no. 1). The Rovuma (also
Ruvuma) forms the border between Tanzania and Mozambique for about 600 km
at 11◦S. Whitehead and Greenwood (1959) reviewed the status of “three closely
related species of Petrocephalus ...recorded from East Africa; these are P. degeni
*Corresponding author. Email: bernd.kramer@biologie.uni-regensburg.de
ISSN 0022-2933 print/ISSN 1464-5262 online
© 2012 Taylor & Francis
http://dx.doi.org/10.1080/00222933.2012.708452
http://www.tandfonline.com
2180 B. Kramer et al.
10° E 20° 30° 40°
40°
30°
20°
10° E
0°
10° S
20°
30° 30°
20°
10° S
0°
1000 km
1
•2
•
3
•4
•
5
•
6
•22
•
7
•
8
•
9
•
10
11
12
•
13
14 •
15
•
16
•
18
17
20
21
•
23
•
24
•
25
•
26
•
19
27
28
29
30
•
•
•
•
•
•
•
31
•
•
36
37
•
38
•
39
Figure 1. Map of southern Africa indicating the origin of samples of the Petrocephalus species
studied. (1) Rovuma (Ruvuma) River, type locality for P. catostoma (Günther 1866) [BMNH
1863.10.12.4]; (2) Ruvu (Kingani) River, type locality for P. stuhlmanni Boulenger 1909 [BMNH
1907.12.3.1]; (3) Sabie River, type locality for P. wesselsi Kramer and Van der Bank 2000 [ZSM
28554 to ZSM 28566, SAIAB 054449]; (4) Groot Letaba River, Limpopo System [SAIAB
85920]; (5) Blyde River, Limpopo System [SAIAB 85923]; (6) Pongola River [SAIAB 85919]
(7) Upper Zambezi River near Katima Mulilo, type locality for P. longicapitis sp. nov. [SAIAB
85916]; (8) Kwando River [ZSM 38658]; (9) Okavango Delta, Nguma Lagoon, type locality for
P. okavangensis sp. nov. [SAIAB 030046]; (10) Tana River, type locality for P. catostoma tanensis
Whitehead and Greenwood, 1959, here recognized as P. tanensis (Whitehead and Greenwood,
1959) [SAIAB 85907]; (11) Lake Rukwa [SAIAB 059515]; (12) Lufubu River, Luapula River sys-
tem, P. longianalis sp. nov. [SAIAB 76758]; (13) East Lungu River, Kafue/Zambezi River system
[SAIAB 040074]; (14) East Lumwana River, Zambezi system [SAIAB 041208]; (15) Mwekera
Stream, Kafue/Zambezi River system [SAIAB 042559]; (16) Kapesha River, Lake Malawi
[SAIAB 039328]; (17) Dwangwa River, Lake Malawi [specimen SAIAB 050065]; (18) Kaombe
River, Lake Malawi [SAIAB 050155]; (19) Lake Chiuta [SAIAB 039264]; (20) Mulela River
[SAIAB 055875]; (21) Zambezi River Delta, type locality for P. pe t e r s i sp. nov. [SAIAB 060846];
(22) Mbuluzi River, Swaziland [SAIAB 067228]; (23) Cunene River, type locality for P. mag-
noculis sp. nov. [SAIAB 78788]; (24) Lukula River, type locality for P. haullevillii Boulenger
1912 [BMNH 1912.4.1.186-188]; (25) Rufiji basin, type locality for P. steindachneri Fowler
1958 [NMW 551181]; (26) Mukishi on Lomami River (Congo River basin), type locality for
Journal of Natural History 2181
BLGR. 1906, P. stuhlmanni BLGR. 1909 and P. catostoma GÜNTHER, 1866”. These
authors concluded “...it became clear that size discrepancies and paucity of mate-
rial could explain why three ‘species’ had been recognised.” Therefore, they united all
three species as members of a single, widespread species, P. catostoma. To the list of
synonyms they added P. stuhlmanni congicus David and Poll, 1937 and P. haullevillii
Boulenger, 1912 from the distant and unconnected upper and lower Congo basins,
respectively, while recognizing that the “Congo form of this species clearly differs
from the eastern and southern subspecies ...”, and that “the two Congoan forms
may yet have to be united”. In consequence, P. steindachneri Fowler, 1958 of East
Africa became the sixth nominal species referred to P. catostoma (Seegers 1996).
Petrocephalus catostoma defined in this way ranged from the Katonga River, Lake
Victoria, in Uganda in the north to the Pongola River in South Africa (a distance of
3000 km); from the Atlantic Congo and Cunene Rivers in the west to the Indian Ocean
in the east (Whitehead and Greenwood 1959; Gosse 1984; Seegers 1996; Eschmeyer
2011) (Figure 2).
A critical comparison among a few allopatric samples from southern local-
ities revealed that South African churchills represented a different species from
P. catostoma (Kramer and Van der Bank 2000). The southern churchill, P. wesselsi
Kramer and Van der Bank, 2000, as it is now called, also differed genetically from
the Petrocephalus sampled from the Upper Zambezi (Van der Bank 1996). Additional
species diversity within Petrocephalus has similarly been discovered in other regions
of sub-Saharan Africa: in Gabon in Central Africa where a new (fourth) species of
Petrocephalus,P. sullivani Lavoué, Hopkins and Kamdem Toham, 2004, was recog-
nized on the basis of anatomical and electrical characters (Lavoué et al. 2004). Five
more species [supported by molecular genetics and electric organ discharge (EOD)
comparisons] have been found in a small region in the northwest of the Republic of
Congo (Lavoué et al. 2008, 2010; Lavoué 2011); one more species in the upper reaches
of the Congo basin in northern Zambia (Lavoué Forthcoming 2012). For a full revision
of only P. catostoma as traditionally understood (including all the nominal species), all
local populations need to be sampled and critically compared. Given the huge distri-
bution and the prevailing sparseness of museum specimens, especially type material,
this goal appeared difficult to achieve.
We took the opportunity to sample additional rivers, among them the Tana River,
type locality for P. c. tanensis Whitehead and Greenwood, 1959, for more material to
extend our comparisons of allopatric churchills (Figure 1). We compared anatomical
P. stuhlmanni congicus David and Poll 1937 [MRAC 30.807–30.808]; (27) Katonga River, Lake
Victoria (Uganda), type locality for P. degeni Boulenger 1906 [BMNH 1906.5.30]; (28) approx-
imate location for our type region material for P. catostoma (SAIAB 73802, 73808, 73887,
73894), (29) Luapula River [SAIAB 76582]; (30) Bangweulu Lake [SAIAB 76859 and SAIAB
76825]; (31) Okavango delta, Boro River, type locality for P. magnitrunci sp. nov. [SAIAB
67069]. (36) Lepalala River, tributary of Limpopo [SAIAB 96537]; (37) Mokolo River, tribu-
tary of Limpopo [SAIAB 95989]; (38) Nwanedzi River, tributary of Limpopo [SAIAB 58157],
(39) “Ruisseau affluent de la Lukinda”, close to Lake Moero, type locality of P. squalostoma
(Boulenger, 1915) [BMNH 1920.5.26.1]. Some rivers and lakes are too small to be shown at the
scale used.
2182 B. Kramer et al.
data and, where possible, also EODs and molecular genetics to test the hypothe-
sis of a P. catostoma species complex for the whole of southern and eastern Africa.
We attempted to reconstruct the systematics and phylogeography in what has tradi-
tionally been considered to represent a single species, the churchill (P. catostoma)of
subcontinental distribution, and to identify some of the local adaptations both for
morphology and the electric communication signal.
Material and methods
Electrical and morphological studies
A total of 566 specimens was examined morphologically and at least 16 measurements
and at least three meristic characters were recorded. Measurements are illustrated in
Figure 3 and were made using vernier calliper readings to 0.1 mm. The following abbre-
viations were used: PDL, predorsal length: distance tip of snout to dorsal fin origin;
PAL, preanal length: distance tip of snout to anal fin origin; LD, dorsal fin length; LA,
anal fin length; pD, distance dorsal fin origin to end of caudal peduncle; CPL, length
of caudal peduncle: end of anal fin base to midbase caudal fin; CPD, depth of caudal
peduncle: the least vertical distance across the caudal peduncle; LSo, length of snout:
distance tip of snout to posterior orbital rim of eye; LSc, length of snout: distance tip
of snout to centre of eye; HL, head length: distance tip of the snout to furthest bony
edge of the operculum; Na, distance between the pair of nares of one side (from cen-
tre to centre); OD, eye diameter: defined by orbital rims; LPF, length of pectoral fins:
from anterior base to tip; PPF, distance between anterior base of pectoral fin to ante-
rior base of pelvic fin; BD, body depth: the greatest vertical distance across the body;
SL, standard length: distance tip of snout to midbase caudal fin; nD, number of dorsal
fin rays; nA, number of anal fin rays; SPc, number of scales around caudal peduncle;
SLS, number of scales in linear series along the lateral line row, as detailed in Skelton
(2001: 67); SLS range of accuracy, ±2 counts.
Abbreviations used to represent institutions and collections follow Leviton
et al. (1985) and Fricke and Eschmeyer (2011). Specimens collected during
the course of the present study are permanently stored at the South African
Figure 2(A–D). Photographs of members of southern and eastern African Petrocephalus species
studied (numbers refer to localities given in Figures 1, 6 and 9). (1) P. catostoma (Günther,
1866), Lectotype, BMNH 1863.10.12.4, right side, SL 4.7 cm. (2) P. stuhlmanni Boulenger, 1909,
holotype, BMNH 1907.12.3.1, SL 7.9 cm. (3) P. wesselsi Kramer and Van der Bank, 2000, right
side, SL 9.7 cm, SAIAB 85922 (R1). (7) P. longicapitis sp. nov., SMF 28265 (R1), SL 9 cm. (9) P.
okavangensis sp. nov., holotype, SAIAB 030046, right side, SL 6.1 cm. (10) P. catostoma tanensis
Whitehead and Greenwood, 1959, holotype, BMNH 1963.11.29.1, right side, SL 6.7 cm. (10a)
P. tanensis, field no. Ta05na, SL 8.7 cm, SAIAB 85907. (12) P. longianalis sp. nov., holotype, SL
8.2 cm, right side, SAIAB 76758. (21) P. pe t e r s i sp. nov., holotype, SAIAB 060846, right side, SL
6.4 cm. (23) P. magnoculis sp. nov., SL 8.9 cm, ZSM 38659. (24) P. haullevillii Boulenger, 1912,
BMNH 1912.4.1.186–188 (R1), right side, SL 5.7 cm. (25) P. steindachneri Fowler 1958, syntype,
NMW 55118:3, right side, SL 6.4 cm. (26) P. stuhlmanni congicus David and Poll, 1937, syntype,
MRAC 30807-30808, SL 7.8 cm. (27) P. degeni Boulenger, 1906, BMNH 1906.5.30.84, right side,
SL 8.12 cm. (28) P. catostoma, SAIAB 73894 (R1), right side, SL 6.4 cm. (31) P. magnitrunci sp.
nov., SL 8.5 cm, SAIAB 67069 (R5, right side), see map Figure 6. Scale bar, 1 cm.
Journal of Natural History 2183
2184 B. Kramer et al.
Figure 2(A–D). (Continued)
Journal of Natural History 2185
Figure 2(A–D). (Continued)
2186 B. Kramer et al.
Figure 2(A–D). (Continued)
Journal of Natural History 2187
SL
LPF
LA
LD
CPD
PDL
PAL
HL
Na
TL
PPF
LSc
LSo
OD
CPL
pD
BD
Figure 3. Schematic sketch of how measurements were taken on Petrocephalus sp. For abbrevi-
ations, see Material and methods.
Institute for Aquatic Biodiversity, Grahamstown, South Africa (SAIAB); at the
Zoologische Staatssammlung, München, Germany (ZSM); and at the Senckenberg
Forschungsinstitut und Naturmuseum Frankfurt, Frankfurt am Main, Germany
(SFM). Specimens studied were initially identified using dichotomous keys in
Bell-Cross and Minshull (1988) and Skelton (1993, 2001).
Fish sampled from the field were transferred into a 37-litre plastic aquarium filled
with river water where the fish were collected for recording their EODs with mini-
mum delay. Conductivity (±1µS/cm) and temperature (±0.1 ◦C) were monitored
using an electronic meter (LF92 by Wissenschaftlich-Technische Werkstätten WTW,
82362 Weilheim, Germany). Conductivity changes possibly affecting EOD waveform
(Bell et al. 1976; Bratton and Kramer 1988; Kramer and Kuhn 1993) were excluded.
Methods for capturing and analysing EODs are as described in Kramer and Van
der Bank (2000). The three phases to an EOD pulse were head-positive, head-negative,
head-positive (P1, N, P2). Before analysis, EODs were temperature-corrected to 25◦C
using a Q10 of 1.5 (Kramer and Westby 1985), and normalized in amplitude (by setting
the peak amplitude of the P1 phase, measured from baseline, equal to 1). Abbreviations
of EOD parameters: P1amp, P2amp, Namp, peak amplitudes from baseline for the P1,
the P2 and the N phases, respectively; P1dur, P2dur, Ndur, durations of the P1, P2 and
N phases, respectively, with P1 and P2 phases slightly shortened by using an amplitude
criterion of ±2% of P1amp for estimating start or termination, respectively; P1Nsep,
P1P2sep, NP2sep, separation (or interval) between the peaks of the P1 and N phases,
the peaks of the P1 and P2 phases, and the peaks of the N and P2 phases, respectively;
P1area, P2area, Narea, areas under the P1, the P2, or the N phase curves measured
from ±2% P1amp of the baseline.
Statistical analyses as indicated in the Results section; Pvalues are two-
tailed unless otherwise stated. For a Principal Components Analysis (PCA) on
correlations among anatomical characters we estimated eigenvalues, eigenvectors
2188 B. Kramer et al.
and, for interpreting the principal components in terms of the anatomical charac-
ters, the component loadings, i.e. the principal component structure (see McGarigal
et al. 2000). For assessing the significance of loadings we followed Tabachnick
and Fidell (2007). These authors recognize five levels of significance: loadings
>0.32 or <– 0.32 are poor, >0.45 or <–0.45fair,>0.55 or <– 0.55 good, >0.63 or
<– 0.63 very good, and >0.71 or <– 0.71 excellent. These benchmarks account for
10%, 20%, 30%, 40% and 50% of the variance in the component, respectively. The soft-
ware used was STATVIEW v. 5 and JMP v. 7.0.2 to 9 (SAS Institute, Cary, NC, USA,
2007).
Genetic studies
DNA isolation
DNA was isolated from muscle or scale tissue, which was preserved in ethanol, using
a standard phenol/chloroform protocol (Sambrook et al. 1989). The mitochondrial
cytochrome b(cyt b) gene was amplified using the published mitochondrial DNA
primers (Kramer et al. 2007).
The PCR amplifications were performed with 50-µl reaction volumes contain-
ing 1 ×PCR buffer (Bioron, Ludwigshafen, Germany), 100 µMdNTPs, 0.2 units of
Taq DNA polymerase (Bioron, Ludwigshafen, Germany), 200 ng DNA and 5 pmol
primers. Thermal cycling was performed under the following conditions: (1) an initial
denaturing step at 94◦C for 5 min; (2) 35 cycles: 1 min at 94◦C, 1 min at 52◦Cand
1 min at 72◦C; and (3) a final 5-min extension at 72◦C. The PCR products were pre-
cipitated with 4 MNH4Ac and ethanol (1:6) and centrifuged for 15 min (15 550 ×g).
Sequencing was carried out on an ABI 3730 automated capillary sequencer (Applied
Biosystems, 64293 Darmstadt, Germany) with the ABI Prism Big Dye Terminator
Cycle Sequencing Ready Reaction Kit 3.1 by STARSEQ GmbH (Mainz, Germany).
Phylogenetic analyses
The tree reconstruction was performed using the maximum likelihood method with the
substitution model Tamura–Nei and the Nearest-Neighbour-Interchange algorithm.
Bootstrap was carried out with 600 replications and the mean pairwise p-distances
were calculated following Nei (1987). All of these analyses were conducted with
MEGA version 5.0 (Tamura et al. 2011).
Genetic samples examined
(IPBM collection nos =Institut für Pharmazie und Molekulare Biotechnologie,
Heidelberg University, Germany).
Petrocephalus magnoculis sp.nov.(n=6), Namibia: Cunene River: just below
Ruacana Falls, coll. B. Kramer and E. Swartz, 17◦2424 S, 14◦1301 E: IPBM
43982 =ZSM38659#A278, 19 August 2006; IPBM 43983 =SAIAB78788#A288,
20 August 2006; IPBM 43984 =SAIAB79480#A302, 20 August 2006; IPBM 43987 =
ZSM38660#A309, 20 August 2006, IPBM 43991 =SAIAB78790#A364, 22 August
2006; IPBM 43992 =SAIAB78788#A280, 22 August 2007, released.
Petrocephalus longicapitis sp. nov. (n=4), Namibia: East Caprivi: Upper Zambezi:
Katima Mulilo, 17◦2930 S, 24◦1618 E, coll. H. van der Bank August 1994, scale
Journal of Natural History 2189
taken from live fish on 9 February 2009: IPBM 51428, #32, IPBM 51429, #33, IPBM
51430, #34, IPBM 51431, #35.
Petrocephalus catostoma (n=2), Mozambique: Rovuma System: Lucombe River,
coll. R. Bills: IPBM 35836 =SAIAB73891#N303, 26 August 2003, 12.0839◦S,
37.5619◦E; IPBM 35837 =SAIAB73889#N317, 22 August 2003, 12.0875◦S,
37.5606◦E.
Petrocephalus wesselsi (n=4), South Africa: Limpopo System: Mogol (Mokolo)
River at Hermanusdorings, 24◦06.823S, 27◦48.153E, coll. A. Hoffman and B.
Kramer, 20 October 2008, IPBM 50695, IPBM 50696, IPBM 50698, IPBM 50699,
released.
Marcusenius altisambesi (n=2), IPBM 57467, #9, Namibia: Cunene River Mouth,
coll. F.H. Van der Bank, 15 December 2009, 17◦15.606S, 11◦45.892E; IPBM 50679,
#15, Namibia: Upper Zambezi River: Kalimbeza, 17◦3227.3 S, 24◦3126.2 E, coll.
F.H.Van der Bank and B. Kramer, 21 August 1999, tissue sample live fish taken on
10 October 2008.
Material examined
Petrocephalus catostoma (Günther, 1866) and its previous synonyms and subspecies
Mormyrus catostoma Günther, 1866. Lectotype BMNH 1863.10.12.4, 4.7 cm SL, and
four paralectotypes BMNH 1863.10.12.5-6(4), 4.1–4.4 cm SL, for Petrocephalus
catostoma catostoma (Günther, 1866),
BMNH 1906.5.30.84, Petrocephalus degeni Boulenger, 1906, holotype (unique), 8.1 cm
SL, Katonga River, Lake Victoria (Uganda),
BMNH 1907.12.3.1 Petrocephalus stuhlmanni Boulenger, 1909, holotype (unique),
7.8 cm SL,
BMNH 1912.4.1.181-185, Petrocephalus haullevillii Boulenger, 1912, syntypes (5),
3.9–5.9 cm SL, Angola, Portuguese Congo, Lundo, Luali River,
MRAC 1496-1501, Petrocephalus haullevillii Boulenger, 1912, syntypes (6), 4.6–6.4 cm
SL, Angola, Portuguese Congo, Lundo,
BMNH 1912.4.1.186-188, Petrocephalus haullevillii Boulenger, 1912, syntypes (3),
5.7–6.5 cm SL, Democratic Republic of the Congo (Belgian Congo), Lukula River,
NMW 55118(3), -117 (half of a fish), Petrocephalus steindachneri Fowler, 1958,
syntypes (1+3), 6.4–6.6 cm SL, Tanzania, Ulanga, Kiperege, Msola-stream,
MRAC 30807–30808, Petrocephalus stuhlmanni congicus David and Poll, 1937,
syntypes (2), 7.2–7.8 cm SL, Zaire, Congo River basin, Mukishi (Lomami River),
08◦30S, 24◦44E,
SAIAB 73887(9), 3.7–4.5 cm SL, Mbatamila-Matondovela Rd, Litungulu stream near
Matondovelo 18 August 2003, Mozambique, Niassa Reserve, Litungulu Rovuma,
12◦0527S, 37◦1940 E, coll. R. Bills,
SAIAB 73802(10), Petrocephalus catostoma, 3.8–4.4 cm SL, Mbatamila-Matondovela
Rd, third river crossing, 14 August 2003, Mozambique, Niassa Reserve, Rovuma,
12◦0805 S, 37◦2418 E, coll. R. Bills,
SAIAB 73808(10), Petrocephalus catostoma 3.8–4.6 cm SL, Mbatamila-Matondovela
Road, third river crossing, 18 August 2003, Mozambique, Niassa Reserve, Rovuma,
12◦0805 S, 37◦2418 E, coll. R. Bills,
2190 B. Kramer et al.
SAIAB 73894(6), Petrocephalus catostoma 3.8–6.4 cm SL, Mbatamila-Mussoma
Rd, Nkupo stream near Mussoma bridge, 22 August 2003, Mozambique,
Niassa Reserve, Nkupo, Lugenda River (confluence of Rovuma R), 12◦2642 S,
37◦4044 E, coll. R. Bills,
SAIAB 050155(2), Petrocephalus cf. catostoma, 3.8–4.7 cm SL, Malawi, Nkhotakota,
Malenga Chanzi, Pool in stream bed above Lake Chiku, Shire River system,
Kaombe River, 12◦58S, 34◦13E, 25 July 1995, coll. D. Tweddle,
SAIAB 050065(2), Petrocephalus cf. catostoma, 4.8–5.7 cm SL, Malawi, Nkhotakota,
Kanyenda, Below main dam for sugar estate tak, Shire River system, Dwangwa
River, 12◦31S, 34◦07E, 20 July 1995, coll. D. Tweddle,
SAIAB 039328(1), Petrocephalus cf. catostoma, 6.1 cm SL, Malawi, South of
Chinteche, Lake Malawi, Kapesha River, 11◦54S, 34◦09E, 5 July 1992, coll. D.
Tweddle.
Petrocephalus frieli Lavoué, 2012
SAIAB 76825(3), 6.1–7.4 cm SL, Zambia Province: Luapula System: Lake Bangweulu
shoreline at rocky point near Samfya Ferry dock, 11◦2119.44 S, 29◦3347.52 E,
coll: R. Bills, A. Chilala, J. Friel, 25. September 2005, field no. JPF-05-014,
SAIAB 76859(1), 5.6 cm SL, Zambia Province: Luapula System: Lake Bangweulu
shoreline at rocky point near Samfya Zambian Fisheries building, 11◦2220.64
S, 29◦3353.64 E, coll: R. Bills, A. Chilala, J. Friel, 25 September 2005, field no.
JPF-05-015,
Petrocephalus longianalis sp. nov.
SAIAB 76758, holotype, specimen R9, 8.2 cm SL, Zambia Province: Luapula System:
Luongo River: Lufubu River, Lufubu River Falls below bridge at Chipili on Mensa-
Mununga road, 10◦4346.92 S, 29◦0536.96 E, coll: R. Bills, A. Chilala, J. Friel,
2 October 2005, field no. JPF-05-025,
SAIAB 186060(48), paratypes, 3.8–8.2 cm SL, Zambia Province: Luapula System:
Luongo River: Lufubu River, Lufubu River Falls below bridge at Chipili on Mensa-
Mununga road, 10◦4346.92 S, 29◦0536.96 E, coll: R. Bills, A. Chilala, J. Friel,
2 October 2005, field no. JPF-05-025,
Non-types: SAIAB 76582(5), 7.0–7.3 cm SL, Zambia Province: Central System:
Luapula River: Luapula, Luapula River Bridge, 12◦0656.16 S, 29◦5049.92 E,
coll: R. Bills, A. Chilala, J. Friel, 22 September 2005, field no. JPF-05-006,
SAIAB 76733(2), 10.1–11.9 cm SL, Zambia Province: Luapula System: Luongo River
at bridge on Kashiba-Mwenda road, 10◦2812.72 S, 29◦0128.2 E, coll: R. Bills,
A. Chilala, J. Friel, 1 October 2005, field no. JPF-05-023.
Petrocephalus longicapitis sp. nov.
SAIAB 85916 Holotype, 19fish, 8.4 cm SL, Upper Zambezi River at Katima Mulilo,
East Caprivi, Namibia, rocks in middle of river (opposite boat landing), approx.
17◦2930 S, 24◦1618 E, 10 September 1993, coll. F.H. Van der Bank and B.
Kramer,
Journal of Natural History 2191
Paratypes: SMF 28265(27), 3.8–9.0 cm SL (one of which 2.8 cm SL); SAIAB 85911(2),
25fish, 26fish, 8.4–8.5 cm SL; SAIAB 85917(2), 14fish, 16fish, 8.0–8.4 cm SL;
SAIAB 85918(3), 37fish, 38fish, 43fish, 7.6–7.9 cm SL; all from same location,
10–13 September 1993, water conductivity and temperature, 81 µS/cm, 21.8◦C, SL
from 28 –105 mm, coll. F.H. Van der Bank and B. Kramer,
Non-types: ZSM 38657(1), L29isi, 10.3 cm SL, Lisikili backwater of Zambezi down-
stream of Katima Mulilo, 17◦29S, 24◦26E, 6 March 1994, 56.1 µS/cm and 26.8◦C,
gravid female, coll. F.H. Van der Bank and B. Kramer,
Non-types: ZSM 38658(1), N53ak, 9.5 cm SL, specimen from Kwando River,
Nakatwa, 18◦06S, 23◦23E, 9 March 1994, 130 µS/cm and 24.9◦C, gravid female,
coll. F.H. Van der Bank and B. Kramer,
Non-types: SAIAB 85909(3), Ven02, Ven03, Ven09, 8.1–9.1 cm SL, from Zambezi
rapids at Wenela just upstream of Katima Mulilo (border post to Zambia;
17◦2921.5 S, 24◦1533 E, 9 September 1997, coll. F.H. Van der Bank and B.
Kramer,
Non-types: SAIAB 041208(5), 6.1–7.3 cm SL, Zambia, North West Province,
E Lumwana, confluence of Mwambezhi and East Lumwana Rivers (Upper
Kabompo/Zambezi system), off Mwinilunga-Solwezi road, 12◦15S, 25◦40E,
31 July 1983, coll. R. Bills,
Non-types: SAIAB 041025(1), 3.6 cm SL, Zambia, North West Province, Off Solwezi-
Mwinilunga Road, Zambezi River system, Kabompo River, Lumwana River, 12◦15
S, 25◦40E, 31 July 1983, coll. R. Bills,
Non-types: SAIAB 042559(18), 4.1–10.4 cm SL, specimens of Petrocephalus from
Zambia, below dam and fish ladder, Kafue/Zambezi River system, Mwekera
Stream, 12◦40S, 28◦30E, 1 July 1983, coll. R. Bills,
Non-types: SAIAB 41224(4), 10.2–11.0 cm SL, specimens of Petrocephalus from
Zambia, Kafue System, Mwekera Stream pool below waterfalls, 12◦4000 S,
28◦3000 E, 4 July 1983, coll. R. Bills,
Non-types: SAIAB 040074(1), 7.7 cm SL, Petrocephalus, Zambia, East Lunga River,
Kafue/Zambezi River system, Lunga River, 14◦00S, 26◦30E, 17 April 1983, coll.
R. Bills.
Petrocephalus magnitrunci sp. nov.
SAIAB 67069, holotype, specimen R2, 8.8 cm SL, Botswana, Okavango Delta, south-
east of Chief’s Island, Boro River, 19◦3157 S, 023◦0521 E, 20 June 2000, coll. D.
Tweddle and B.C.W. van der Waal,
SAIAB 186057(10), paratypes, 7.7–8.8 cm SL, Botswana, Okavango Delta, southeast
of Chief’s Island, Boro River, 19◦3157 S, 023◦0521 E, 20 June 2000, coll. D.
Tweddle and B.C.W. Van der Waal.
Petrocephalus magnoculis sp. nov.
SAIAB 78788, holotype, specimen Ruac06, 9.6 cm SL, from Cunene River, Ruacana
Falls, Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about
800 m altitude; from 19 August 2006, coll. B. Kramer and E. Swartz,
ZSM 38659(1), paratype, Ruac07, 8.9 cm SL, from Cunene River, Ruacana Falls,
Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about 800 m
altitude; 19 August 2006, coll. B. Kramer and E. Swartz,
2192 B. Kramer et al.
SAIAB 186053(1), paratype, Ruac09, 10.5 cm SL, from Cunene River, Ruacana Falls,
Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about 800 m
altitude; 20 August 2006, coll. B. Kramer and E. Swartz,
SAIAB 79480(1), paratype, Ruac10, 9.4 cm SL, from Cunene River, Ruacana Falls,
Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about 800 m
altitude; 19 August 2006, coll. B. Kramer and E. Swartz,
ZSM 38660(1), paratype, Ruac13, 9.1 cm SL, from Cunene River, Ruacana Falls,
Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about 800 m
altitude; 21 August 2006, coll. B. Kramer and E. Swartz,
SAIAB 78790(1), paratype, Ruac17, 9.6 cm SL, from Cunene River, Ruacana Falls,
Hippo Pool Campsite, just below the Falls, 17◦2424 S, 14◦1301 E, about 800 m
altitude; from 22 August 2006, coll. B. Kramer and E. Swartz,
SAIAB 028120(3), non-types, 9.1–10.5 cm SL, Petrocephalus sp., Namibia, “Hippo
Pool”, Ruacana Falls, Cunene River system, Cunene River, 17◦24S, 14◦12E,
5 October 1986, coll B. van Zyl.
Petrocephalus okavangensis sp. nov.
SAIAB 030046, holotype, specimen R22, 6.1 cm SL, Botswana, Okavango, Thoage
River, Nguma (Guma) Lagoon, 18◦5660 S, 22◦2259.99 E, 3 January 1987, coll.
G. Merron,
SAIAB 186062(41), paratypes, 4.2–7.8 cm SL, Botswana, Okavango, Thoage River,
Nguma (Guma) Lagoon, 18◦5660 S, 22◦2259.99 E, 3 January 1987, coll. G.
Merron,
Non-types: ZSM 38665(3), 8.0–8.1 cm SL, Botswana, Okavango, Thoage River, Guma
Lagoon, 18◦5746.6 S, 22◦2225.3 E, 10 August 2004, coll. F. H. Van der Bank and
B. Kramer,
one specimen for EOD only, Botswana, Okavango, Thoage River, 19◦0345.3 S,
22◦2324.3 E, 23 March 2002, 26◦C water temperature, 50 µS/cm, EOD recorded
7 March 2003 Regensburg,
one specimen for EOD only, Botswana, Okavango, Thoage River, Guma Lagoon,
18◦5746.6 S, 22◦2225.3 E, 17.1◦C water temperature, 38 µS/cm, EOD recorded
12 August 2004,
SAIAB 36841(24), 3.7–6.6 cm SL, Namibia, Okavango River, Popa Rapids,
18◦0600 S, 21◦3600 E, 15 July 1986, coll. P. Skelton,
SAIAB 36823(12), 3.6–7.8 cm SL, same location, second island, 14 July 1986, coll. T.
Andrew, G. Merron, P. Skelton,
SAIAB 19769(21), 4.5–8.3 cm SL, Botswana, Okavango Delta, Moanachira River,
Gadikwe Lagoon, mid island 19◦1000 S, 23◦1400 E, 3 November 1983, coll. G.
Merron,
SAIAB 19705(11), 5.6–9.5 cm SL, Botswana, Okavango Delta, Moremi Game
Reserve, Xakanixa Channel, opposite Safari Lodge, upper swamp 19◦1000 S,
23◦2400 E, 2 November 1983, coll. G. Merron,
SAIAB 21271(38), 3.6–9.3 cm SL, Botswana, Okavango Delta, Moremi Game
Reserve, Xakanixa River, 19◦1500 S, 23◦1500 E, 24 June 1984, coll. G. Merron
andG.May.
Journal of Natural History 2193
Petrocephalus petersi sp. nov.
SAIAB 060846, holotype, specimen R1, 6.4 cm SL, Mozambique, stream near
campsite 1, edge of wet Zambezi River System, Zambezi River, 18◦3354 S,
35◦3946 E, 1 August 1999, coll. R. Bills,
SAIAB 186054(10), paratypes, 4.1–6.1 cm SL, Mozambique, stream near campsite
1, edge of wet Zambezi River System, Zambezi River, 18◦3354 S, 35◦3946 E,
1 August 1999, coll. R. Bills,
SAIAB 055875(4), non-types, 7.1–7.3 cm SL, Mozambique, Zambesia, Mulela Village,
Mulela River, Mulalae, 16◦5342 S, 38◦1727 E, 20 July 1997, coll. R. Bills.
Petrocephalus tanensis (Whitehead and Greenwood, 1959), elevated to species rank
Holotype BMNH 1963.11.29.1, Petrocephalus catostoma tanensis Whitehead and
Greenwood, 1959, 6.7 cm SL, Lower Tana River, Garsen, Kenya; BMNH
1963.11.29.2-8, 8 paratypes, 3.0–4.5 cm SL, same locality,
Non-types: NMK nos 24102, 24108, 24109, 24111, 24112, 24126, 24143, 24144, 24151,
24159, 23902, 23903, 23905, 23906, 23910, 23911, 23912, 23913, 239NotLegible,
23915, 6371, 6372, 6373, 6375, 6378, 63713, 63714, 63715, 63717, 29 specimens
of “Petrocephalus catostoma tanensis”, 5.6–8.1 cm SL, Kenya, Lower Tana River,
Garsen,
ZSM 38661(3), Ta01na, Ta04na, Ta09na, 8.0–8.2 cm SL; ZSM 38662(3), Ta15na,
Ta17na, Ta26na, 8.3–9.7 cm SL; ZSM 38663(1), Ta06na, 9.6 cm SL; ZSM 38664(2),
Ta39na, Ta44na, 6.6–6.7 cm SL; SAIAB 85906(8), Ta35na, Ta36na, Ta37na,
Ta38na, Ta40na, Ta41na, Ta42na, Ta43na, 5.8–8.5 cm SL; SAIAB 8907(5), Ta02na,
Ta03na, Ta05na, Ta07na, Ta08na, 8.3–8.8 cm SL, SAIAB 85908(5), SinEOD2,
SinEOD3, SinEOD6, R1, R2, 6.4–9.8 cm SL; Lower Tana River at Tana Primate
Research Reserve near village Wenje, east of road B8, 1◦5238.1 S, 40◦822.5 E,
48 m above sea level, 3–6 September 2001, coll. L. De Vos and B. Kramer,
23 specimens for EOD, 186 µS/cm and 25.7◦C
Petrocephalus wesselsi Kramer and Van der Bank, 2000
ZSM 28556, holotype, 10.8 cm SL; ZSM 28554–ZSM 28555, ZSM 28557–ZSM 28566
(12 paratypes, 5.6–10.1 cm SL), SMF 28266 (13 paratypes, 5.7–8.9 cm SL), SAIAB
054449 (13 paratypes, 5.4–8.4 cm SL), all from Sabie River, Kruger National
Park, South Africa, bridge near Lower Sabie tourist camp (25◦07S, 31◦55E),
29–30 March 1996, coll. F.H. Van der Bank and B. Kramer, 139 µS/cm and 25.1◦C,
Non-types: SAIAB 85922(5), 2.0–3.0 cm SL, same time and place etc. as previous
paragraph,
SAIAB 58157(9), SL 4.7–9.8 cm, Nwanedzi River (Limpopo system) at Nwanedi,
Northern Province, South Africa, below dams, 22◦3745 S, 30◦2352 E, 25 March
1997, coll B. van der Waal,
SAIAB 85920(2), 4.8–5.7 cm SL, Groot Letaba River (Olifants System, Limpopo
drainage) just below Tzaneen Dam, Northern Province, South Africa, 23◦4900
S, 30◦1000 E, 22 September 1998, coll. W. Vlok and B. Kramer, 114 µS/cm and
21.4◦C,
2194 B. Kramer et al.
SAIAB 85923(6), 3.8–9.0 cm SL, Blyde River (Olifants System, Limpopo drainage)
just below Blydepoortriviers Dam, Mpumalanga, South Africa, 24◦3200 S,
30◦4705 E, 25/26 September 1998, coll. J. Engelbrecht and B. Kramer, 154 µS/cm
and 16.7◦C, 82–90 mm SL,
SAIAB 85919(1), 7.2 cm SL, Pongola River, KwaZulu-Natal, South Africa, at bridge
on road from Ndumo to Kosibay, 27◦0115 S, 32◦18E, 14 August 1999, coll. J.
Engelbrecht and B. Kramer, 600 µS/cm and 22◦C,
SAIAB 068279(2), 4.6–8.8 cm SL, Mnjoli Dam Wall, Mbuluzi River, Swaziland,
26◦0941 S, 31◦4014 E, 29 January 2003, coll. R.C. Boycott,
SAIAB 067228(1), 8.9 cm SL, Mnjoli Dam Wall, Mbuluzi River, Swaziland,
26◦0941 S, 31◦4014 E, 14 August 2002, coll. R.C. Boycott,
SAIAB 066355(1), 9.8 cm SL, Mnjoli Dam Wall, Mbuluzi River, Swaziland,
26◦0926 S, 31◦4010 E, 19 July 2002, coll. R.C. Boycott,
SAIAB 95989(30), 3.7–6.0 cm SL, specimens: Mogol05–Mogol14, Mogol17,
Mogol20, Mogol21, Mogol30, Mogol31a, Mogol31b, Mogol39–Mogol51; and
ZSM 39537(8), 4.5–5.9 cm SL, specimens: Mogol52–Mogol56, Mogol58–Mogol60;
and ZSM 39538 (10), 4.3–5.4 cm SL, specimens: Mogol61–Mogol66, MogolDead1–
MogolDead3, ohne Fish-ID; Limpopo System: Mokolo River, 24◦06.823S,
27◦48.153E, altitude 932 m, near Hermanusdorings, 20 October 2008, coll. A.
Hoffman and B. Kramer,
SAIAB 96537(2), 4.8 cm SL, specimens Palala04, Palala05, Limpopo System: Lepalala
River, 23◦59.049S, 28◦24.281E, 1144 m altitude, near Melkrivier, 22 October 2008,
coll. A. Hoffman and Bernd Kramer.
Other material examined, of uncertain status
SAIAB 059515(1), 8.5 cm SL, Petrocephalus sp., Tanzania, Lake Rukwa at mouth of
Luika River, Lake Rukwa, Luika, 8◦245 S, 32◦5420 E, 19 November 1995, coll.
P. Skelton,
SAIAB 039264(1), 7.7 cm SL, Petrocephalus sp., Malawi, Zikanyeka Beach, Lake
Chiuta, 14◦43S, 35◦51E, 13 July 1992, coll. D. Tweddle.
Results
Morphological comparisons
Comparisons between nominal species
Our first concern was to investigate whether or not our own sample from the type
region represented P. catostoma. David Livingstone collected the five type specimens
from the lower reaches of the Rovuma River in 1859, but the exact type locality is
unknown (no. 1, Figure 1; Livingstone 1865). The fresh material from locality no. 28
(Figure 1; n=35) corresponds very well to the type material. The medians for the
meristic characters were identical, and the mensural characters similar, with the bigger
sample usually showing a wider range that is overlapping the smaller. We conclude that
our fresh specimens from the Rovuma drainage represent the species P. catostoma,and
will use them for comparisons with nominal species and statistical comparisons with
allopatric populations (Table 1).
Petrocephalus stuhlmanni congicus David and Poll, 1937 (n=2; locality 26,
Figure 1). Whitehead and Greenwood (1959) had seen differentiation from “the
eastern and southern subspecies, particularly in its lower modal number of dorsal fin
Journal of Natural History 2195
Table 1. Morphometrics of the southern and eastern African Petrocephalus species.
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL LSo/HL HL/SL HL/Na BD/SL nD nA SPc SLS OD/HL LPF/HL PPF/SL SL (cm)
P. catostoma - Types
Lecto-, BMNH
1863.10.12.4
0.631 0.56 0.175 0.229 0.441 0.237 0.318 0.318 0.441 0.249 18.47 0.292 21 29 12 4.74
Mean/Median∗0.628 0.581 0.167 0.227 0.423 0.22 0.338 0.319 0.436 0.259 18.82 0.289 20 27 14 4.32
SE/SIQ∗0.001 0.006 0.004 0.008 0.007 0.006 0.008 0.003 0.005 0.004 0.21 0.002 0.5 1 0.25 0.12
Min 0.625 0.56 0.159 0.196 0.41 0.207 0.318 0.309 0.422 0.249 18.47 0.283 20 27 12 4.1
Max 0.631 0.593 0.18 0.244 0.441 0.237 0.356 0.33 0.45 0.274 19.64 0.293 21 29 14 4.74
n555555 5 5 5555555 5
Rovuma System
Mean/Median∗0.625 0.578 0.174 0.233 0.411 0.21 0.331 0.326 0.423 0.277 17.433 0.274 20 27 14 4.2
SE/SIQ∗0.002 0.002 0.001 0.001 0.002 0.002 0.003 0.003 0.002 0.001 0.21 0.002 0.75 0.5 0.75 0.08
Min 0.602 0.549 0.16 0.219 0.383 0.195 0.296 0.287 0.398 0.259 16.07 0.252 18 25 12 3.7
Max 0.656 0.602 0.188 0.252 0.434 0.238 0.385 0.356 0.454 0.291 23.21 0.306 22 28 16 6.4
n35 35 35 35 35 35 35 35 35 35 35 35 35 35 35 35
Lake Malawi
confluences
Mean/Median∗0.637 0.584 0.174 0.239 0.421 0.21 0.341 0.337 0.451 0.267 19.58 0.292 21 26 16 5
SE/SIQ∗0.005 0.004 0.002 0.002 0.007 0.005 0.016 0.003 0.006 0.003 1.43 0.007 0.125 0.5 2 0.39
Min 0.627 0.57 0.167 0.233 0.393 0.201 0.289 0.328 0.44 0.256 16 0.266 20 26 12 3.9
Max 0.652 0.596 0.179 0.246 0.435 0.228 0.386 0.347 0.473 0.276 24.81 0.305 21 27 16 6.1
n555555 5 5 5555555 5
P. steindachneri -
Types
1 of 3 syn-, NMW
55118:1
0.623 0.61 0.206 0.252 0.425 0.204 0.299 0.312 0.437 0.273 27.77 0.294 26 31 12 6.5
Mean/Median∗0.632 0.609 0.193 0.238 0.42 0.204 0.3 0.322 0.436 0.273 26.25 0.308 24.5 30 12 6.5
SE/SIQ∗0.005 0.004 0.01 0.009 0.009 0.004 0.006 0.011 0.001 0.001 2.72 0.01 2 1.75 0 0.054
Min 0.623 0.601 0.173 0.222 0.402 0.195 0.285 0.311 0.434 0.272 20.98 0.294 22 26 12 6.4
Max 0.641 0.616 0.206 0.252 0.432 0.211 0.313 0.343 0.438 0.274 30 0.328 27 31 12 6.6
n333333 4 3 3333444 3
P. stuhlmanni - Type,
BMNH 1907.12.3.1
(n=1)
0.643 0.596 0.152 0.239 0.398 0.216 0.389 0.288 0.382 0.276 25.83 0.301 21 27 12 7.86
P. catostoma tanensis -
Types
Holo-, BMNH
1963.11.29.1
0.598 0.594 0.205 0.247 0.449 0.217 0.343 0.325 0.407 0.294 31.13 0.318 25 28 16 6.7
(Continued)
2196 B. Kramer et al.
Table 1. (Continued).
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL LSo/HL HL/SL HL/Na BD/SL nD nA SPc SLS OD/HL LPF/HL PPF/SL SL (cm)
Mean/Median∗0.598 0.562 0.198 0.246 0.453 0.215 0.386 0.328 0.43 0.283 23.94 0.305 25 28 14 4.1
SE/SIQ∗0.003 0.006 0.002 0.003 0.001 0.002 0.007 0.005 0.006 0.003 1.06 0.003 0 0.5 0 0.35
Min 0.589 0.545 0.189 0.234 0.449 0.207 0.343 0.305 0.407 0.269 19.51 0.281 24 26 12 3
Max 0.619 0.594 0.205 0.257 0.462 0.225 0.413 0.354 0.458 0.301 31.13 0.318 25 28 16 6.7
n999999 9 9 99 99999 9
Tan a River
Mean/Median∗0.612 0.583 0.192 0.239 0.448 0.223 0.346 0.327 0.424 0.266 32.85 0.299 24 28 12 37 0.232 0.742 0.164 7.4
SE/SIQ∗0.002 0.002 0.001 0.001 0.001 0.001 0.004 0.002 0.002 0.001 0.46 0.003 0.625 0.5 1 0.5 0.003 0.015 0.002 0.15
Min 0.574 0.554 0.167 0.214 0.427 0.193 0.291 0.297 0.399 0.237 27.31 0.257 22 26 12 36 0.206 0.507 0.143 5.6
Max 0.645 0.618 0.222 0.257 0.472 0.247 0.439 0.382 0.457 0.282 42.83 0.356 27 29 14 38 0.276 0.862 0.180 9.8
n54 54 54 54 54 54 54 54 54 54 54 54 53 54 54 27 27 27 27 54
P. degeni - Type,
BMNH
1906.5.30.84 (n=1) 0.634 0.641 0.155 0.221 0.427 0.202 0.369 0.258 0.411 0.261 20.59 0.29 19 27 12 8.12
P. stuhlmanni congicus - Syntypes (n=2)
MRAC 30807-30808 0.635 0.535 0.166 0.241 0.438 0.237 0.297 0.343 0.429 0.231 18.87 0.319 2712 7.2
MRAC 30808-30808 0.618 0.578 0.155 0.234 0.417 0.228 0.296 0.326 0.408 0.238 21.32 0.286 18 28 12 7.8
P. haullevillii – Syntypes (n=5) BMNH
1912.4.1.181–185 0.607 0.549 0.177 0.258 0.44 0.228 0.306 0.292 0.40.226 21.71 0.266 18 27 12 5.9
Mean/Median∗0.618 0.564 0.168 0.254 0.421 0.22 0.326 0.322 0.429 0.24 20.03 0.283 19 27 12 5.41
SE/SIQ∗0.003 0.003 0.002 0.002 0.003 0.002 0.008 0.005 0.006 0.002 0.21 0.005 0.5 0 0 0.23
Min 0.605 0.542 0.154 0.24 0.401 0.209 0.277 0.292 0.4 0.225 18.86 0.24 18 26 12 3.9
Max 0.641 0.586 0.177 0.269 0.44 0.231 0.365 0.362 0.484 0.248 21.71 0.312 21 28 12 6.5
n14 14 14 14 14 14 14 14 14 14 14 14 14 14 14 14
Lower Zambezi - P. p e t e rs i sp. nov.
Holo-, SAIAB
060846
0.656 0.625 0.166 0.235 0.406 0.197 0.378 0.355 0.449 0.268 22.96 0.323 18 26 16 6.4
Mean/Median∗0.642 0.616 0.166 0.228 0.409 0.197 0.356 0.364 0.462 0.278 21.87 0.299 20 26 16 5
SE/SIQ∗0.003 0.004 0.002 0.002 0.003 0.002 0.006 0.003 0.004 0.002 0.52 0.003 0.5 0.5 1 0.2
Min 0.623 0.595 0.151 0.215 0.391 0.185 0.324 0.345 0.442 0.268 20.15 0.287 18 25 12 4.1
Max 0.658 0.637 0.182 0.242 0.425 0.208 0.39 0.383 0.479 0.288 26.35 0.323 20 26 16 6.4
n11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11
(Continued)
Journal of Natural History 2197
Table 1. (Continued).
Mulela River
Mean/Median∗0.645 0.617 0.176 0.24 0.405 0.188 0.402 0.359 0.453 0.264 29 0.308 20 27 16 7.2
SE/SIQ∗0.005 0.003 0.004 0.05 0.005 0.007 0.013 0.004 0.005 0.001 0.78 0.002 0 1.25 0 0.04
Min 0.634 0.607 0.168 0.229 0.391 0.167 0.378 0.346 0.44 0.263 27.9 0.303 20 25 16 7.1
Max 0.655 0.623 0.185 0.253 0.413 0.199 0.438 0.367 0.461 0.266 31.23 0.312 20 28 16 7.3
n444444 4 4 44 44 444 4
Lake Rukwa (n=1) 0.573 0.638 0.264 0.191 0.461 0.225 0.365 0.385 0.515 0.239 14.49 0.288 28 21 12 8.5
Lake Chiuta (n=1) 0.632 0.592 0.168 0.222 0.441 0.213 0.36 0.336 0.429 0.264 27.35 0.28 21 28 16 7.7
Upper Zambezi System
P. longicapitis sp. nov.
Holotype, SAIAB
85916
0.647 0.612 0.18 0.242 0.40.194 0.369 0.333 0.425 0.277 26.86 0.307 22 28 12 37 0.251 0.749 0.187 8.4
Katima -
Mean/Median∗
0.635 0.596 0.184 0.237 0.411 0.199 0.36 0.321 0.431 0.279 23.5 0.295 23 28 12 38 0.261 0.727 0.171 6.8
SE/SIQ∗0.001 0.002 0.001 0.001 0.001 0.001 0.003 0.003 0.003 0.001 0.59 0.002 0.5 0.5 0 0.25 0.003 0.01 0.004 0.29
Min 0.619 0.575 0.169 0.221 0.391 0.18 0.334 0.277 0.389 0.258 17.29 0.277 21 26 12 37 0.249 0.655 0.153 3.8
Max 0.657 0.617 0.208 0.252 0.427 0.215 0.398 0.347 0.458 0.296 30.11 0.315 25 30 12 39 0.281 0.775 0.196 10.3
n38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 12 12 12 12 38
East Lumwana R -
Mean/M∗
0.64 0.603 0.177 0.224 0.417 0.202 0.368 0.346 0.47 0.281 26.2 0.294 22 27.5 12 6.3
SE/SIQ∗0.008 0.01 0.001 0.005 0.007 0.004 0.012 0.006 0.008 0.007 2.32 0.008 0.5 1.5 0 0.58
Min 0.616 0.575 0.174 0.216 0.401 0.186 0.325 0.322 0.439 0.254 18.91 0.27 21 26 12 3.6
Max 0.664 0.629 0.182 0.251 0.45 0.21 0.403 0.36 0.487 0.298 33.09 0.317 23 29 12 7.3
n666666 6 6 66 66 666 6
Waya ma L -
Mean/Median∗
0.63 0.587 0.172 0.256 0.425 0.207 0.321 0.363 0.462 0.254 23.83 0.316 22.5 31 12 38 0.237 0.687 0.159 8.0
SE/SIQ∗0.008 0.007 0.004 0.009 0.004 0.005 0.02 0.004 0.004 0.007 1.32 0.005 1 0.25 0 1 0.004 0.013 0.004 0.42
Min 0.614 0.572 0.162 0.232 0.416 0.195 0.287 0.356 0.455 0.244 20.59 0.306 21 30 12 37 0.225 0.658 0.149 7.3
Max 0.65 0.607 0.18 0.268 0.436 0.215 0.377 0.372 0.474 0.273 27.04 0.328 24 31 12 41 0.246 0.719 0.168 9.1
n444444 4 444 44 44444 4 4 4
Kwando River (n=1) 0.647 0.631 0.187 0.228 0.402 0.183 0.407 0.274 0.378 0.259 29.02 0.309 22 28 12 38 0.255 0.784 0.196 9.5
(Continued)
2198 B. Kramer et al.
Table 1. (Continued).
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL LSo/HL HL/SL HL/Na BD/SL nD nA SPc SLS OD/HL LPF/HL PPF/SL SL (cm)
Kafue System:MwekeraStream
Mean/Median∗0.636 0.593 0.182 0.23 0.422 0.202 0.367 0.343 0.465 0.273 25.52 0.287 23 28 12 38.5 0.245 0.703 0.169 6.9
SE/SIQ∗0.002 0.002 0.001 0.002 0.002 0.002 0.006 0.003 0.003 0.002 1.05 0.004 0.5 0 0 0.75 0.004 0.016 0.004 0.54
Min 0.622 0.577 0.171 0.202 0.402 0.184 0.314 0.315 0.44 0.257 18.79 0.267 21 27 12 38 0.236 0.669 0.16 4.1
Max 0.658 0.628 0.193 0.243 0.439 0.215 0.439 0.368 0.483 0.29 36.74 0.328 25 30 14 40 0.252 0.743 0.176 11
n22 22 22 22 22 22 22 22 22 22 22 22 22 22 22 4 4 4 4 22
Kafue S: Lunga R (n
=1)
0.6573 0.6158 0.1625 0.2491 0.4278 0.2104 0.3665 0.3465 0.4798 0.2653 28.14 0.3191 21 30 12 7.64
Okavango River,P. okavangensis sp. nov.
Holo-, SAIAB
030046
0.64 0.587 0.177 0.275 0.408 0.189 0.333 0.277 0.408 0.264 21.48 0.303 22 30 12 6.1
Guma L -
Mean/Median∗
0.623 0.576 0.17 0.264 0.426 0.207 0.311 0.309 0.439 0.25 19.68 0.291 22 30 12 37 0.253 0.685 0.163 5.6
SE/SIQ∗0.002 0.002 0.001 0.001 0.002 0.001 0.003 0.004 0.003 0.001 0.51 0.003 1 0.5 0 0.375 0.014 −0.001 0.14
Min 0.589 0.538 0.145 0.247 0.403 0.182 0.264 0.261 0.385 0.232 15.11 0.261 20 27 12 37 0.229 0.641 0.161 4.2
Max 0.66 0.619 0.195 0.286 0.454 0.23 0.349 0.372 0.478 0.274 28.97 0.348 24 32 12 38 0.277 0.728 0.165 8.1
n45 45 45 45 45 45 45 45 45 45 45 45 45 45 45 3 3 2 3 45
Gadikwe L -
Mean/Median∗
0.614 0.578 0.171 0.256 0.426 0.211 0.295 0.355 0.466 0.261 20.54 0.3 22 30 12 38 0.256 0.729 0.146 5.7
SE/SIQ∗0.002 0.004 0.002 0.002 0.002 0.002 0.004 0.002 0.003 0.002 0.57 0.002 1 0.125 0 0.5 0.003 0.007 0.001 0.2
Min 0.595 0.553 0.153 0.242 0.404 0.191 0.276 0.334 0.434 0.247 16.89 0.279 20 29 11 36 0.234 0.675 0.135 4.5
Max 0.636 0.626 0.192 0.272 0.444 0.226 0.333 0.376 0.494 0.296 27.54 0.322 26 32 14 39 0.285 0.786 0.161 8.3
n21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21
Popa Rap -
Mean/Median∗
0.618 0.576 0.181 0.252 0.428 0.207 0.306 0.364 0.479 0.271 19.72 0.29 23 29 12 37 0.265 0.703 0.148 4.9
SE/SIQ∗0.003 0.002 0.001 0.002 0.002 0.002 0.004 0.002 0.002 0.003 0.49 0.002 0.75 0.75 0 1 0.002 0.008 0.001 0.16
Min 0.58 0.554 0.165 0.229 0.399 0.191 0.252 0.347 0.454 0.247 15.22 0.259 22 27 12 35 0.242 0.6 0.132 3.7
Max 0.658 0.619 0.195 0.275 0.457 0.229 0.352 0.387 0.509 0.32 25.9 0.329 25 32 12 39 0.289 0.799 0.170 7.8
n36 36 36 36 36 36 36 36 36 36 36 36 36 36 36 36 36 36 36 36
Xakanixa Ch -
Mean/Med∗
0.624 0.586 0.172 0.255 0.418 0.204 0.317 0.342 0.456 0.262 20.96 0.307 22 29 12 37 0.241 0.69 0.151 6.5
SE/SIQ∗0.004 0.007 0.002 0.004 0.005 0.004 0.009 0.003 0.004 0.004 1.18 0.006 0.75 0.5 0 0.875 0.003 0.007 0.004 0.35
Min 0.604 0.556 0.159 0.234 0.386 0.183 0.282 0.322 0.431 0.243 17.72 0.265 20 27 12 36 0.228 0.650 0.137 5.6
Max 0.652 0.624 0.179 0.284 0.437 0.221 0.359 0.360 0.481 0.290 30.96 0.351 23 32 12 38 0.258 0.721 0.176 9.5
n11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11
(Continued)
Journal of Natural History 2199
Table 1. (Continued).
Xakanixa R -
Mean/Med∗
0.615 0.575 0.182 0.259 0.433 0.211 0.302 0.341 0.459 0.255 19.48 0.291 22 30 12 37 0.256 0.697 0.144 6.4
SE/SIQ∗0.003 0.004 0.001 0.002 0.002 0.002 0.007 0.003 0.003 0.002 0.5 0.003 0.5 1 0 0.5 0.002 0.006 0.003 0.25
Min 0.585 0.537 0.168 0.214 0.397 0.178 0.216 0.310 0.425 0.239 15.1 0.254 21 27 12 36 0.226 0.618 0.122 3.6
Max 0.655 0.64 0.199 0.278 0.458 0.233 0.382 0.381 0.5 0.281 26.6 0.346 25 32 12 39 0.289 0.778 0.183 9.3
n38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 38 38
Okavango delta: Boro R., P. magnitrunci sp. nov.
Holotype, SAIAB
67069
0.641 0.597 0.178 0.27 0.429 0.194 0.34 0.358 0.467 0.248 24.10.345 21 31 12 40 0.25 0.671 0.171 8.9
Boro R -
Mean/Median∗
0.643 0.607 0.172 0.254 0.413 0.192 0.337 0.358 0.464 0.25 21.19 0.344 21 29 12 40 0.232 0.658 0.173 8.3
SE/SIQ∗0.004 0.003 0.002 0.003 0.003 0.003 0.008 0.002 0.003 0.002 0.68 0.004 0.5 1.25 0 0.875 0.002 0.008 0.002 0.12
Min 0.621 0.59 0.162 0.235 0.396 0.176 0.296 0.346 0.449 0.239 17.61 0.327 19 27 11 39 0.218 0.609 0.160 7.7
Max 0.658 0.626 0.178 0.27 0.43 0.215 0.376 0.369 0.481 0.258 24.1 0.368 22 31 12 41 0.250 0.698 0.185 8.9
n11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11 11
Cunene River, P. magnoculis sp. nov.
Holo-, SAIAB 78788 0.617 0.607 0.179 0.238 0.427 0.201 0.354 0.391 0.47 0.256 24.34 0.314 22 29 13 38 0.292 0.685 0.173 9.6
Mean/Median∗0.632 0.611 0.18 0.234 0.423 0.199 0.327 0.37 0.485 0.258 27.89 0.312 23 29 12 40 0.294 0.706 0.167 9.7
SE/SIQ∗0.005 0.005 0.003 0.004 0.003 0.003 0.008 0.005 0.005 0.003 0.83 0.006 1 1 0 1 0.01 0.013 0.003 0.22
Min 0.612 0.59 0.17 0.215 0.408 0.184 0.283 0.352 0.47 0.25 24.34 0.283 20 26 11 38 0.259 0.662 0.158 8.9
Max 0.661 0.628 0.194 0.256 0.433 0.218 0.354 0.391 0.514 0.277 31.09 0.34 24 31 13 42 0.329 0.746 0.179 10.9
n999999 9 999 99 99966 6 6 9
P. wesselsi (Sabie River)
Holotype, ZSM
28556
0.621 0.595 0.178 0.248 0.43 0.199 0.42 0.296 0.382 0.26 24.09 0.307 20 26 16 10.8
Mean/Median∗0.634 0.593 0.171 0.238 0.417 0.212 0.381 0.34 0.432 0.271 21.71 0.318 20 26 16 37 0.212 0.709 0.167 6.8
SE/SIQ∗0.002 0.001 0.001 0.001 0.002 0.001 0.003 0.002 0.003 0.001 0.32 0.002 0.5 0.5 0 0.625 0.006 0.021 0.001 0.2
Min 0.611 0.572 0.156 0.225 0.392 0.197 0.342 0.296 0.382 0.253 19.08 0.283 18 25 16 35 0.2 0.643 0.164 5.4
Max 0.655 0.61 0.183 0.251 0.444 0.226 0.42 0.375 0.461 0.289 27.34 0.343 21 28 16 37 0.233 0.754 0.169 10.8
n44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 5 5 5 5 44
P. wesselsi (Mbuluzi River)
Mean/Median∗0.628 0.599 0.161 0.231 0.416 0.21 0.367 0.354 0.452 0.264 22.83 0.293 19 26 16 8
SE/SIQ∗0.002 0.009 0.001 0.008 0.006 0.006 0.019 0.006 0.007 0.005 1.89 0.009 0.5 0.5 0 1.17
Min 0.624 0.576 0.158 0.21 0.406 0.199 0.329 0.342 0.436 0.251 18.63 0.27 18 24 16 4.6
Max 0.632 0.62 0.162 0.244 0.432 0.226 0.408 0.368 0.47 0.272 26.72 0.309 20 26 16 9.8
n444444 4 4 44 44 444 4
(Continued)
2200 B. Kramer et al.
Table 1. (Continued).
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL LSo/HL HL/SL HL/Na BD/SL nD nA SPc SLS OD/HL LPF/HL PPF/SL SL (cm)
Limpopo System
P. wesselsi (Blyde River)
Mean/Median∗0.620 0.589 0.167 0.24 0.423 0.209 0.361 0.316 0.395 0.264 22.97 0.291 20 27 16 35 0.218 0.699 0.165 8.1
SE/SIQ∗0.002 0.004 0.003 0.004 0.004 0.002 0.005 0.003 0.004 0.002 0.35 0.004 0.75 0.625 0 0.25 0.009 0.008 0.003 0.26
Min 0.617 0.578 0.161 0.225 0.408 0.204 0.349 0.307 0.385 0.259 21.93 0.28 18 26 16 34 0.19 0.669 0.158 7.6
Max 0.627 0.602 0.177 0.249 0.432 0.213 0.374 0.325 0.405 0.271 24.08 0.303 21 28 16 36 0.238 0.715 0.173 9.0
n5 55555 5 5 5 5 5555555 5 5 5
P. wesselsi (Groot Letaba River)
Mean/Median∗0.626 0.587 0.168 0.225 0.418 0.216 0.348 0.305 0.389 0.276 19.05 0.28 20 26 16 34 0.253 0.741 0.161 5.2
Min 0.626 0.585 0.165 0.221 0.405 0.214 0.347 0.303 0.382 0.27 18.93 0.279 20 25 16 34 0.245 0.729 0.159 4.8
Max 0.627 0.589 0.172 0.228 0.431 0.219 0.349 0.306 0.396 0.282 19.18 0.28 20 27 16 34 0.261 0.753 0.164 5.7
n2 22222 2 2 2 2 2222222 2 2 2
P. wesselsi (Nwanedzi River)
Mean/Median∗0.625 0.6 0.17 0.23 0.422 0.212 0.34 0.371 0.474 0.265 21.42 0.293 21 27 16 37 0.231 0.655 0.158 6.7
SE/SIQ∗0.003 0.005 0.002 0.003 0.004 0.002 0.013 0.005 0.006 0.003 0.78 0.006 0.625 1 0.25 0.5 0.004 0.015 0.005 0.55
Min 0.612 0.582 0.158 0.222 0.409 0.203 0.285 0.345 0.451 0.25 18.06 0.266 19 26 16 36 0.214 0.591 0.14 4.7
Max 0.642 0.622 0.179 0.248 0.453 0.22 0.399 0.389 0.496 0.276 24.29 0.313 22 28 18 38 0.252 0.731 0.19 9.8
n9 99999 9 9 9 9 9999999 9 9 9
P. wesselsi (Mokolo River)
Mean/Median∗0.627 0.587 0.167 0.226 0.417 0.208 0.311 0.361 0.494 0.269 21.2 0.258 20 26 16 38 0.276 0.727 0.154 5.1
SE/SIQ∗0.001 0.001 0.001 0.001 0.002 0.002 0.004 0.003 0.003 0.001 0.63 0.002 0.5 0.5 0 1 0.003 0.007 0.001 0.07
Min 0.606 0.564 0.155 0.209 0.393 0.188 0.252 0.302 0.445 0.257 14.87 0.236 19 25 16 36 0.221 0.633 0.136 3.7
Max 0.648 0.617 0.183 0.239 0.451 0.228 0.37 0.4 0.531 0.285 31.11 0.29 21 28 17 40 0.323 0.794 0.171 6
n48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 47 48 48
P. wesselsi (Lepalala River)
Mean/Median∗0.62 0.587 0.18 0.238 0.413 0.195 0.321 0.326 0.476 0.273 19.26 0.254 20.5 26.5 16 37.5 0.245 0.677 0.162 4.8
Min 0.618 0.562 0.177 0.231 0.413 0.188 0.285 0.316 0.45 0.262 18.32 0.243 20 26 16 37 0.206 0.632 0.156 4.8
Max 0.621 0.612 0.183 0.246 0.413 0.202 0.357 0.336 0.5 0.283 20.19 0.264 21 27 16 38 0.285 0.722 0.168 4.8
n2 22222 2 2 2 2 2222222 2 2 2
P. wesselsi (Pongola)
(n=1)
0.61 0.576 0.172 0.243 0.431 0.213 0.387 0.315 0.438 0.273 23.52 0.294 21 27 16 37 0.238 0.742 0.156 7.2
(Continued)
Journal of Natural History 2201
Table 1. (Continued).
P. squalostoma Syntypes (n =2)
BMNH 1920.5.26.1
(R1)
0.667 0.603 0.174 0.241 0.412 0.184 0.388 0.304 0.441 0.263 19.12 0.329 21 30 12 0.257 0.689 0.198 6.8
BMNH 1920.5.26.1
(R2)
0.673 0.599 0.160 0.250 0.406 0.192 0.346 0.259 0.409 0.280 27.60.335 18 30 12 0.223 0.638 0.208 6.6
P. longianalis sp.nov.(LufubuRiver)
Holotype SAIAB
76758
0.636 0.579 0.175 0.268 0.401 0.192 0.291 0.334 0.455 0.243 24.15 0.274 22 31 12 40 0.213 0.694 0.164 8.2
Mean/Median∗0.622 0.569 0.178 0.261 0.417 0.194 0.298 0.337 0.464 0.268 18.8 0.267 24 33 12 40 0.248 0.664 0.152 4.8
SE/SIQ∗0.002 0.002 0.001 0.001 0.001 0.001 0.002 0.002 0.003 0.001 0.234 0.002 0.5 0.5 0 0.5 0.002 0.005 0.001 0.1
Min 0.592 0.543 0.162 0.244 0.402 0.172 0.269 0.29 0.407 0.243 15.93 0.225 22 30 11 39 0.213 0.599 0.128 3.8
Max 0.64 0.59 0.191 0.278 0.444 0.215 0.331 0.362 0.496 0.279 24.15 0.293 26 35 12 42 0.273 0.789 0.169 8.2
n49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49 49
Luongo River
Mean/Median∗0.627 0.572 0.184 0.287 0.415 0.179 0.298 0.346 0.447 0.237 25.96 0.294 24 33.5 12 42 0.223 0.698 0.152 11
Min 0.623 0.568 0.18 0.28 0.413 0.176 0.29 0.344 0.439 0.236 25.74 0.294 24 33 12 42 0.214 0.695 0.151 10.2
Max 0.632 0.575 0.187 0.295 0.417 0.181 0.305 0.348 0.456 0.238 26.18 0.294 24 34 12 42 0.233 0.702 0.153 11.9
n2 22222 2 2 2 2 2 2 22222 2 2 2
Luapula River
Mean/Median∗0.627 0.569 0.176 0.27 0.431 0.193 0.295 0.357 0.462 0.25 20.88 0.298 23 32 12 39 0.255 0.665 0.15 7.0
SE/SIQ∗0.006 0.007 0.004 0.004 0.008 0.003 0.004 0.005 0.005 0.003 0.276 0.003 0.75 0.625 0 1 0.007 0.014 0.005 0.12
Min 0.611 0.546 0.168 0.257 0.408 0.184 0.284 0.344 0.448 0.244 20.13 0.288 21 31 12 38 0.235 0.631 0.138 6.6
Max 0.638 0.585 0.19 0.278 0.455 0.199 0.305 0.369 0.479 0.259 21.82 0.304 24 33 12 40 0.27 0.718 0.166 7.3
n5 55555 5 5 5 5 5 5 55555 5 5 5
P. frieli (Lake Bangweulu)
Mean/Median∗0.614 0.567 0.208 0.268 0.436 0.201 0.29 0.36 0.494 0.265 23.78 0.317 26 30 12 38 0.283 0.728 0.148 6.6
SE/SIQ∗0.001 0.005 0.004 0.004 0.004 0.006 0.012 0.007 0.009 0.001 0.843 0.008 0.5 0.25 0 0 0.008 0.01 0.002 0.42
Min 0.612 0.557 0.201 0.261 0.429 0.185 0.264 0.339 0.48 0.261 21.89 0.3 25 30 12 38 0.259 0.712 0.141 5.6
Max 0.618 0.579 0.216 0.281 0.446 0.209 0.322 0.372 0.519 0.267 25.69 0.338 27 31 12 38 0.297 0.755 0.152 7.4
n4 44444 4 4 4 4 4 4 44444 4 4 4
∗Median and SIQ, semi-interquartile range, for meristic data. Abbreviations of anatomical characters, see Material and Methods.
Fish from “Rovuma System”: SAIAB 73887(9), SAIAB 73802(10), SAIAB 73808(10), SAIAB 73894(6). “Upper Zambezi System”: “Katima”: SMF 28265(26), SAIAB 85909(3), 85911(2), 85916(1), 85917(2), 85918(3),
ZSM 38657(1); “East Lumwana River”: SAIAB 041208(5), and SAIAB 041025(1); “Wayama Lagoon”:SAIAB 72842(2); SAIAB 72670(1) pooled with Kama Lagoon, SAIAB 71792(1); “Kwando River”:ZSM 38658(1).
“Lake Malawi confluences”: SAIAB 039328(1), SAIAB 050065(2), SAIAB 050155(2). “Kafue System: Mwekera Stream”: SAIAB 042559(18), SAIAB 41224(4). “Lunga River”: SAIAB 040074(1). “Okavango River”:
“Guma Lagoon”, SAIAB 030046(1), SAIAB 186062(41), ZSM 38665(3); “Boro River”: SAIAB 67069(1), SAIAB 186057(10); “Gadikwe Lagoon”: SAIAB 19769(21); “Popa Rapids”: SAIAB 36841(24), SAIAB
36823(12); “Xakanixa Channel”: SAIAB 19705(11); “Xakanixa River”: SAIAB 21271(38). “Cunene River”: SAIAB 028120(3), SAIAB 78788(1), SAIAB 186053(1), SAIAB 79480(1), SAIAB 78790(1), ZSM 38659(1),
ZSM 38660(1). P. longianalis sp. nov. from “Lufubu River”: SAIAB 76758(1), SAIAB 186060(48); P. longianalis sp. nov. from “Luongo River”: SAIAB 76733(2); P. longianalis sp. nov. from “Luapula River”: SAIAB
76582(5); P. frieli from “Lake Bangweulu”: SAIAB 76825(3) and SAIAB 76859(1). P. wesselsi from “Sabie River”: SAIAB 05449(13), ZSM 28554 to 28566(13), SMF 28266(13), SAIAB 85922(5); “Mbuluzi River”:
SAIAB 067228(1), SAIAB 066355(1), SAIAB 068279(2). P. wesselsi from “LimpopoSystem”: Blyde River”: SAIAB 85923(5); Groot Letaba River”: SAIAB 85920(2); Nwanedzi River: SAIAB 58157(9). P. wesselsi from
“Pongola” River: SAIAB 85919(1).
2202 B. Kramer et al.
rays (19 in the Congo subspecies cf. 21 for P. catostoma and 24 for P. tanensis)”
and recognized it as a subspecies of P. catostoma.Petrocephalus stuhlmanni congi-
cus appears clearly more differentiated from P. catostoma than that by its extremely
short HL, short LD, CPD, long CPL and low number of SPc and nD. Furthermore,
its inferior mouth position even behind the centre of the eye (rare: Bigorne 2003:
158–159) and its reduced dorsal fin that originates far behind the origin of the anal
fin set it apart from P. catostoma. Species status is more appropriate than its present
subspecific designation.
Petrocephalus stuhlmanni Boulenger, 1909 (n=1) cannot be referred to P.
catostoma because of its low LD, LSo (also LSc), SPc, Na and high CPD and
BD (locality 2, Figure 1). The synonymization appears unjustified. Whitehead and
Greenwood’s (1959) synonymization with P. catostoma, that was not commented
upon, again seems mainly based on the similar number of dorsal fin rays: 19–20–21 in
P. stuhlmanni (n=9) vs 19–22–23 in P. catostoma (n=17).
Petrocephalus haullevillii Boulenger, 1912 (n=14; locality 24, Figure 1), a valid
subspecies, is differentiated from P. catostoma: its low SPc, HL and Na, and high LA
and its very inferior mouth also set it clearly apart from P. catostoma. [Whitehead and
Greenwood’s (1959) comment on the status of P. stuhlmanni congicus (cited above)
applies also for P. haullevillii.] The hypothesis of no difference (from Rovuma speci-
mens) among 13 anatomical characters (the ones listed on Table 2) was rejected by
multivariate analysis of variance (MANOVA;F
13,35 =47.76, P<0.0001 for all four test
variables, Wilks’ Lambda, Roy’s Greatest Root, Hotelling–Lawley Trace and Pillai
Trace). Subsequent univariate ANOVAs identified PAL, LD, LA, pD, CPL, HL, nD,
nA and SPc as sources of the difference (F1,47 ≥5.516, P≤0.0231). PCA on the same
set of 13 anatomical characters revealed complete separation of populations already
for principal components PC1 and PC2 (not shown). The synonymization appears
unjustified (locality 24, Figure 1).
With extreme values for PAL (highest), LSc and LD lowest among all nominal
species, the lowest possible number of 12 for SPc, and with CPD and Na in the
extreme range for the species, P. degeni Boulenger, 1906 (n=1) clearly does not rep-
resent P. catostoma, and the synonymization appears mistaken (locality 27, Figure 1).
Whitehead and Greenwood’s (1959) reasons for full synonymization of P. degeni with
P. catostoma (not a subspecies) are not explained expressis verbis, but the near identity
of their dorsal fin ray counts (19–21–22) with that for P. catostoma (19–22–23) may
have played a role.
Petrocephalus steindachneri Fowler, 1958 (n=3) is well differentiated from
P. catostoma by its high nD, LD, nA, PAL, BD, and low SPc, CPD and Na (local-
ity 25, Figure 1). The synonymization with P. catostoma cannot be supported.
Given Whitehead and Greenwood’s (1959) synonymization of all eastern African
Petrocephalus species with P. catostoma, including even two Congoan forms such as
P. haullevillii from near the Atlantic coast, Seegers (1996) logically united the East
African species P. steindachneri with P. catostoma.
Petrocephalus catostoma tanensis Whitehead and Greenwood, 1959 (n=9) was
recognized as a subspecies when discovered (locality 10, Figure 1). With very high val-
ues for LD, nD, pD, high BD, LA, CPD and low PDL and Na there is a marked degree
of differentiation present in the type material when compared with P. catostoma;
subspecies status appears inadequate and species status more appropriate. This is
supported by statistical comparisons using fresh samples from the Tana River (below).
Journal of Natural History 2203
Table 2. Comparison of anatomical characters in allopatric Petrocephalus species from southern and eastern Africa; multivariate analysis of variance
(MANOVA) followed by univariate ANOVAs.
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL HL/SL BD/SL nD nA SPc
MANOVA <10−4
ANOVA <10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4
Post tests
Rovuma, Tana <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, U Zambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, Okavango <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, P.wesselsi∗<0.01 0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Tana, U Zambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Tana, Okavango <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Tana, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Tana, P.wesselsi∗<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
U Zambezi, Okvgo <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
U Zambezi, Kafue <0.01
U Zambezi,
P.wesselsi∗
<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Okavango, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Okavango, P.wesselsi∗<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Kafue, P.wesselsi∗<0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu, Tana <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu,UZambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu, P. wesselsi∗<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu, Okavango <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Lufubu, Rovuma <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Pvalues in the body of the table not shown when P>0.01. For sample sizes, see footnote to Table 1.
Abbreviations of anatomical characters, see Material and methods. U Zambezi, Upper Zambezi. MANOVA F≥13,277 ≥33.51, depending on procedure;
MANOVA Pvalue: same for Wilks’ Lambda, Roy’s Greatest Root, Hotelling–Lawley Trace, and Pillai Trace tests. ANOVA F6,284 ≥16.55. Post tests
followed the Games/Howell procedure.
∗Petrocephalus wesselsi from Incomati River system; Kafue sample, n=22.
2204 B. Kramer et al.
Comparisons between allopatric populations referred to P. catostoma
Whereas firm conclusions for the relationships among most of the above nominal
species and P. catostoma can only be drawn with caution because of the general dearth
of material, multivariate statistics can be used to characterize the differences among
the allopatric populations we have sampled. The first three principal components
(PC1–PC3) on correlations accounted for almost two-thirds (65%; Appendix 1) of
the morphological variation in the data set. This shows that there was considerable
redundancy, and PCA was quite successful. Therefore, in order not to overestimate
differentiation when examining the hypothesisof no morphological difference between
fish from different origins by inferential statistics, a MANOVA was required (McGarigal
et al. 2000).
Included in both MANOVA and PCA were specimens from (1) the Rovuma region
(n=35, representing the type species in the present study), (2) Tana River (n=54),
(3) Upper Zambezi River (n=44), (4) Okavango River, Guma Lagoon (n=45),
(5) Kafue River (n=22), (6) the Luapula System, Lufubu River (n=49), and (7)
P. wesselsi from the Sabie River (n=44). Characters excluded from both PCA and
MANOVA were LSo (for its high degree of redundancy with LSc) and Na (because of
the danger of measurement error of this very small measure); thus, 13 anatomical char-
acters (the dependent variables) were compared by PCA and MANOVA/ANOVA, using
group (origin) as an independent variable in the latter two. (Additional samples and
characters were included in certain instances, as indicated where appropriate.)
The null hypothesis of no difference among the seven allopatric groups was clearly
rejected by MANOVA (P<0.0001, Table 2). Subsequent univariate ANOVAs identified
all 13 anatomical characters included in the analysis as contributing to the differentia-
tion (P<0.0001 for each). The PCA identified the main characters responsible for this
differentiation. PC1 captured 32.9% of the variation in the data set and was correlated
with positive and negative loadings (Appendix 1). Characters loading strongest on PC1
(“excellent”, in that order) were PAL, PDL and nA, the loading by CPD was “very
good”, the loadings by LA, SPc, nD and pD “good”, the ones by HL and BD “fair”,
and the one by LSc “poor”. PC1 therefore represented a gradient for “length and
depth of anterior trunk and depth of caudal peduncle vs length of rear section, espe-
cially of anal fin”, signifying that a long PAL and PDL and high CPD were associated
with a short anal fin, small number of rays and short pD (and vice versa). PC2 cap-
tured an additional 17.5% of the variation, representing a gradient for characteristics
of “caudal peduncle and peduncle-to-dorsalis length vs anal fin and anterior body
length“. “Excellent” was the loading by CPL, “good” those by pD and nA, “fair”
the ones by LA and PDL, and “poor” the one by SPc. PC3 captured an additional
14.7% of the variation, and was strongly loaded by LD and nD (“excellent”), but only
“poorly” by PAL, HL and SPc. PC3 seemed to represent a gradient for the dorsal fin,
and also head and trunk, being long when SPc was small (or vice versa). LSc was the
only character loading no more than “poorly” on any one of the first three (and even
four) PCs, but loaded strongly on PC5 (“excellent”). PC5 accounted for only 6.8%
of the variation, and LSc does not seem to contribute significantly to any dominant
morphological trait in the present data sample set.
Pairwise post-hoc tests showed significant differentiation between Rovuma spec-
imens and each one of the other populations in 7–11 characters (P<0.01,
Games/Howell procedure; Table 2); that is, none of the latter represents P.
catostoma. Furthermore, all possible pairwise comparisons among the six allopatric
Journal of Natural History 2205
populations yielded significant differences in 6–11 characters, except for the pair Upper
Zambezi–Kafue that differed significantly only in LSc (PCA had identified LSc as the
trait explaining least of the variation; see above). That is, with the exception of the
latter pair, all allopatric populations studied are well differentiated from each other.
We compared the small Lower Zambezi sample (n=11) to its neighbouring
populations only, to keep the number of pairwise comparisons manageable (and the
result meaningful). MANOVA rejected the null hypothesis of no difference among
13 anatomical characters when comparing Rovuma, Lower Zambezi, Upper Zambezi,
Kafue and Sabie (P. wesselsi) samples with one another. Subsequent ANOVAs showed
that each one of the 13 characters contributed significantly to this result. Lower
Zambezi samples differed from the standard, Rovuma samples, in seven characters,
from Kafue samples also in seven, from Upper Zambezi samples in six (among them
the three meristic characters), and from Sabie samples (P. wesselsi) in seven characters
(Table 3).
Plots of the principal component axes PC1 vs PC2 confirmed differentiation from
the Rovuma samples by a separation of clouds of points for samples from (1) the
Tana River (Figure 4A), the Upper Zambezi River (Figure 4B), the Okavango River
(Guma Lagoon; Figure 4C), the Lower Zambezi delta (Figure 4D), the Lufubu River
(Figure 4E), the Sabie (P. wesselsi, Figure 4F), the Cunene River (Figure 4G), and
the Boro River (Okavango delta; Figure 4H). (Where an individual point fell into
the region of the other sample, as in Figures 4A and 4F, tilting the graph slightly
by the third dimension PC3 revealed complete separation in separate spaces; not
shown for economy of presentation.) The samples from the Cunene River also proved
differentiated from another neighbouring population, those of the Upper Zambezi
(Figure 4I).
The two systems neighbouring the Okavango River, with sporadically intercon-
nected waterways, are the Upper Zambezi and the Cunene rivers. Samples of the latter
two were differentiated from the two samples of the Okavango River proper: Guma
and Popa. This is shown by non-overlapping ranges in PC1–PC3 coordinates, inde-
pendent of whether the Okavango River was represented by the Guma Lagoon sample
(Figure 5A, B) or the Popa Rapids sample (Figure 5C, D).
Despite the small Cunene sample size, MANOVA/ANOVA analysis (Table 4) con-
firmed differentiation from Rovuma specimens (in eight characters), from Lower
Zambezi specimens in seven characters, from Upper Zambezi specimens in four char-
acters, and from Okavango (Guma) specimens in three characters. Cunene samples
differed from Guma samples by their greater PAL and LSo, and smaller LA; addi-
tional differentiation is present in HL/Na (P<0.001, t=6.836), and SLS and OD
(not testable at present for insufficient sample size). In conclusion, we recognize mor-
phological differentiation on the species level for the samples from (1) the Tana River,
(2) the Lower Zambezi delta, (3) the Upper Zambezi River (including Kafue), (4) the
Okavango (Guma), (5) the Lufubu River, and (6) the Cunene River, and confirm
such differentiation for Sabie River samples (i.e. P. wesselsi). (A further species for
the Okavango, Boro River, is recognized below).
Comparisons of putative new species with nominal species for P. catostoma
Tana sample. The geographically closest nominal species, P. stuhlmanni (type, n=1),
showed values below the lowest of our large Tana sample for the characters LD,
2206 B. Kramer et al.
Table 3. Comparison of anatomical characters in allopatric Petrocephalus species focussing on the Lower Zambezi sample; multivariate analysis of
variance (MANOVA) followed by univariate ANOVAs.
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL HL/SL BD/SL nD nA SPc
MANOVA <10−4
ANOVA <10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4
Post tests
Rovuma, L Zambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, U Zambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
L Zambezi, Kafue <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
L Zambezi,
U Zambezi
<0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Kafue,UZambezi <0.01 <0.01
L Zambezi, P.wesselsi∗<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Pvalues in the body of the table not shown when P>0.01. For sample sizes, see footnote to Table 1.
Abbreviations of anatomical characters, see Material and methods. U Zambezi, Upper Zambezi; L Zambezi, Lower Zambezi. MANOVA F13,138 to F52,552
≥19.8, depending on procedure; MANOVA Pvalue: same for Wilks’ Lambda, Roy’s Greatest Root, Hotelling–Lawley Trace, and Pillai Trace tests.
ANOVA F4,147 ≥6.14. Post tests followed the Games/Howell procedure.
∗Petrocephalus wesselsi from Incomati River system; Kafue sample, n=18.
Journal of Natural History 2207
ACB
DEF
GI
GH
I
Figure 4(A–H). Principal component analysis for 13 anatomical characters of Petrocephalus
catostoma from Rovuma System (red triangles; n=35) compared (one by one) with vari-
ous allopatric Petrocephalus populations (blue squares): (A) with P. tanensis from Tana River
(n=52); (B) P. longicapitis sp. nov. from Upper Zambezi River (n=38); (C) P. okavangensis sp.
nov. from Guma Lagoon, Okavango (n=45); (D) P. p e t e r s i sp. nov. from Lower Zambezi River
(n=11); (E) P. longianalis sp.nov.fromLufubuRiver(n=49); (F) P. wesselsi from Incomati
System (n=44); (G) P. magnoculis sp. nov. from Cunene River (n=9); (H) P. magnitrunci sp.
nov. from Boro River (n=11). (I) compares P. longicapitis sp. nov. from Upper Zambezi River
(n=38, red triangles) with P. magnoculis sp. nov. from the Cunene River (n=9, blue squares).
Prin1, Prin2, for Principal Components 1 and 2.
pD, LSc, LSo, nD, and greater than (or greater than the 90th percentile) the high-
est Tana values for PDL, CPD and Na. Therefore, the Tana samples do not represent
P. stuhlmanni.
Petrocephalus degeni (type, n=1) of Lake Victoria, also in East Africa, is the
next closest nominal species. However, the measurements for its anatomical characters
were below the lowest of the Tana samples for LD, LA, pD, CPL, LSc, Na, nD and
2208 B. Kramer et al.
BA
DC
Figure 5. Principal component analysis for anatomy of Petrocephalus okavangensis sp. nov. (red
triangles) from (A, B) Guma Lagoon (n=45) and (C, D) Popa Rapids (n=36), compared
with (A, C) P. longicapitis sp. nov. from Upper Zambezi River (blue squares) and (B, D) P.
magnoculis sp. nov. (blue squares). Upper panels, analyses on 13 characters (see Table 3), lower
panels, analyses on 17 characters (see Table 1, with HL/Na and LSc/HL excluded). Prin1–Prin3,
for Principal Components 1–3.
greater than the Tana sample’s upper range for PAL and PDL (PDL, 90th percentile).
Therefore, the Tana samples do not represent P. degeni.
From geography an unlikely species to associate with the Tana samples is P. s t e i n -
dachneri (types, n=3) because the origin of P. stuhlmanni is between the two (all three
inhabiting independent rivers draining into the Indian Ocean, with 600 km between
the mouths of the Tana River and the Rufiji River inhabited by P. steindachneri).
Anatomical measures below the 90th percentile of our large Tana samples’ ranges were
pD and CPD, and above, PAL (all characters loading strongest on PC1). Given these
differences, the Tana samples cannot be conspecific with P. steindachneri.
Journal of Natural History 2209
Table 4. Comparison of anatomical characters in allopatric Petrocephalus species focusing on the Cunene River sample; multivariate analysis of variance
(MANOVA) followed by univariate ANOVAs.
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSc/HL HL/SL BD/SL nD nA SPc
MANOVA <10−4
ANOVA <10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4<10−4
Post tests
Rovuma, L Zambezi <0.01 <0.01 <0.05 <0.01 <0.01 <0.01 <0.01 <0.01
Cunene, L Zambezi <0.05 <0.05 <0.05 <0.01 <0.01 <0.01 <0.01
L Zambezi, Okavango <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
L Zambezi,
U Zambezi
<0.01 <0.01 <0.05 <0.01 <0.01 <0.01 <0.01
Rovuma, Cunene <0.01 <0.05 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, Okavango <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Rovuma, U Zambezi <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Cunene, Okavango <0.01 <0.01 <0.01
Cunene, U Zambezi <0.05 <0.05 <0.01 <0.01
Okavango, U
Zambezi
<0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Pvalues in the body of the table not shown when P>0.05. For sample sizes, see footnote to Table 1.
Abbreviations of anatomical characters, see Material and methods. U Zambezi, Upper Zambezi; L Zambezi, Lower Zambezi. MANOVA F≥13,124 ≥13.4;
MANOVA Pvalue: same for Wilks’ Lambda, Roy’s Greatest Root, Hotelling–Lawley Trace, and Pillai Trace tests. ANOVA F4,133 ≥9.33. Post tests followed
the Games/Howell procedure.
2210 B. Kramer et al.
The remaining two nominal species, P. stuhlmanni congicus (types, n=2) and
P. haullevillii (types, n=14), inhabit distant basins draining into the Atlantic Ocean,
with no connections to East Africa. For the closer of the two species, P. s. congicus from
the Congo River basin, values lower than the Tana samples’ lowest ranges were found
for HL, Na and nD, with in addition both LD and CPD below the 90th Tana per-
centile. Regarding P. haullevillii from near the Atlantic coast and north of the Congo
River, the hypothesis of no anatomical difference from Tana samples (n=53) was
rejected by MANOVA (F13,52 =47.24, P<0.0001 for all four multivariate test proce-
dures). Subsequent univariate ANOVAs identified PAL, LD, LA, pD, CPD, HL, nD,
nA and SPc as sources of the difference (F1,64 ≥5.259, P≤0.0251). The Tana samples
represent neither of these two nominal species, nor any other, except, of course, P. c.
tanensis Whitehead and Greenwood, 1959, which we elevate to species rank, P. tanensis
(Whitehead and Greenwood, 1959), in Systematics (below).
Upper Zambezi sample. The origins of the nominal species are all far off the origin
of the present Upper Zambezi sample, the closest geographical association probably
being with the Congo basin. Anatomical character measures for P. s. congicus were
above those for the Upper Zambezi sample for CPL, and below for LD, CPD, HL,
nD and PAL (PAL, 90th percentile). Therefore, the Upper Zambezi sample does not
represent P. s. congicus.
Nor are the Upper Zambezi samples representing P. haullevillii. The hypothesis of
no difference from Upper Zambezi samples when considering all anatomical charac-
ters together (as in Table 2) was rejected by MANOVA (F12,45 =45.16; P<0.0001 for
all four multivariate test procedures). Subsequent univariate ANOVAs identified all
anatomical characters except LSc and SPc as sources of the difference (F1,56 ≥7.741,
P≤0.0073).
The remaining nominal species are all in distant and isolated basins in East Africa.
The P. steindachneri sample of three specimens overlaps with Upper Zambezi samples
in most characters, except for its very low CPD, which is even below outliers of the
Upper Zambezi sample, and among the characters loading strongest on PC1. Given
its provenance even further from the Upper Zambezi than P. catostoma, we conclude
that P. steindachneri is not conspecific with the Upper Zambezi samples.
Petrocephalus stuhlmanni differs from the Upper Zambezi sample by low LD and
LSo values, and high CPL values beyond the range of Upper Zambezi sample outliers.
In addition, its values for pD, LSc, nD and nA are below and CPD above the 90th
Upper Zambezi sample percentile.
Petrocephalus degeni from tropical Lake Victoria cannot be associated with the
Upper Zambezi sample because of its low values for LD, LSc, nD and high PAL (all
beyond the range of outliers of the Upper Zambezi sample), with in addition an HL
below and a pD measure above the 90th Upper Zambezi percentile. Its short dorsal fin
originates above (and not behind) its anal fin.
We conclude that the Upper Zambezi sample cannot be conspecific with any of the
nominal species previously referred to P. catostoma, and recognize it as P. longicapitis
sp. nov. in Systematics (below).
Guma Lagoon sample (Okavango).Among the morphological measurements of P. s.
congicus, LA and nD are below the outlier range of the Guma sample, with in addition
HL and nA below the lower 90th percentile and CPL greater than the upper 90th
percentile. P. s. congicus is quite clearly not the species we find in Guma Lagoon.
Journal of Natural History 2211
Regarding P. haullevillii, the null hypothesis of no difference from Guma samples
in anatomy (as represented by 13 characters) is rejected by MANOVA (F12,46 =21.13,
P<0.0001 for all four test procedures). Subsequent univariate ANOVAs identified
CPD, CPL, LA, PAL, HL, nA and nD as sources of the difference (F1,57 ≥4.123,
P≤0.047), and the Guma sample is considered a different species.
Petrocephalus steindachneri features an LA shorter than the lower 90th percentile
of the Guma sample, and PAL and HL longer than the upper 90th percentile.
Petrocephalus steindachneri is not the species we find in the Guma Lagoon.
The same holds for P. stuhlmanni. LA, pD, LSo were smaller than, and CPD and
HL greater than the Okavango sample range. With in addition LD, nD and Na smaller
than or equal to, and PDL and PAL greater than the lower or upper 90th percentiles,
respectively. The Guma sample is well differentiated from P. stuhlmanni.
Petrocephalus degeni is not a possibility for the Guma sample. Its LA, LSc and nD
were smaller, and its PAL and CPD greater than the most extreme values observed in
the Guma sample. In addition, LD and nA were smaller than, and HL greater than
the lower and upper 90th percentile, respectively. The origins of the dorsal fin differ:
above anal fin in P. degeni, more posterior in the Guma sample of specimens.
We conclude that the Guma Lagoon sample has no close affinities with any of the
nominal species referred to P. catostoma, and we recognize P. okavangenis sp.nov.in
Systematics (below).
Boro River sample (Okavango).Samples from the Boro River in the Okavango delta
were clearly differentiated from all other Okavango samples. This discovery raises the
question of differentiation from nominal species also here. Petrocephalus s. congicus
did not overlap with the Boro River sample’s greater values for PAL, LSo, HL and
BD, whereas for CPL its range of values was below that of P. s. congicus. Therefore, P.
s. congicus is not the species found in the Boro River.
Petrocephalus haullevillii and the Boro River sample were drawn from clearly dif-
ferentiated populations, as shown by MANOVA (F13,11 =26.58, P<0.0001). This result
was brought about by significantly different distributions for PDL, PAL, CPL, LSo,
HL, BD, nD and nA (ANOVA,F
1,23 ≥8.56, P <0.01).
Comparing the Boro River sample with P. steindachneri yielded four non-
overlapping characters (LSc, LSo and BD smaller, HL greater in P. steindachneri),
in addition to further marked differences (nD, CPD). We therefore do not refer the
Boro River sample to P. steindachneri.
The Boro River sample is quite clearly not referable to P. stuhlmanni whose LD,
LSc, LSo and BD are all smaller, and CPD and HL greater than the most extreme
values found in the Boro River sample.
The comparison of the Boro River sample with P. degeni shows clear differentiation
in the characters (1) LD, LA, LSc, LSo and BD, and (2) PAL and HL, the distributions
of which do not overlap. The first group of characters are all smaller, the second greater
in P. degeni. The position of the dorsal fin origin with respect to that of the anal fin
differs between the two species.
We conclude there are no close affinities with any of the nominal species, and we
therefore recognize P. magnitrunci sp. nov. for the Boro River sample in Systematics.
Lower Zambezi sample. The most relevant nominal species for comparison with type-
locality P. catostoma are those from rivers also discharging into the Indian Ocean.
The Lower Zambezi sample proved to be anatomically well differentiated from
2212 B. Kramer et al.
P. steindachneri, the geographically closest P. catostoma synonym. Petrocephalus stein-
dachneri’s CPD, LSc and LSo were lower, and nD and nA higher than the most
extreme values observed in the Lower Zambezi sample (that also differed by a median
SPc of 16 vs 12 as observed in P. steindachneri). In addition, P. steindachneri’s LD was
greater than the 90th LD percentile of the Lower Zambezi sample.
Similarly, the Lower Zambezi sample cannot be associated with P. stuhlmanni.The
P. stuhlmanni’s LD, LSc and LSo were lower, and its CPL, nD and nA were higher than
the most extreme observed in the Lower Zambezi sample. An SPc of a median count
of 16 in the latter also contrasts with 12 in the former. Beyond the 90th percentiles of
the Lower Zambezi sample were PAL and Na (low end of the distribution), LA and
CPD (high end) in P. stuhlmanni.
The Lower Zambezi sample is clearly differentiated from P. catostoma tanensis
(represented by a sample of n=54). The hypothesis of no anatomical overall differ-
ence is rejected by MANOVA (F13,49 =41.285, P<0.0001 for all four test procedures).
Subsequent univariate ANOVAs identified PDL, PAL, LD, LA, pD, CPL, LSc, HL,
nD, nA and SPc as sources of the difference (F1,61 ≥15.67, P≤0.0002).
Petrocephalus degeni of Lake Victoria is also not the species we find in the Lower
Zambezi. With LSc, LSo and HL below, and PAL, pD and nA above the range
observed in the Lower Zambezi sample, and an SPc of 12 vs a median 16, P. degeni
and the Lower Zambezi species have little in common. In addition, P. degeni’s LD was
below the 90th LD percentile of the Lower Zambezi sample.
Petrocephalus stuhlmanni congicus is clearly differentiated from the Lower Zambezi
sample by anatomical character measures more extreme than the range of the latter in:
PAL, CPD, LSc, LSo, HL (lower), and CPL and nA (higher); in addition, by SPc (12 vs
a median 16 in the Lower Zambezi sample).
The hypothesis of no anatomical differences between P. haullevillii and the Lower
Zambezi sample is rejected by MANOVA (F13,11 =45, P<0.0001 for all four
multivariate test procedures; for the characters included, see Table 2). As shown by
subsequent univariate ANOVAs, characters significantly contributing to this difference
were PDL, PAL, LA, pD, CPL, CPD, LSc, HL, BD, nA and SPc (F1,23 ≥5.732, P≤
0.0252).
We conclude that the Lower Zambezi sample is not represented by any of the
nominal species of P. catostoma, and recognize P. p e t e rs i sp. nov. in Systematics
(below).
Lufubu sample. Samples from the Lufubu River are clearly differentiated from P. s.
congicus even though both occur in the same river system, the Congo. The two samples
show non-overlapping ranges in LA, CPL, LSo, HL, nD and nA.
A similarly clear differentiation was found for P. degeni: the Lufubu samples’
ranges were below or above the values for this species in PAL, LD, LA, LSc, nD
and nA.
Petrocephalus steindachneri from East Africa cannot be associated with the Lufubu
samples for their non-overlapping ranges in PAL and BD, characters loading signifi-
cantly on PC1. The 90th percentile ranges for LA, LSo and nA did not overlap with
those for P. steindachneri.
Petrocephalus stuhlmanni. There was no overlap with Lufubu samples in any
character except in SPc, that is, no affinity whatsoever.
Journal of Natural History 2213
Petrocephalus haullevillii. The hypothesis of no morphological difference between
P. haullevillii (n=14) and the Lufubu sample (n=49) is rejected by MANOVA (F13,49 =
54.45; P<0.0001). Significant differences were apparent in LD, CPL, CPD, HL, nD,
LA and LSc (ANOVA,F
1,61 ≥8.85; P<0.01); that is, the Lufubu sample is well
differentiated.
The Lufubu sample is also clearly differentiated from P. c. tanensis (n=54). Among
the 13 anatomical characters included in MANOVA/ANOVAs, 11 proved to be sig-
nificantly different (P<0.01; Table 2). We conclude the Lufubu sample cannot be
associated with any of the nominal species of P. catostoma.
However, there is another hypothesis to test. Petrocephalus squalostoma
(Boulenger, 1915) resembles churchills and was recorded from a small tributary of
Lake Moero, that is, from a region further downstream the Luapula River system com-
pared with the Lufubu (Figure 1, no. 39). Therefore, we compared the Lufubu sample
with two syntypes of P. squalostoma.
Petrocephalus squalostoma (n=2) features the most extreme PDL/SL and PPF/SL
values of all the samples studied in the present paper, with no overlap with Lufubu
samples, including outliers (n=49; Table 1). With no overlap, PAL/SL, CPD/CPL
and BD/SL were all greater, and nD and nA smaller in the Types than in the Lufubu
sample (nA: a single outlier of the Lufubu sample also had only 30 rays). LA/SL was
longer in Lufubu samples than in Types that did not reach the lower 90th percentile
of Lufubu samples. LSo/HL values in Types were below the lower 90th percentile
of the Lufubu sample. Given these differences, we recognize the Lufubu samples as
representing the new species P. longianalis sp. nov. in Systematics (below).
Cunene sample. The closest affinities of samples from the isolated Cunene River may
be expected with the Guma Lagoon or Boro samples (however, Figures 5B,D, 8F con-
firmed differentiation), and with the Upper Zambezi sample (Figure 4I, also showing
differentiation), whereas all other origins are so far away and unconnected that close
affinities are unlikely (e.g. Figure 4G).
Comparison of other samples with nominal or putative new species
Petrocephalus catostoma of the type region. The specimens from Lake Malawi conflu-
ences (SAIAB 050155, 050065, 039328; localities 16–18, Figure 1) do not present clear
differentiation in any anatomical character (Table 1), and are regarded as represent-
ing P. catostoma. However, SAIAB 039264, a specimen from Lake Chiuta (Malawi,
locality 19), appears not to represent P. catostoma for its higher pD and SPc, smaller
Na, as well as measures more extreme than the 90th percentile range for LA and HL.
Considering the close association of Lake Chiuta with the Rovuma system through the
Lugenda River, this degree of differentiation should be confirmed by more material.
Lower Zambezi samples. Specimens from Mulela (SAIAB 055875, n=4) and Lower
Zambezi (SAIAB 060846, n=11) were mostly similar to each other, except for HL
and Na being smaller, and CPD (90th percentile of Lower Zambezi) greater in the
Mulela than Lower Zambezi sample. Mulela specimens appear rather well differenti-
ated from P. catostoma (PAL and Na more extreme than the range of P. catostoma,
CPD and BD equal to an outlier, and several characters as extreme or more than the
2214 B. Kramer et al.
90th percentile range: LSc, LSo, CPL, Spc). Therefore, Mulela specimens are regarded
as representing the Lower Zambezi species (P. p e t e r s i sp. nov.). The specimen from
Lake Chiuta (SAIAB 039264) is well differentiated from both Lower Zambezi and
Mulela specimens by nine and 12 anatomical measures at least at the 90th percentile
level, respectively.
Lake Rukwa specimen. The specimen from Lake Rukwa (SAIAB 059515, locality 11,
Figure 1) is well differentiated from P. catostoma (as well as from any other nominal
species in a huge perimeter: P. steindachneri,P. stuhlmanni, the Tana sample, P. s. con-
gicus,P. degeni) in all anatomical characters listed on Table 1 but BD, CPD and CPL.
The Rukwa specimen’s anatomical measurements or counts are also below or above
the ranges of any one of the Luapula System samples for 11 characters; that is, there
are no affinities with a new species (designated in Systematics). If confirmed by more
material, a new species is suggested for Lake Rukwa.
Upper Zambezi sample. Kafue system churchills (locality 13, SAIAB 042559, 18 spec-
imens; and locality 15, SAIAB 040074, one specimen) appear closely associated with
Upper Zambezi churchills (only a single difference identified by MANOVA/ANOVA
analysis, in LSc, see Table 2), despite geography from which one would expect affini-
ties with the sample from the Lower rather than the Upper Zambezi (the Kafue River
joins the Zambezi River below Victoria Falls and Lake Kariba). The specimen from
the Kwando River differs from the Upper Zambezi specimens, particularly so from the
Guma Lagoon (Okavango) specimens. Compared with both species’ anatomical char-
acter ranges, the Kwando specimen was above range for PAL and CPD, below range
for LSo; compared with the Guma sample, in addition below range for LA and pD.
Compared with the 90th percentile ranges of both species, the Kwando specimen was
also more extreme in CPL, LSc and Na. When exclusively compared with Guma sam-
ples, additional differences at or beyond the 90th percentile ranges occurred for nA,
BD, LD, whereas when compared with Upper Zambezi samples, only HL was added
to the list. The Kwando churchill therefore is closer to the Upper Zambezi churchill
but may, with more material, prove to be differentiated.
Okavango samples. Churchills from six locations on the Okavango or its delta
(Figure 6) are clearly differentiated among each other, as confirmed by MANOVA on
13 anatomical characters (Tables 4, 5). Subsequent ANOVAs eliminated only SPc as a
source for this differentiation. The Boro River population from deep down the delta
appeared strongly differentiated from all other, more northern or peripheral popula-
tions by significantly different mensural or meristic characters (compared to Gadikwe
fish, six characters; Guma, five characters; Popa, seven characters; Xakanixa River,
seven characters; all P<0.01, Table 5). Some tendency for segregation in terms of
PCA coordinates was obvious in all pairwise comparisons with Guma specimens, but
only the Boro population approached separation from the Guma population, despite
its low sample size and only 13 characters analysed (Figure 7E). When replacing the
Guma by the more northern (more distant) Popa population from the Okavango River,
the separation from Boro in 13 characters was complete (Figure 7F).
The characters in Boro River fish that differed most consistently from those of
the other Okavango populations were PDL, PAL, BD, nD and SLS. A PCA-based
comparison of Boro fish with all other Okavango samples in pairwise comparisons
Journal of Natural History 2215
22° E 23°
18° S
19°
20°
24°
22° E 23° 24°
18° S
19°
20°
9
34
32
33
31
8
35
100 km
Figure 6. As Figure 1, but area of Petrocephalus okavangensis sp. nov. and location from
where samples were taken shown at better resolution. (9) Nguma (Guma) Lagoon. (31) Boro
River. (32) Gadikwe Lagoon. (33) Xakanixa River. (34) Xakanixa Channel. (35) Popa Rapids.
(8) unique specimen from Kwando River.
confirmed strong differentiation and complete separation (Figure 8A–D), including
the two neighbouring populations from further east (Upper Zambezi, Figure 8E) and
west (Cunene River, Figure 8F). Also remarkable was the differentiation between
fish from the Popa Rapids and Guma Lagoon in five characters, which we regard
as infrasubspecific because of the considerable overlap in terms of PCA coordinates
(Figure 7A, marked degree of overlap confirmed in three dimensions).
Whereas the Popa Rapids represent the Okavango River proper in our data set,
Guma Lagoon in its southeast connects in addition to all major waterways of the delta,
including the partly seasonal channel to the Upper Zambezi system (Magwegqana).
In spite of much similarity with the Guma population, the Kwando specimen dif-
fers markedly from it in CPD, LSc and LSo. More specimens are needed before any
conclusions can be drawn.
The differentiation presented here is based on a vastly incomplete picture of
the variability within the Okavango and its delta. The six populations represent the
Okavango River, the panhandle of its delta, and points on the northeastern edge of
the delta, whereas central, western and southern regions are poorly represented or not
at all. Furthermore, EOD and DNA samples are largely lacking. We still feel that the
Boro population surpasses subspecific variation and recognize P. magnitrunci sp. nov.
in Systematics.
Tana sample. No neighbouring samples present in our material.
2216 B. Kramer et al.
Table 5. Comparison of anatomical characters in different populations of Petrocephalus species from the Okavango River or delta: multivariate analysis
of variance (MANOVA) followed by univariate ANOVAs.
PDL/SL PAL/SL LD/SL LA/SL pD/SL CPL/SL CPD/CPL LSo/HL HL/SL BD/SL nD nA SPc
MANOVA <0.0001
ANOVA <10−4<10−4<10−40.011 0.0002 <10−40.0038 <10−4<10−4<10−4<10−40.0102
Post tests
Boro, Gadikwe <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Boro, Guma <0.01 <0.01 <0.01 <0.01 <0.01
Boro, Popa <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Boro, Xakanixa
Channel
<0.01
Boro, Xakanixa River <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01
Gadikwe, Guma <0.01 <0.01
Gadikwe, Popa
Gadikwe, Xakanixa
Channel
Gadikwe, Xakanixa
River
Guma, Popa <0.01 <0.01 <0.01 <0.01 <0.01
Guma, Xakanixa
Channel
Guma, Xakanixa
River
<0.01 <0.01
Popa, Xakanixa
Channel
<0.01 <0.01 <0.01
Popa, Xakanixa River <0.01
Xakanixa Ch,
Xakanixa River
<0.01
Pvalues in the body of the table not shown when P>0.01. For sample sizes, see footnote of Table 1.
Abbreviations of anatomical characters, see Material and methods. MANOVA F13,148 to F65,740 ≥7.357, depending on procedure; MANOVA Pvalue: same
for Wilks’ Lambda, Roy’s Greatest Root, Hotelling–Lawley Trace, and Pillai Trace tests. ANOVA F5,156 =3.128 for P=0.0102. Post tests followed the
Games/Howell procedure.
Journal of Natural History 2217
ABC
FED
Figure 7(A–E). Principal component analysis on the 13 anatomical characters of Table 3, for six
Okavango populations, focusing on the Guma Lagoon sample (Petrocephalus okavangensis sp.
nov., red triangles). (A) Guma/Popa Falls (blue squares). (B) Guma/Gadikwe (blue squares).
(C) Guma/Xakanixa Channel (blue squares). (D) Guma/Xakanixa River (blue squares). (E)
Guma/Boro River (blue squares). (F) Popa Falls (red triangles)/Boro River (blue squares).
Lufubu sample. In addition to SAIAB 76758 from the Lufubu River, there were three
small samples from (1) the close-by Luongo River (SAIAB 76733), (2) from a location
considerably further upstream the Luapula River (SAIAB 76582), and (3) from Lake
Bangweulu (SAIAB 76825 and 76859), all forming part of the Luapula drainage in
the north of Zambia. Specimens from the closest location, the Luongo River, differed
from the Lufubu sample in the characters LA, HL, HL/Na and BD (no overlap),
and CPL (90th percentile), whereas Luapula River samples differed by BD at the
90th percentile level. Lake Bangweulu specimens were more markedly different by
their LD and BD presenting no overlap, and pD, HL/Na, nD and nA differing at
the 90th percentile level. The latter sample represents P. frieli Lavoué, 2012 (Lavoué
Forthcoming 2012), whereas the former samples seem to demonstrate a rather large
degree of intraspecific variation within the new species P. longianalis sp. nov. designated
in Systematics (below)
Petrocephalus wesselsi. There is morphological differentiation among some of the eight
P. wesselsi samples from different origins within South Africa and Swaziland (Table 1;
geography, Figure 9). As demonstrated by PCA on correlations for anatomical charac-
ters of four of these samples [from Sabie River (n=45), Blyde River (n=5), Nwanedzi
River (n=9) and Mokolo River (n=48)], PC1 captured 36.4% of the variation in the
data set (33.6% when P. longicapitis sp. nov. from Namibia was included as shown
2218 B. Kramer et al.
ABC
FED
Figure 8. Principal component analysis on 17 anatomical characters (see Table 1), for six
Okavango samples, focusing on that of the Boro River (Petrocephalus magnitrunci sp. nov.,
red triangles). (A) Boro/Xakanixa River (blue squares); (B) Boro/Gadikwe Lagoon (blue
squares); (C) Boro/Xakanixa Channel (blue squares); (D) Boro/Popa Rapids (blue squares);
(E) Boro/Upper Zambezi River (blue squares); (F) Boro/Cunene River (blue squares). The
17 characters included in the analysis were: PDL/SL, PAL/SL, LD/SL, LA/SL, pD/SL,
CPL/SL, CPD/CPL, Lso/HL, HL/SL, BD/SL, nD, nA, SPc, SLS, OD/HL, LPF/HL,
PPF/SL. Characters of Table 1 that were excluded: HL/Na, LSc/HL.
in Figure 10A). Positively loading on PC1 were LD, nD, nA (all “excellent”), CPD
(“good”), LA, BD, HL (“fair”) and PDL (“poor”). Negatively loading on PC1 were
SPc (“excellent”), LSc (“very good”) and CPL (“good”). PC1 seems to represent a
gradient mainly for characteristics of the unpaired fins vs characteristics of the caudal
peduncle, such as its length and SPc. PC2 captured an additional 15.7% of the varia-
tion, and characters loading positively on PC2 were PDL (“excellent”), PAL (“good”),
BD (“fair”) and CPD (“poor”). Negatively loading on PC2 were LD, pD, nD, nA (all
“poor”). PC2 therefore is mainly a gradient for trunk length and height vs unpaired
fin development. PC3 captured a further 14.7% of the variation, however, none of the
characters loading on PC3 did so better than “good” (LA, pD, BD), “fair” (CPD)
or even “poor” (CPL, PDL, the latter being the only negatively loading character).
The characters loading on PC1 had by far the greatest weight and separated the
populations best.
The differentiation of populations seems to follow a north–south transect, with
almost total separation of the two populations farthest apart (Sabie vs Mokolo River;
Figure 10A). On the graph they are connected by the two populations that are also
geographically intermediate (Blyde River, Nwanedzi River). Discriminant Analysis on
the same data set confirms a very marked differentiation between Sabie and Mokolo,
Journal of Natural History 2219
22° S
28°
26°
24°
26°
22° S
26°
250 km
Maputo
3
22
5
4
6
Maputo
Incomati
Komati
Limpopo
Changane
Olifants
Vaal
Limpopo
Pongola
30° 32° 34°
26°
24°
Indian Ocean
Sabie
Crocodile
Bloedrivier
Kosimeer
Olifants
Letaba
Groot Letaba
Blyde
30° 34° E32°
Mbuluzi
Mokolo
Lepalala
Johannesburg
Gaborone
Limpopo
36
37
28°
Nwanedzi
38
Figure 9. As Figure 1, but area of Petrocephalus wesselsi is shown at better resolution.
(3) Incomati System: Sabie River. (4) Groot Letaba River (5) Blyde River, (36) Lepalala
River, (37) Mokolo River, (38) Nwanedzi River, all Limpopo System. (6) Pongola
River. (22) Swaziland: Mbuluzi River.
with Nwanedzi specimens again at an intermediate position (Figure 10B). However,
compared with the differentiation from P. longicapitis sp. nov., the differentiation
within South Africa is perhaps best regarded as indicating intraspecific variation,
possibly in the form of a geographical cline).
Electric organ discharge comparisons
Allopatric churchill species
For P. tanensis,P. longicapitis sp. nov., P. wesselsi,P. okavangensis sp. nov. and P.
magnoculis sp. nov. we confirm a degree of similarity of electric organ discharge
waveforms among each other. In all species, a head-positive P1 phase is followed, in
turn, by a strong head-negative N phase and a weaker positive P2 phase (Figure 11).
Unsupported by references and without further comment, such triphasic EODs with
head-negative main phase have been termed “atypical” for Petrocephalus by Lavoué
et al. (2000), but Lavoué et al. (2004, 2008) show Petrocephalus EODs from more trop-
ical species very similar to ours (acknowledged by Lavoué et al. 2004). It seems that
the more tropical the sampling origin, the briefer the pulse. It is only by quantitative
analysis of samples that a considerable degree of differentiation is revealed.
For reasons of sample size, the following statistical comparisons focus on
P. tanensis,P. longicapitis sp. nov. and P. wesselsi (n≥22). We considered only adult
2220 B. Kramer et al.
Petrocephalus Anat: alle 15 Variablen
AB
C
D
Component 1 (64.9 %)
Component 1 (33.6 %)
Figure 10. Differentiation within Petrocephalus wesselsi of different South African origins, stud-
ied using Principal Component (PCA) and Discriminant (DA) Analyses, as compared with
another species (P. longicapitis sp. nov., Namibia, Upper Zambezi). Circles in DA, 95% con-
fidence circles to contain true mean of group. (A) PCA and (B) DA on correlations among
anatomical characters. (C) PCA and (D) DA same as (A) and (B), respectively, but for anal-
yses on characters of Electric Organ Discharges (EODs). Green M symbols, specimens from
Mokolo (Mogol) River (n=48; 43, i.e. 48 for anatomy and 43 for EOD); blue-green N sym-
bols, Nwanedzi River (n=9; 0); orange B symbols, Blyde River (n=5; 5); red S symbols, Sabie
River (n=44; 39); blue Z symbols, P. longicapitis sp. nov. from the Upper Zambezi (type local-
ity Katima Mulilo; n=38; 42); bluer shade of blue-green coloured U symbols, specimens from
Mbuluzi River, Swaziland (n=4; 0). Excluded from DA but shown on DA graphs: lilac G sym-
bols, Groot Letaba River (n=2; 2); sand-coloured L symbols, Lepalala River (n=2;1); redder
shade of lilac P symbol, Pongola River (n=1; 1). The 15 anatomical characters included in the
anatomical analyses were: PDL/SL, PAL/SL, LD/SL, LA/SL, pD/SL, CPL/SL, CPD/CPL,
LSc/HL, LSo/HL, HL/Na, HL/SL, BD/SL, nD, nA, SPc. The characters used in the EOD
analyses were: Namp, P2amp, P1dur, Ndur, P2dur, P1Nsep, P1P2sep, NP2sep, P1area, Narea,
P2 area.
fish (that is, SL ≥5.2 cm, or 40% of the maximum species size; see Kramer 1997 for
discussion). Except for P1dur, P. tanensis EODs differed between the sexes for all char-
acters studied [P≤0.0223, analysis of covariance (ANCOVA); Table 6]. An ANCOVA
with sex as a factor and SL as a covariate was chosen to control for any dependencies of
EOD characters on size (SL). Such dependencies were found in all samples of sufficient
size at least for P2amp (Table 6, underlined; for details, see Appendix 2).
Journal of Natural History 2221
Figure 11. Oscilloscope traces of Electric Organ Discharges (EODs) of members of southern
and eastern African Petrocephalus species. (A) P. longicapitis sp.nov.(B)P. tanensis,(C)P. w e s -
selsi,(D)P. okavangensis sp. nov. Ordinate, voltage, with head-positivity upwards from baseline: