T here are not many late Cenozoic mammalian localities in north-western Africa. Our knowledge of many of these localities owes much to Camille Arambourg who conducted numerous excavations in this area (Arambourg 1970, 1979). More recently, important new sites have been excavated in Morocco under a Franco-Moroccan program, the ‘Programme Casablanca’, of the ‘Institut National des Sciences de l’Archéologie et du Patrimoine’ (INSAP), Rabat, and other excavations are under way in some other localities. A few of these sites have yielded significant collections of Carnivora, but detailed studies are extremely rare. Currently, a general picture is badly lacking, as the recent reviews (Olive 2006; Geraads 2008) do not go deep into details. The purpose of the present paper is to provide a detailed account of the whole fossil record of the family Canidae in this area, from the Miocene–Pliocene to the late Pleistocene and to place these North African representatives in the evolutionary framework of the family.
Abstract: The fossil record of the Canidae in North-western Africa begins near the Miocene–Pliocene boundary with a form close to Nyctereutes, a genus best known in the late Pliocene of Ahl al Oughlam. This site yields two other canids. Vulpes hassani sp. nov. is a small fox, probably ancestral to the modern V. rueppelli, recorded from the Middle Pleistocene onwards. Lupulella paralius sp. nov. is a primitive jackal that probably belongs to the clade of modern African jackals. In the middle Pleistocene, the most common canid is Lupulella mohibi sp. nov., remarkable by its Nyctereutes-like dentition and primitive skull-features. These are all endemic forms, but V. vulpes and C. aureus, of northern origin, appear in the course of the middle Pleistocene. Lycaon has a sparse record in the middle and late Pleistocene.
The main sites, all located in Morocco unless otherwise indicated (Text-fig. 1), are the following, in decreasing geological age (to avoid confusion with earlier publications [Geraads 1997, 2002, 2006, 2008], the Pliocene/Pleistocene boundary is set at 1.8 Ma):
- 1 Lissasfa, dated to 5.5–6 Ma, by rodents (Geraads 1998), is thus close in age to the Miocene/Pliocene boundary. It yielded only one canid specimen.
- 2 Ahl al Oughlam (AaO), of late Pliocene age, at about 2.5 Ma (Geraads 2002, 2006), has a very rich carnivore fauna (Geraads 1997, 2008), but some complements to the Canidae are given here.
- 3 Aïn Boucherit and Aïn Hanech, Algeria, were excavated by Arambourg but their faunas include only a few carnivores (Arambourg 1979). The former site is probably close in age to the Pliocene/Pleistocene boundary, at c. 1.8–2 Ma, whereas the latter is more probably in the 1.5–1.2 Ma range (Geraads et al. 2004 vs. Sahnouni et al. 2002).
- 4 Tighenif (formerly Ternifine), Algeria, is close to the early/middle Pleistocene boundary, at c. 0.8 Ma (Geraads et al. 1986). The site is famous for its hominid remains, but its carnivores have only been briefly mentioned (Geraads 2008).
- 5 Thomas 1 Quarry includes several levels, the main one being the ‘Grotte à Hominidés’ (Th1-GH) that may be about 0.6 Ma (Raynal et al. 2001).
- 6 The ‘Grotte des Rhinocéros’ (OH1-GDR) in Oulad Hamida 1 Quarry is slightly younger according to rodents, at c. 0.5 Ma (Raynal et al. 2001; Geraads 2002).
- 7 Level D0 of Sidi Abderrahmane (SAR) is certainly younger, at perhaps 0.4 Ma, and level D2 is younger still (Biberson 1961; Raynal et al. 2001).
- 8 Cave fillings in Doukkala Quarries (Michel 1989) are not well dated, but could be of the same broad age.
- 9 Tihodaïne is an open air site in Algeria, probably of similar age (Thomas 1977).
- 10 Jebel Irhoud, famous for its hominid remains, is probably of late middle Pleistocene age (Amani and Geraads 1993; Geraads and Amani 1998).
Other sites are of late Pleistocene age, but their relative chronology is unknown. These are El Harhoura-1 (Aouraghe 2000); El Harhoura-2 (Michel et al. 2009); Sidi Bouknadel (Michel 1989); Kifan bel Ghomari (Mas 1955); ‘Grotte des Félins’ (Raynal et al. 2008) and ‘Grotte des Gazelles’ (unpublished) at Dar Bouazza.
Abbreviations. CCEC: Centre de Conservation et d’Etude des Collections, Lyon; ENM: Ethiopia National Museum, Addis Ababa; INSAP: Institut National des Sciences de l’Archéologie et du Patrimoine, Rabat; FSUR: Faculté des Sciences, Université de Rabat, Maroc; KNM: Kenya National Museums, Nairobi; MNHN: Muséum National d’Histoire Naturelle, Paris. Upper case denotes upper teeth, lower case denotes lower teeth. L = length; W = width. Length of P4 is taken on the labial side, excluding protocone; width of M1 is taken perpendicular to a line tangent to the labial sides of paracone and metacone.
Repository. The material from Sidi Abderrahmane and some of the material from Sidi Bouknadel is housed in MNHN; the material from Jebel Irhoud and the rest of the material from Sidi Bouknadel is housed in FSUR; all other Moroccan specimens are housed in INSAP. The material from Algeria is housed in MNHN.
Order CARNIVORA Bowdich, 1821
Family CANIDAE Gray, 1821
Genus NYCTEREUTES Temminck, 1838
Type species. Nyctereutes procyonoides (Gray, 1834).
Plate 1, figures 3–4
Material. Lissasfa (5.5–6 Ma): unnumbered fragment of maxilla with P4 and slightly incomplete M1.
Description. The large mammal collection from Lissasfa is very poor, but it documents the earliest North African canid, the size of a red fox. This maxilla has a large infra-orbital foramen above the anterior root of P4 and very close to the orbit, denoting a short snout, a P4 very short in comparison to M1 and with a medium size protocone. M1 is large, long relative to its width, with a large hypocone proceeding into a mesial cingulum, but with only a slight distal concavity; what remains of the M2 alveoli shows that this tooth was also large.
Measurements. LP4 = 11.7 mm; WP4 = 6.9 mm; LM1 = 10.0 mm; WM1 = 12.0 mm.
Comparisons. The large crushing component of the dentition is unlike European ‘Canis’ of similar age such as C. cipio from Concud (Pons-Moyá and Crusafont-Pairó 1979), Eucyon debonisi from Venta del Moro (Montoya et al. 2009), and probably also unlike the Vulpes from the late Miocene of Toros Menalla in Chad, whose posterior lower teeth are not large (de Bonis et al. 2007). In most of its characters, the Lissasfa canid resembles Pliocene forms of Nyctereutes, such as N. donnezani from Europe (Deperet 1890, and collections in the CCEC), N. lockwoodi from the Afar in Ethiopia (Geraads et al. 2010), and N. abdeslami from Morocco (Geraads 1997). However, it differs from most of them in the small size of the P4 protocone (a character shared only by N. lockwoodi), and in the large size of P3, as indicated by the posterior root of this tooth. In these two characters, the Lissasfa canid clearly differs from the later N. abdeslami from the same area. The former also resembles the Laetoli canid first described as ‘aff. Canis brevirostris’ (Barry 1987), later referred to Nyctereutes by Werdelin and Lewis (2005), but this eastern African species also has small anterior premolars. Reference to Nyctereutes is the most satisfactory, but I prefer not to attempt species identification or erection of a new name upon this single specimen. Still, it must be noted that if assignment to Nyctereutes is correct, the Lissasfa find is the earliest securely dated record of this genus, pre-dating the European and Chinese ones.
Nyctereutes abdeslami Geraads, 1997
Plate 1, figures 5, 8
Material. Ahl al Oughlam (c. 2.5 Ma): about 100 specimens, mostly isolated teeth, but also including several maxillas and mandibles (most of them listed in Geraads 1997).
Discussion. A canid readily identifiable as a Nyctereutes is the most common member of this family at Ahl al Oughlam. Descriptions, measurements and comparisons can be found in Geraads (1997). It has a well-developed subangular lobe, very large m2 and m3, and an m1 entoconid that is as large as the hypoconid, unlike in most other Nyctereutes. In the PCA (Geraads 1997, fig. 22), it is close to some Spanish forms, but some specimens have an outlier position and I gave it a new species name, N. abdeslami Geraads, 1997. It likely resulted from local evolution of a form of northern origin, as it does not seem to be closely related to other African Nyctereutes, N. terblanchei (Broom, 1948), N. lockwoodi Geraads et al. (2010), and the Laetoli form.
Genus VULPES Frisch, 1775
Type species. Vulpes vulpes (Linnaeus, 1758). Extant.
Vulpes hassani sp. nov.
Plate 1, figures 1–2; Tables 1–2
|L P3||W P3||L P4||W P4||L M1||W M1||L M2||W M2|
|Ahl al Oughlam|
|Vulpes cf. rueppelli|
|Grotte des Gazelles|
|L p3||W p3||L p4||W p4||L m1||W m1||L m2||W m2|
|Ahl al Oughlam|
|Vulpes cf. rueppelli|
|Grotte des Gazelles|
|Mean of 12 m1s3||14.7 (13.5−15.6)||5.9 (5.4−6.4)|
.1997 Vulpes aff. rueppelli Schinz; Geraads, p. 151, fig. 25.
Derivation of name. Dedicated to Hassan Caid Rha, the first excavator of Ahl al Oughlam.
Holotype. AaO-179 (Pl. 1, fig. 1), maxilla with P4, M1 and slightly incomplete M2.
Referred material. Ahl al Oughlam (c. 2.5 Ma): AaO-147, right mandibular ramus complete except for the tip of the coronoid process, but with all teeth missing except m1; AaO-658, maxilla with P2–P4; AaO-3193, maxilla with P4–M1; AaO-652, incomplete P4; AaO-1371, AaO-1374, AaO-1375, M1s; AaO-3150, AaO-3185, M2s; AaO-3070, right mandibular ramus with m1; AaO-2616, right mandibular ramus with incomplete p4 and m1, and m2; AaO-3071, AaO-4111, m1s.
Measurements. Measurements of upper and lower teeth are given in Tables 1 and 2.
Diagnosis. A species of Vulpes of small size (L P4–M2 c. 20 mm); P4 short relative to the molars, and especially relative to M2; M1 with reduced parastyle; lower premolars contiguous; m1 short relative to m2, with an entoconid almost as large as the hypoconid, and sometimes connected to it by a transverse crest.
Description. P4 has a large prominent protocone; this tooth is short compared to the crushing component of the dentition. M1 is broad and twisted, as usual in foxes, but the metaconule is not much reduced; the paracone is distinctly higher than the metacone; these cusps are circled by a cingulum, but the parastyle is not prominent.
The mandibular corpus is slender and shallow (Pl. 1, fig. 2); the slightly incomplete coronoid process was not very high; the angular process is long and slender, the masseteric fossa is very deep. The premolars are broken off on all mandibular rami, but their alveoli show that, except p1, these teeth were contiguous. The metaconid of m1 is rather large; the trigonid angle is about 110°. In the talonid, the hypoconid is larger but not much higher than the entoconid. These cuspids are connected by an incipient crest in AaO-3071, which is unworn; this crest is lacking on AaO-4111, only slightly touched by wear; other teeth are too much worn to ascertain its presence. AaO-2616 preserves the talonid and the anterior alveolus of m1, showing that this tooth was short relative to m2; that this is not just individual variation is shown by two other mandibles that preserve m1 and the alveoli of m2, and display the same proportions (ratio mean L of 2 m2s/mean L of 4 m1s = 0.66).
Comparisons. Vulpes hassani differs from the sand fox of the Sahara, V. rueppelli, in its slightly smaller size, its weaker buccal cingulum of P4 that does not develop into a parastyle, its shorter P4 relative to the molars, and especially relative to M2 (see Text-fig. 2), its less obtuse trigonid angle of m1, and in its m1 shorter relative to m2 (the ratio mean L of m2/mean L of m1 is 0.45–0.58 in 19 V. rueppelli).
Vulpes hassani differs from the Sahelian pale fox V. pallida in its larger size, its weaker buccal cingulum of P4 that does not develop into a parastyle, its slightly longer P4 relative to M2 (see Text-fig. 2), its more obtuse trigonid angle of m1, and in its m1 shorter relative to m2 (the ratio mean L of m2/mean L of m1 is 0.43–0.64 in 14 V. pallida).
Comparisons with V. riffautae from the upper Miocene of Chad (de Bonis et al. 2007) are limited, because this species is known by mandibular remains only, but it differs from V. hassani in its lower premolars separated by diastemas, and shorter m2 (ratio mean L of m2/mean L of m1 is 0.53).
Vulpes hassani differs from the contemporaneous V. alopecoides from Europe in its much shorter m1 compared to m2.
Vulpes cf. rueppelli (Schinz, 1825)
Plate 1, figure 11; Tables 1–2
Material. Tighenif (c. 0.8 Ma): TER-2055, maxilla with P1–M2; TER-2075, 2077, 2078 and no N°, P4s; two unnumbered M1s; an unnumbered M2; TER-2122 and TER-2123: m1s. Fissures in OH1 Quarry, equivalent in age to OH1-GDR (c. 0.5 Ma): anterior portion of a cranium. Pink fillings above level L of Thomas 1 Quarry, equivalent in age to Th1-GH (c. 0.6 Ma): unnumbered m1.
Description. The cranial piece from OH1 preserves P2 to M2, but is strongly crushed, and the molars are heavily worn. The post-orbital processes are depressed dorsally, showing that the frontal sinus had a limited extent. The muzzle is narrow. P4 has a large protocone, and is short relative to M1. The unnumbered m1 from the pink fillings of Thomas 1 has a broad talonid with a large entoconid and a pre-entoconid; there is a distinct crest between the talonid cuspids.
Most specimens from Tighenif are about the size of the one from OH1; the most complete specimen, TER-2055 (Pl. 1, fig. 11), is slightly larger, but its morphology and tooth proportions are similar (Text-fig. 2).
Comparisons and discussion. These foxes differ little from the modern sample of V. rueppelli: their P4s are slightly smaller relative to M2 (Text-fig. 2) and the measurements of the most complete specimen from Tighenif, TER-2055, lie slightly outside the sampled range of the modern form. However, other specimens fall within the size range of the modern form, and it is more parsimonious to include them all in a single species close to V. rueppelli. It would be less satisfactory to split the sample into two groups of closely similar size, and refer a single specimen, TER-2055, to a small V. vulpes.
Vulpes vulpes (Linnaeus, 1758)
Material. Th1-GH (c. 0.6 Ma.): NA-22-21, complete mandible. Doukkala: piece of mandible and some isolated teeth (list in Michel 1989, p. 286). ‘Grotte des Gazelles’: 11 isolated teeth.
Measurements. Measurements of the upper and lower teeth are given in Tables 1 and 2.
Comparisons. The mandible NA-22-21 is of the size and morphology of a small V. vulpes. Its m1 differs from the unnumbered m1 from the pink fillings referred above to V. cf. rueppelli in being larger, narrower, and in having a hypoconid distinctly larger than the entoconid. It is preferable to assign it to a different species rather than lumping all medium-sized Vulpes from Thomas 1 and Oulad Hamida 1 into a single species that would be intermediate in size, but distinct from both V. rueppelli and V. vulpes. Specimens from Doukkala and ‘Grotte des Gazelles’ do not differ from V. vulpes.
Discussion. The red fox, today represented in North Africa by the subspecies V. vulpes atlantica, is present in many upper Pleistocene sites (Michel 1989; Aouraghe 2000), but was not common in middle Pleistocene. Arambourg (1979) recorded the weathered mandible of a medium-sized Vulpes, with much worn p4 and m1, and assumed that it originated from the level of Aïn Hanech, rather than from the earlier one of Aïn Boucherit. It is from an animal smaller than the ‘Canis’ from this latter site, being about the size of the modern V. vulpes atlantica, to which it was referred by Arambourg, but survival of this subspecies for more than 1 Ma. is unlikely, and the species itself is not definitely known anywhere at that time. In any case, this poor specimen, being of very doubtful age and perhaps even sub-fossil, is of little significance.
Genus LUPULELLA Hilzheimer, 1906
Type-species. Canis mesomelas Schreber, 1775. Extant.
Lupulella paralius sp. nov.
Plate 1, figures 5–7; Table 3
|Upper teeth||L P4||W P4||L M1||W M1||L M2||W M2|
|Ahl al Oughlam|
|Lower teeth||L p2||W p2||L p3||W p3||L p4||W p4||L m1||W m1||L m2||W m2|
|Ahl al Oughlam|
.1997 Canis aff. aureus Linnaeus, 1758; Geraads, p. 150, fig. 23.
Derivation of name. From Greek παραλία, seaside, because the type-locality was on the beach at the time of its formation.
Holotype. Ahl al Oughlam (c. 2.5 Ma): AaO-4057, incomplete cranium, strongly crushed and distorted, with P4–M2 (Pl. 1, fig. 5).
Referred material. Ahl al Oughlam: AaO-4119, right mandibular corpus with p2–m2 (Pl. 1, figs 6–7); AaO-3499, right mandibular corpus with p4–m1; 11 isolated teeth.
Diagnosis. A species of medium size. Skull with fully pneumatised postorbital processes, short sagittal crest restricted to the caudal part, occipital surface bell-shaped, inion not projecting. M1 of medium length relative to its width and to P4 length; paracone higher than the metacone. Lower p4 long relative to m1, without second posterior cuspid; m1 with at most incipient transverse cristid between the main talonid cuspids; m2 long relative to m1, but with a narrow talonid.
Description. This canid is larger than foxes, and about the size of the modern jackals. The strong distortion of the skull AaO-4057 allows for only very few measurements (minimum post-orbital width = 34; width over post-orbital processes = 48) and morphological observations. The post-orbital processes are well-inflated and dorsally convex, and the sinus certainly widely invaded the frontals. The outline of the occipital, in caudal view, sharply widens immediately below its dorsal summit, so that this outline is bell-shaped and not triangular. Although crushed, the braincase shows that the sagittal crest was very short and restricted to its most caudal part. The upper carnassial has a moderate protocone, and is moderately long relative to M1. The M1 has a strong antero-labial cingulum almost forming a parastyle, a paracone distinctly higher than the metacone, and a lingual part that is not expanded, with a small hypocone; three other isolated M1s display the same morphology. M2 is small, also with a small hypocone.
On AaO-4119, there is a short diastema between p2 and p3, but none between p3 and p4; the premolars are still more crowded on AaO-3499, probably because this specimen is ontogenetically younger. The p4 has the usual canid morphology, without extra posterior cuspid. There are four specimens of m1, rather homogeneous in size and morphology; the entoconid is always significantly smaller than the hypoconid. The cristid connecting these cuspids is at most incipient. There are five m2s, all with a vestigial paraconid, and a talonid much narrower than the trigonid.
Comparisons. Lupulella paralius is unlike Vulpes and Nyctereutes in its dorsally inflated postorbital processes, but cannot easily be distinguished from the modern jackals (Canis aureus, Lupulella adustus and Lupulella mesomelas). Although there is variation in these features, it differs from most modern specimens of these three species in its short sagittal crest and weak cristid between hypoconid and entoconid of m1.
It further differs from Canis aureus in its less triangular occipital, P4 shorter relative to M1 and M1 longer relative to its width than in most specimens of this species (because of the strong mesio-labial cingulum and small hypocone), p4 slightly longer relative to m1, and narrow talonid on m2 (in C. aureus it is usually as broad as the trigonid, with a well-developed entoconid).
It further differs from modern Lupulella mesomelas in its P4 that is shorter relative to M1, and p4 and m2 longer relative to m1. Fossil jackals have been described from Olduvai as Lupulella mesomelas latirostris (Pohle 1928; Petter 1973), and from South Africa by Ewer (1956) as C. mesomelas pappos. Size range is wide in South African forms, but not more than in modern C. aureus, and the range is smaller within a single site, suggesting that there were regional or temporal variants, although there is no obvious connection with geological age. In their large m2s, like the AaO form, these South African jackals are more like C. aureus, and geography is the only reason for connecting them to Lupulella mesomelas. The Olduvai forms, instead, in their long P4s and shorter m2s with a rather narrow talonid, are more like modern Lupulella mesomelas, and more different from Lupulella paralius.
Lupulella paralius further differs from modern Lupulella adustus in its small M1 hypocone, longer P4 relative to M1, and longer p4 relative to m1.
The holotype m2 of Canis anthus primaevus Arambourg, 1979, from Aïn Boucherit (a site about 0.5 Ma. younger than AaO), does not significantly differ from those of Ahl al Oughlam. The same is true of three P4s from the same locality, but an unnumbered M1, although of similar size, has a more expanded hypocone and stronger lingual cingulum, both features being resemblances with the next species. Therefore, I refrained from assigning them to Lupulella paralius; in any case, there is no positive evidence that they are identical, and the holotype of ‘C. anthus primaevus’ is very unsatisfactory, as it includes only one tooth.
European Plio–Pleistocene canids are larger than Lupulella paralius, even the relatively small C. accitanus Garrido and Arribas, 2008, from Fonelas P1 in Southern Spain, dated to c. 2 Ma., which further differs in its divided M1 hypocone. Canis adoxus Martin, 1973, from Saint-Estève in Southern France, a site that is probably earlier than Ahl al Oughlam, has very similar upper tooth measurements, but its skull is longer and narrower, with a strong sagittal crest, a more triangular occipital (but the slenderness of the skull owes much to crushing), an M1 more expanded lingually, a p4 as short as in C. aureus, and a large central protoconid on m2. None of these differences is great, but their association precludes species identity.
Lupulella mohibi sp. nov.
Plate 1, figure 10; Plate 2, figures 1–8; Tables 4–5
|Condylo-basal length (EG1) = 141.7|
|Length from prosthion to occipital crest (RA3) = 148|
|Length from foramen magnum to front of choanae (EG4) = 56.9|
|Length from prosthion to front of choanae (EG5) = 77|
|Length from prosthion to back of M2 (EG7-RA6) = 78|
|Length from prosthion to orbit (RA3) = 64.7|
|Length from orbit to infra-orbital foramen (RA21) = 16.9|
|Length P1–M2 (RA5) = 53|
|Width over bases of canines (EG11) = 24.2|
|Minimum width of snout (EG12) = 24|
|Bi-mastoid width (EG17-RA22) = 45.4|
|Bi-condylar width (EG18-RA10) = 27.4|
|Braincase width (EG20-RA11) = 48.5|
|Minimum post-orbital width (EG22-RA15) = 29.7|
|Width over post-orbital processes (EG24-RA14) = 37.4|
|Bi-orbital width (EG25-RA13) = 27.5|
|Upper teeth||L P3||W P3||L P4||W P4||L M1||W M1||L M2||W M2|
|Means of isolated teeth||13.9 (13.4−15.1)||7.5 (6.5−8.0)||11.6 (10.5−12.4)||13.6 (12.6−14.5)||7.7 (7.2−8.2)||10.2 (9.5−11.5)|
|n = 6||n = 6||n = 11||n = 11||n = 7||n = 7|
|Means of other teeth||13.7 (12.3−15.0)||7.3 (5.5−8.2)||11.5 (10.1−12.8)||13.2 (10.8−14.8)||8.4 (8.2−8.6)||10.6 (10.5−10.7)|
|n = 6||n = 6||n = 8||n = 8||n = 2||n = 2|
|Lower teeth||L p3||W p3||L p4||W p4||L m1||W m1||L m2||W m2|
|Means of other teeth||17.8 (16.0−19.3)||7.8 (6.4−8.5)||9.8 (9.3−10.5)||6.7 (5.7−7.3)|
|n = 14||n = 14||n = 8||n = 8|
|Means of other teeth||16.6 (15.3−18)||7.1 (6.1−7.7)||9.3 (8.7−10.4)||6.5 (6−7.1)|
|n = 13||n = 13||n = 9||n = 9|
|Means of 7 other m1s||16.1 (15.6−17.1)||6.8 (6.5−7.1)|
.1980 Vulpes cf. atlantica Wagner; Geraads, p. 78, pl. 2, fig. 4.
Derivation of name. Dedicated to Abderrahim Mohib, whose invaluable help greatly facilitates the excavations at Casablanca.
Holotype. OH1-GDR (c. 0.5 Ma): GDR-7966 (Pl. 2, figs 1–5), relatively complete cranium, lacking the auditory bullae, zygomatic arches, nasal bones, and all teeth anterior to the carnassials except the left P3.
Referred material. OH1-GDR: D12-94, left maxilla with P4–M2 (Pl. 2, fig. 6); M10-00, anterior part of a muzzle and fragment of the left orbit; M10-44, maxilla with incomplete P4, M1 and M2; I10-58, M2; C14-00a, mandible with p4–m1; C14-00b, D12-111, D13-38, E13-00, G11-104, X00-034, X00-035, X00-37, m1s; F10-00, GDR-7260, m2s. Th1-GH (c. 0.6 Ma): Th1-10728, right mandible, complete but lacking all teeth except m1 (Pl. 2, fig. 8); Th1-10332, mandible with m1–m2; SA26-97, right mandibular ramus with p3–m2 (Pl. 1, fig. 10); Th1-C, maxilla with P4–M2 (Geraads 1980, pl. 2, fig. 4); seven isolated m1s. Thomas 1 or Oulad Hamida 1 Quarry: Th/OH-A, fragment of orbit and muzzle with M1-M2; Th/OH-B, maxilla with P3–M2. Sidi-Abderrahmane D0 (c. 0.4 Ma): 1957-9, maxilla with P3–M1. Tighenif (c. 0.8 Ma): TER-2057, mandible ramus with p4–m2 (Pl. 2, fig. 7); TER-2058, mandible ramus with p3–m1; about 50 isolated teeth (10 P4s, 11 M1s, 7 M2s, 1 p4, 14 m1s and 8 m2s).
Diagnosis. A medium-sized canid, remarkable by the large size of the crushing part of its dentition. Sagittal crest very short. Third incisors small. M1 trapezoidal rather than triangular, as long lingually as labially, with a very strong hypocone, and a continuous labial cingulum.
Differential characteristics. Lupulella mohibi differs from Nyctereutes in its weaker postorbital processes, narrower occipital, and complete lack of sub-angular lobe at the mandible. It differs from Canis aureus and Lupulella mesomelas in the short sectorial part of the dentition. It differs from Lupulella adustus in a less low and less narrow cranium in the orbital region, lack of inflated fronto-nasal area, more posterior infra-orbital foramen, M1 that is shorter in respect to its width, and p4 that is shorter relative to m1 and m2.
Description. The holotype, found in 2009, is the best specimen of a fossil North African canid, and one of the most complete crania in Africa. Its most striking cranial features are the small size of the occipital, and the shortness of the sagittal crest. The temporal lines remain about 10 mm apart for most of their lengths, and unite only for a short distance immediately before joining the nuchal crest, but this occipital protuberance remains weak. These features give the skull a juvenile aspect, although it is fully adult. The frontal bones are well inflated on either sides of the depressed mid-line, and no depression remains in the post-orbital processes.
From Thomas 1 and Oulad Hamida 1, the vast majority of canid teeth are of the same size and basic morphology as the holotype; therefore, the tooth sample is large enough to assess its metric features and its variability. There is no in situ incisor, but the holotype and M10-00 preserve the alveoli of the upper ones; that of I3 is not much larger than that of I2, and is contiguous with it, showing that both I3 and i3 were small. The P4 is short relative to M1 and M2, and so is the crushing part of the lower dentition (talonid of m1, and m2). The carnassial has a well-developed protocone, reaching farther mesially than the rest of the tooth. The molars are bunodont, with low tubercles. The paracone and metacone of M1 are sub-equal in size and height, and bordered by a continuous cingulum, albeit variably developed. The metaconule is large in most specimens, its area being often larger than that of the protocone; only in the holotype is it much reduced and almost fused with the hypocone. The hypocone rarely shows an incipient division; it is always very strong and protruding disto-lingually, so that the distal border of the tooth is distinctly concave. A well-marked anterior cingulum, usually interrupted at the base of the protocone, contributes to the expansion of the lingual half of the tooth. The hypocone is also remarkably expanded on the large M2, where it continues, as a robust cingulum, to the disto-labial angle. The trigonid of m1 has relatively low cuspids; both the hypoconid and entoconid send weak cristids towards each other, but the intervening valley remains visible until late wear.
The morphology and relative proportions of the teeth from Tighenif do not differ from those of Th1-OH1, but they average slightly larger (Text-fig. 3). As in these sites, the mandibles lack a sub-angular lobe.
Mammalian fossils found in level D0 of the Sidi Abderrahmane Quarry a long time ago (Biberson 1961, p. 153) include some upper teeth of medium-sized canids that were identified by C. Arambourg as Canis‘anthus’ (=C. aureus). They share the same morphology as the Tighenif and Th1-OH1 species, but are slightly smaller.
Comparisons. The large size of the crushing part of the dentition calls for comparisons with Nyctereutes, to which this species was tentatively referred (Geraads 2008), before the discovery of the holotype. Indeed, the proportions LP4 vs. LM1 and LM1 vs. WM1 are similar, isolated M1s are virtually identical with those of Nyctereutes abdeslami, and the dentition of the canid from Th1-OH1 is strongly reminiscent of this genus, but the skull precludes such an assignment. Middle Pleistocene Nyctereutes, all from Asia, are similar to, or perhaps even identical with, the modern raccoon-dog N. procyonoides, and are quite distinctive in their shortened skull with reduced snout and strong mandibular subangular lobe. Lupulella mohibi, instead, has a snout that is not reduced, and all mandibles fully lack a pre-angular lobe. Even the late Pliocene N. abdeslami, also from Casablanca, is already a highly derived representative of the genus. Early forms of Nyctereutes, in the early Pliocene, are more primitive, but their cranium is already broad at mastoid level and over the postorbital processes, which are not fully pneumatised. Assigning Lupulella mohibi to Nyctereutes would require several evolutionary changes highly divergent from those followed by other species of this genus, whereas inclusion into the Canina (‘Canis s.l.’) requires no such unlikely change.
Present-day similar-sized members of this group in Africa are the golden jackal Canis aureus, the side-striped jackal Lupulella adustus, and the black-backed jackal Lupulella mesomelas. Their skulls are very similar to each other, and differences between them amount mostly to proportions, but the very short sagittal crest of Lupulella mohibi is a difference from virtually all modern specimens. It also differs from Lupulella mesomelas and C. aureus in the small size of the occipital (as shown by the comparison of bi-mastoid widths), but other cranial features do not differ from these modern jackals. Differences from Lupulella adustus are more obvious, because the skull of the latter is remarkably long and narrow, with an inflation of the fronto-nasal area in front of the orbits, giving it a characteristic lateral profile, the dorso-ventral depth of the cranium remaining almost unchanged from the occipital to the level of P4; by contrast, GDR-7966 has a normal profile, the depth decreasing forward, but its narrow occipital is a cranial resemblance with Lupulella adustus.
The jugal teeth most resemble those of Lupulella adustus and contrast with those of other jackals (and canids of the wolf group) in the large size of the crushing component of the dentition and bunodont tubercles. There are some differences with Lupulella adustus: in the fossil form, M1 is shorter relative to its width, that of Lupulella adustus being more triangular, whereas the lingual part is well-developed in Lupulella mohibi, giving the tooth a more trapezoidal outline; the metaconule of this tooth is larger than in the modern form, and p4 is shorter relative to m1 and m2.
In its short P4 compared to the molars, Lupulella mohibi resembles Canis simensis, the present-day Simien wolf of Ethiopia, and the Th1-OH1 canid was referred to this species by Olive (2006), but C. simensis is quite unlike the North African fossil form in its large size, strong sagittal crest, long slender snout with widely spaced premolars and posteriorly located orbit.
Genus CANIS Linnaeus, 1758
Canis aureus Linnaeus, 1758
Material. Doukkala: skull fragments, tooth rows and isolated teeth (most of them not seen by me; list in Michel 1989, p. 218); Jebel Irhoud: maxilla and mandible fragments, some isolated teeth (list in Amani 1991, p. 137); Tihodaïne: a mandible (Thomas 1977); Sidi Abderrahmane D2: a maxilla. The species is also present in several Upper Pleistocene sites: Sidi Bouknadel, ‘Grotte des Félins’, ‘Grotte des Gazelles’, Kifan Bel Ghomari, El Harhoura-1 and -2. (Mas 1955; Michel 1989; Aouraghe 2000; Raynal et al. 2008; Michel et al. 2009).
Discussion. Canis aureus is the only present-day jackal of North Africa and, as the tooth proportions of the fossils listed above agree better with this species than with other jackals, there is no doubt that they are closely related to it, but display significant size variations (Text-fig. 4). In some localities (El Harhoura-1, Sidi Bouknadel), their size-range is similar to that of the modern form (which is often larger in the Maghreb than in sub-Saharan Africa, and still larger in Egypt); in those sites, the smallest specimens are but slightly larger than the largest co-occurring foxes. In most sites, instead, measurements correspond to the upper part of the size-range of the modern form, even exceeding the size of the north-western African ones, and matching that of the very large C. aureus lupaster of Egypt. This is most clear at Jebel Irhoud (Amani 1991) and at the recently studied ‘Grotte des Gazelles’, two sites where there are enough specimens to rule out a sampling artefact. The large size of this canid prompted some authors (Michel 1989; Amani and Geraads 1993; Michel et al. 2009) to assume that it was distinct from C. aureus (just like C. aureus lupaster was sometimes considered a species of its own). However, I now consider unlikely that, in some sites only, a different species, morphologically similar and only slightly larger, would have replaced C. aureus, and I include them all in this species.
Canis africanus Pohle, 1928
Plate 1, figure 12
- * 1928 Canis africanus Pohle, p. 52, figs 1, 4, 7.
v. 1979 Canis atrox Broom; Arambourg, pl. 58, fig. 8.
1994 Canis (Xenocyon) africanus Pohle; Rook, p. 78.
p. 2003 Lycaon lycaonoides (Kretzoi); Martínez-Navarro and Rook, p. 699, fig. 1H (not fig. 1G as indicated in their caption).
Material. Aïn Hanech: 1949-2-723, left maxilla.
Description. The maxilla described and illustrated by Arambourg (1979, pl. 58, fig. 8) is large. The anterior premolars are missing, but the tooth-row was distinctly concave labially. On M1, the hypocone is unreduced and separated from the trigon by a well-marked valley, indicative of a relatively large entoconid on m1; the metaconule is reduced; the difference in size between paracone and metacone is not extreme, and the tooth has the usual canid proportions. The M2 is distinctly reduced, but not extremely so.
Measurements. P4: L = c. 24.9 mm, W = 12.8 mm; M1: L = c. 15.7 mm, W = 18.6 mm; M2: L = 8.5 mm, W = 11.6 mm.
Comparisons. The Aïn Hanech canid is larger than the jackals, and resembles Canis africanus Pohle, 1928, from Olduvai bed II. This species is based upon a cranium from Olduvai that is probably lost (L. Werdelin, pers. comm.), but a second skull with associated lower jaw, from bed I, is preserved in the KNM (Old-74, Long K1 bone bed 87). The metaconule of M1 is missing on Old-74, but looks faint on the type specimen (Pohle 1928, pl. 4, fig. 1); Old-74 clearly shows that the hypocone of that tooth is not reduced (and m1 has a well developed entoconid), and both specimens show that M2 is reduced. All these features match those of the Aïn Hanech maxilla. Measurements of the Aïn Hanech maxilla are also quite similar to those of C. africanus, to which it can confidently be referred. As supposed by Rook (1994), this species might be identical with Canis atrox Broom in Broom and Schepers, 1946, to which it was referred by Arambourg (1979), but this species, defined at Kromdraai upon an m1 and an M1 (Broom and Schepers 1946, Broom 1948), remains so poorly known that assignment of any other fossil to it is highly speculative; in any case Pohle’s name has priority.
Genus LYCAON Brookes in Griffith et al., 1827
Lycaon magnus Ewer and Singer, 1956
Plate 1, figure 9
Material. Tighenif (c. 0.8 Ma): TER-2053 and 2054: M1s; two unnumbered premolars. Th1-GH (c. 0.6 Ma): unnumbered m1 (Geraads 1980, pl. 2, fig. 2). OH1-GDR (c. 0.5 Ma): F10-27, m1.
Description. The M1s (Pl. 1, fig. 9) are long in comparison to their width (TER-2053: L = 16.7 mm, W = 18.6 mm; TER-2054: L = 16.5 mm, W = 17.5 mm), and their hypocone is much reduced. The two premolars can probably be assigned to the upper jaw because of their slanting main cusp. The m1s (F10-27: L = 30.2 mm, W = 13.0 mm) have no entoconid and no hypoconulid, and a very small metaconid.
Comparisons. These teeth are characteristic of Lycaon by the reduced hypocone and great length of the M1s compared to their width, and by the reduced metaconid and absence of entoconid on the m1s. They differ from the modern Lycaon pictus in their larger size, and in the very reduced metaconid of m1 and complete lack of hypoconulid, which is often present in the modern form. If the premolars are indeed upper teeth, they differ from those of Lycaon pictus by their lower and less lanceolate main cusp.
These premolars much differ from the lower premolars of the modern form, but less so from the p2 and p3 of the Elandsfontein Lycaon (Hendey 1974, fig. 56), which are much less different from the upper ones than those of the modern form, as they lack anterior accessory cuspid and convex anterior edge; in these features, they resemble the Tighenif premolars. In its large size and reduced m1 metaconid, the North African Lycaon also resemble the South African form that Ewer and Singer (1956) and Hendey (1974) assigned to a subspecies of Lycaon pictus, Lycaon pictus magnus, but the differences in size, lack of anterior cuspids on the lower premolars, and size of the m1 metaconid, warrant species distinction. Lycaon magnus therefore appears to have had a pan-African distribution in early and middle Pleistocene times.
Lycaon pictus (Temminck, 1820)
Material. Doukkala II (0.2 Ma ?): 86-DKII-2, mandible with front teeth and premolars.
Remarks. The Doukkala mandible (Michel 1989) shows that by the late middle Pleistocene the lower premolars had acquired the same morphology as in the modern form; they are also smaller than those of Lycaon magnus, and fall within the range of the modern species, to which they can accordingly be referred.
Lycaon has also been recorded from the upper Pleistocene Sintès Quarry (Arambourg 1932) on the basis of an upper M1, similar to that of Lycaon pictus.
Material. Th1-GH, TA28-5, m1 lacking part of the talonid; from the upper part of the cave filling.
Remarks. This single unworn tooth belonged to an animal the size of a large jackal (L = c. 21.0; W = 8.0). It is slightly (especially the protoconid) lower-crowned than that of C. aureus, and the hypoconid has a less distinct ridge descending lingually. These features are more like Lupulella mohibi, but the tooth is significantly larger than in this species, and there is no tooth of intermediate size at Thomas 1.
Fossil North African Canidae significantly contribute to the phylogeny of the African representatives of this family (Text-fig. 5). They may document the earliest Nyctereutes, confirming the early divergence of this genus, and they include several new species, in both the Vulpes and Canina clades.
Vulpes hassani may not be far remote from the common ancestor of the modern North African foxes V. rueppelli and V. pallida. Lindblad-Toh et al. (2005) reached the conclusion that the red fox V. vulpes is the closest living relative of V. rueppelli, but did not include V. pallida in their analysis; the fossils from Ahl al Oughlam strongly suggest that they are more closely related than either is to V. vulpes. The virtual absence of overlap in their geographic ranges (Brito et al. 2009) suggests allopatric speciation. The V. cf. rueppelli from Tighenif and Th1-OH1, although closer to the modern V. rueppelli than to any other species, is reminiscent of the earlier V. hassani in its small size and in its P4 that is slightly shorter than in the modern form, lending some support to the hypothesis of an ancestor-descendant relationship between these species. If all specimens from Tighenif are assigned to V. cf. rueppelli, and if the Aïn Hanech specimen is dismissed, the arrival of the red fox V. vulpes in North Africa can be dated to c. 0.7 Ma., probably not much after the emergence of this species in Eurasia from a form close to V. alopecoides (Kurten and Crusafont-Pairó 1977).
Systematics of the Canina (Canis s.l.) is still debatable, as recent analyses differ in the placement of several taxa, especially Cuon and Lycaon, but all agree in recognizing that Canis is not monophyletic. Lindblad-Toh et al. (2005) and Bardeleben et al. (2005) agree in grouping ‘C.’ adustus with ‘C.’ mesomelas as the sister-group of Canis s.s., thus recognizing an African group that could be called either Lupulella Hilzheimer, 1906 (based upon C. mesomelas) or Schaeffia Hilzheimer, 1906 (based upon C. adustus). The arrangement proposed by Zrzavý and Řičánková (2004) differs slightly, but these two species also occupy a basal position with respect to Canis s.s. Agnarsson et al. (2010) place ‘C.’ adustus as the sister-taxon of a group including all other members of the Canis s.l. group except ‘C.’ mesomelas, a species that branches below most American canids in their tree, a very unlikely position. On the whole, the most likely hypothesis is a dichotomy between an African group (Lupulella mesomelas + Schaeffia adustus) that I call Lupulella, and a mostly extra-African Canis group (but including also C. simensis). Most of the fossil African forms, even including those from Lemudong’o (Howell and García 2007) and Lukeino (Morales et al. 2005), probably belong to the former group, if indeed the dichotomy dates from the Late Miocene (Lindblad-Toh et al. 2005). The main problem with fossil forms is that morphology does not follow genetics, as Lupulella adustus and Lupulella mesomelas are the most divergent in dental morphology, perhaps because of character displacement, as these species are largely sympatric (van Valkenburg and Wayne 1994). The skull offers few clues, but occipital shape of the Ahl al Oughlam canid is definitely more like those of African jackals; hence I refer it to Lupulella, as Lupulella paralius. It is likely that the relatively long M1s and m2s that it shares with modern C. aureus are primitive features for jackals (see also Hendey 1974, fig. 46), not indicative of close relationships between this latter species and fossil African forms. Among medium-sized African canids, only the ‘Canis new species A’ (Werdelin and Lewis 2000) from South Turkwel has a long m1 compared to m2.
By contrast, several authors (Spassov and Rook 2006; Rook 2009; Tedford et al. 2009) have assigned the Ahl al Oughlam medium-sized canid to Eucyon Tedford and Qiu, 1996. Indeed, some of its characters, in particular the bell-shaped occipital and the faint crest joining the hypoconid and entoconid on m1, attest an evolutionary grade more primitive than in Pleistocene Old World Canis s.l., and similar to that of the species included in Eucyon. However, the latter genus ‘has no autapomorphies’ (Tedford et al. 2009, p. 89), and is characterized only as a Canina lacking the derived features of Canis. It is significant that, although several species have been assigned to Eucyon, no diagnosis of the genus has ever been provided; indeed, species grouped in this genus compose a paraphyletic assemblage, and adding the Ahl al Oughlam species to it would further increase its heterogeneity. Ironically, several fossil forms have been included in Eucyon because of the purported occurrence of a second posterior cuspid on p4, a feature first said to characterize Eucyon (Tedford and Qiu, 1996), but latter recognized in the Canina as a whole (Tedford et al. 2009); it is in fact rather uncommon in this sub-tribe.
The middle Pleistocene Lupulella mohibi is so unlike contemporaneous Eurasian canids of the wolf group that its ancestry is certainly not to be sought in the North; therefore, it can confidently be assigned to the African group. However, most canids from earlier North African sites can be also dismissed as its possible ancestors, because they can be assigned to the Vulpes or Nyctereutes lineages. The maxilla from the Miocene–Pliocene of Lissasfa tentatively referred above to Nyctereutes has a predominantly crushing dentition, but its P4 is even shorter and the snout is also short, two characters that are probably derived, and ill-suited for an ancestor. Only Lupulella paralius from Ahl al Oughlam, with its unspecialized dentition, cannot be discarded as a plausible ancestor, and this relationship is supported by the (limited) material from Aïn Boucherit, intermediate in age and morphology. It seems therefore reasonable to include them both in the African clade Lupulella rather than in an ill-defined Canis s.l. There is little doubt that the populations of Lupulella mohibi from Tighenif, Th1-OH1 and SAR-D0 form a single lineage, and we may conclude, from a relatively large sample (altogether, about 60 P4s, M1s and m1s) that size decreased in this species during the first half of the middle Pleistocene. No remain is known in younger sites, and the species certainly went extinct before the end of the middle Pleistocene, when it became replaced by Canis aureus; all late Pleistocene North African jackals can be assigned to this latter species.
The issue of the origin of Lycaon and of its relationships with the Eurasian Pleistocene canid known as Xenocyon has never been seriously addressed. Inclusion of the type-species X. lycaonoides Kretzoi, 1938 in Lycaon is often rejected because this species lacks some of the dental specializations of the modern form, but so does Lycaon magnus. Evidence about the occurrence of a MC I (the loss of which is a derived feature of Lycaon pictus) is contradictory (Martinez-Navarro and Rook 2003 vs. Moulle and Echassoux 2005). According to Schütt (1973), the m1 talonid of X. lycaonoides is shorter than that of Lycaon pictus; this would speak against an ancestor-descendant relationship, but this difference is not obvious. The absence in Africa of any canid sharing clear synapomorphies with the hunting dog before the latest early Pleistocene speaks against a local origin, but the skull of the European X. falconeri (Palmqvist et al. 1999) is not Lycaon-like either, so that its origin remains disputable.
Canis africanus was called Canis (Xenocyon) africanus by Rook (1994), and included in Lycaon lycaonoides (Kretzoi, 1938) by Martinez-Navarro and Rook (2003). However, besides the fact that Pohle’s name has priority over Kretzoi’s, these authors failed to mention any character shared by C. africanus and the modern Lycaon pictus, so that the purported phyletic link is unsupported. In fact, both skulls of C. africanus from Olduvai are as different from the skull of Lycaon pictus as those of similar-sized canids could be: whereas the latter is almost hyaenid-like, short and broad, with especially a broad and short snout, and straight premolar rows, the Olduvai form has a long narrow snout with concave dorsal profile, narrow palate, labially concave tooth-rows, narrow braincase, and much inflated frontals. The only tooth resemblance with Lycaon is the small size of M2, indicative of hypercarnivory (also found in C. lupus), but neither the Olduvai specimens nor the Aïn Hanech one show(s) reduction of the M1 hypocone. This could of course be because they retain this primitive feature, but the skull morphology speaks against a close relationship, and Canis africanus is more probably a true Canis s.s.
Climatic and biogeographic inferences
Carnivores are usually not very helpful paleoclimatic indicators, but the Canidae do provide some hints about past North African climates. Today, the sand fox Vulpes rueppelli is restricted to the Sahara and its periphery, being absent north of the Atlas, and if we assume similar ecological requirements for the fossil form, this would suggest sub-desert conditions for Tighenif and the Thomas Quarries. In the late Pleistocene, the size of North African V. vulpes may exceed that of the living form. This contrasts with the small size of modern North African V. vulpes, in which the maximum length of 27 m1s collected from Morocco to Egypt (most measurements kindly provided by T. Dayan) is 15.2 mm. From El Harhoura-1 (Aouraghe 2000), Sidi Bouknadel (Michel 1989, and pers. obs.), and ‘Grotte des Gazelles’, five out of 18 m1s exceed this length. Davis (1977) surmised that size in the red fox is negatively correlated with temperature, which would suggest that the North African late Pleistocene climate was significantly cooler than today’s, a result that can easily be accepted. Yet, size variations in Canis aureus do not parallel those of foxes, as would be expected if they were caused by temperature changes or character displacement: there are some large Vulpes at El Harhoura-1 and Sidi Bouknadel, where the size gap between them and the smallest Canis specimens is narrow, and at the ‘Grotte des Gazelles’, where it is very wide. Prey size could have been invoked to explain size variations in C. aureus, but at both El Harhoura-1 and ‘Grotte des Gazelles’, the largest potential preys are medium-sized gazelles. The badly needed relative dating of these upper Pleistocene sites might help solving these issues by detecting temporal trends, if any.
Biogeographic affinities of the North African canids are mixed. They definitely include some endemic forms, such as the strange Lupulella mohibi and V. hassani (as well as its likely descendant V. rueppelli), which look rather distant from other Eurasian or African forms. If the Lissasfa canid is indeed a Nyctereutes, it also has no equivalent elsewhere. However, other forms attest some exchanges with Eurasia. The Ahl al Oughlam Nyctereutes is closer to Spanish forms than to the earlier East African N. lockwoodi or later South-African N. terblanchei, probably implying crossing of the Gibraltar Straits around 3 Ma. Vulpes vulpes is certainly also a Eurasian immigrant around 0.7 Ma, like C. aureus slightly later. On the whole, as in several other mammalian groups, the Palearctic stamp on the canid fauna increases during the Pleistocene, both because of arrival of new forms from the North, and extinction of endemic or African lineages.
As in most other sites of the late Cenozoic, the Canidae are one of the main carnivore families in North-western Africa. At Ahl al Oughlam, a site that probably acted as a natural trap but was also used as a den by large carnivores, Canidae are about as common as Felidae, but less common than Hyaenidae. Sample size is small in the lower Pleistocene, but the Canidae became the most abundant family at Tighenif, and their dominance increased in middle and upper Pleistocene sites, even though most of them are cave sites. They usually include three species with diverse dietary adaptations. At Ahl al Oughlam, only Lupulella paralius was probably a predator of small mammals, while the dominant form, Nyctereutes, was certainly omnivorous. At some time during the early Pleistocene, Nyctereutes went extinct in North Africa, and its niche was filled by Lupulella mohibi, a jackal whose dentition paralleled that of the form that it replaced. It seems that during the Pliocene, early Pleistocene, and most of the middle Pleistocene, the dominant forms among canids were omnivorous; surprisingly, felids of small to medium size are also uncommon, perhaps leaving the niche of predators of this size partly empty. The only truly predator canid during most of the middle Pleistocene was Lycaon, and it is only towards the end of this period that the golden jackal C. aureus appeared. It is remarkable that its chronological range does not overlap that of Lupulella mohibi that it replaced between Sidi Abderrahmane D0 and SAR-D2, suggesting that the diets of these species were less distinct than their dental features.
Acknowledgements. Most of the data for this paper were collected in Morocco in the frame of the ‘Programme Casablanca’ of the Institut National des Sciences de l’Archéologie et du Patrimoine’, and I thank its leaders, J.-P. Raynal and F.Z. Sbihi-Alaoui, for constant support. Field work was funded by the ‘Ministère des Affaires Etrangères et Européennes’, ‘Région Aquitaine’ (grants ‘Origines’ and ‘Origines II’) and Department of Human Evolution of the Max Planck Institute in Leipzig. I also thank people who gave access to collections in their care: C. Argot, J. Lesur, and C. Sagne (MNHN), D. Berthet (CCEC), M. Boutakiout (FSUR), E. Mbua and F. Kyalo (KNM), and M. Yilma (ENM); thanks also to D. Reed, S. Stouge, and the reviewers, S. Peigné, L. Werdelin, and X. Wang, for helpful comments.
Editor. Dr. Marcello Ruta