An evaluation of classical morphologic and morphometric parameters reported to distinguish wolves and dogs

Highlights

• We tested classical morphometric criteria, generally accepted to discern archaeological dogs and wolves.

• Most criteria were proved to be non-diagnostic.

• P4 mesio-distal diameter and skull length are diagnostic, four other criteria are helpful but only in extreme values or means.

• Upper-Paleolithic proto-dogs are not different from contemporary wolves.

Abstract

Morphological and morphometric bone variation between archaeological wolves and the oldest domestic dogs commonly are used to define species differences. However, reference data often have been based on small numbers, without robust statistical support. We consulted the literature on these matters in all possible languages and tested many of the proposed species differences by examining wolf and dog skeletons from several collections, accompanied by an extensive synthesis of existing literature. We thus created large reference groups, assessing data distributions and variability. We examined mandible height, width, length, and convexity; contact points of the skull on a horizontal plane; caudal shifting of the border of the hard palate; skull size; carnassials tooth size reduction; micro-anatomical differences in teeth, snout, and skull height; and snout length and width.

Our results show that skull length and related size; skull height; snout width; orbital angle; P4 and M1 mesio-distal diameter can help (albeit to a limited extent) to distinguish the oldest archaeological dogs from wolves. Based on our observations, we re-evaluated recent large Pleistocene canids reported as Paleolithic dogs and concluded instead that they fit well within the morphomentric distributions seen with Pleistocene wolves.

The research presented here reflects the recent trend to critically re-evaluate axiomatic assumptions about wolf-dog differences, and to rephrase the morphological and morphometric definition of an early archaeological dog in a more suitable manner.

These results are important to the international archaeological community because they place historical reports in a newer context, and create a robust (although narrow) framework for further evaluation of archaeological dogs and wolves.

Introduction

Researchers have examined the origins of dogs for over a century (Galton, 1865; Gaudry and Boule, 1892; Huxley, 1880; Nehring, 1888; Rütimeyer, 1861; Studer, 1901; Verworn et al., 1919; Wolfgram, 1894) and have used specific morphometric and morphological criteria to assign specimens as dog or wolf.

Some of these assignment criteria have been rejected, including:

• backward turning of the dorsal part of the vertical mandibular ramus (Olsen and Olsen, 1977), because the feature is present in both dogs and wolves (Janssens et al., 2016a);

• variations in types and frequencies of oral and dental abnormalities such as tooth agenesis or occlusion-related pathology (Andersone and Ozolins, 2000; Stockhaus, 1965; Vila et al., 1993; Wobeser, 1992) that are similar among wolves and dogs (Janssens et al., 2016b);

• paedomorphosis (juvenile appearance in dogs) (Morey, 1994; Waller et al., 2013) for which the typical morphological criteria are lacking in dogs (Drake, 2011);

• tooth crowding in dogs (Benecke, 1994; Wolfgram, 1894) that occurs also in wolves (Ameen et al., 2017);

• larger tympanic bulla in wolves (Benecke, 1987; Bökönyi, 1975; Zeuner, 1963) that is not species-specific, but is related isometrically to stature (Stockhaus, 1965);

• differences in the sum of maxillary mesio-distal diameters of M1 + M2, being less than P4 in dogs and greater in wolves (Clutton-Brock, 1963); many possible variations exist in dogs and wolves (Gaudry and Boule, 1892; Wolfgram, 1894);

• reduced sagittal crest in dogs (Studer, 1901) (bony protuberance in midline sagittal suture line located at the level of the parietal bones) given that dogs are quite variable in this criteria (Lawrence and Bossert, 1967; Rizk, 2012);

• clear difference in orbital angle (OA) between dogs and wolves (Bockelmann, 1920; Iljin, 1941; Studer, 1901) that was contradicted by a recent study showing there was full overlap between archaeological dogs and modern wolves (Janssens et al., 2016c);

• convex mandible in dogs versus a straight one in wolves (Germonpré et al., 2015; Lawrence and Reed, 1983), that was agreed on in modern samples but incorrect in archaeological samples (Drake et al., 2017).

Several anatomical criteria are accepted today, and are used routinely to classify candidate archaeological canids as dog or wolf. However, documenting observations may include relatively few reference individuals, or results can be contradictory (Aaris-Sørensen, 2004; Clutton-Brock, 1962; Clutton-Brock, 1963; Davis and Valla, 1978; Dayan, 1994a; Degerbøl, 1961; Lawrence and Bossert, 1967; Musil, 2000). Among these criteria are:

Differences in mass have been reported between wolf and dog mandibles. Massive mandibles were defined as deep dorso-ventrally (high) and thick latero-medially (wide) in some prehistoric canids that were identified as dogs (Clutton-Brock, 1962; Germonpré et al., 2015; Lawrence and Bossert, 1967; Tchernov and Horwitz, 1991). The problem is that mass is not defined objectively as an index of width (breadth), length, and height. Here we discuss published mandibular measurements to define mass, comparing mandibular length, width, and height.

There are two useful studies on mandible width (maximal transverse distance at M1-measure 17), in Von den Driesch (1976). Lawrence and Reed (1983: 490–494) found dog mandibles from Jarmo (Iraq, 8 kya) to be wider than those of local recent wolves (C. l. pallipes or arabs) and comparable to “Eskimo dogs”, which the authors consider to be an “archaic breed”. This finding was used as an argument to identify the specimens as dogs. The second study (Germonpré et al., 2015: 12) measured canid mandibles from the archaeological site of Předmostí (Czech Republic, 27–24 kya) and divided them into two morphotypes: a dog-like massive type (proposed to be a Paleolithic dog) and a slender type (proposed to be a Pleistocene wolf). There was a significant difference between modern wolves and the proposed Paleolithic dogs, but not between the latter group and the Pleistocene wolves.

Three studies on mandibular height (the maximal sagittal height of the horizontal ramus of the mandible at the molars - Von den Driesch (1976; 60); measure 19) present conflicting data. Tchernov and Valla (1997) found this height significantly greater (p < 0.0001) in modern Middle Eastern wolves, as compared to Natufian dogs (Levant, 15–11.5 kya) (Grossman, 2013). This seems logical as mandible height has been correlated with body mass (Losey et al., 2015). Dimitrijević (2006) and Dimitrijević and Vuković (2012) found mandibles from Mesolithic and Neolithic dogs in the Danube Gorges to be taller than in archaeological and recent local wolves. Germonpré et al. (2015: 12) measured three different mandibular heights (between P2–3 (Von den Driesch measurement 20), P3–4, and M1–2) but no significant statistical difference was found between the mandibular heights of proposed Paleolithic dogs and Pleistocene wolves at Předmostí.

One study reports mandible length (Von den Driesch (1976), measures 1, 2, 3) and other dimensions of 130 canid skulls from the Gravettian site of Předmostí (Czech Republic) and compared these to modern wolves and dogs, using uni- and multivariate statistical methods (Germonpré et al., 2015). Ten metrics were evaluated (for details see Table 1) including (a) three mandibular lengths; (b) several tooth row lengths (total length, several premolar and molar lengths); (c) two individual tooth lengths (M1 and C); (d) two widths (mandible and M1); and (e) three heights. Tooth row lengths also have been studied by several other authors reporting on tooth row length reduction (most premolar) in dogs compared to wolves (Benecke, 1987; Benecke, 1994; Boudadi-Maligne, 2010; Bökönyi, 1975; Clutton-Brock, 1995; Dayan, 1994b; Dimitrijević and Vuković, 2012; Morey, 2010; Nehring, 1888; Napierala and Uerpmann, 2012; Studer, 1901; Tchernov and Horwitz, 1991; Tchernov and Valla, 1997; Wolfgram, 1894; Zeuner, 1963). The results of the Předmostí study (Germonpré et al., 2015) were that 28% of the total canids at the site had nine statistically significant shorter length metrics (Von den Driesch (1976) mandibular measures 1, 3, 4, 8, 9, 10, 11, 12 and13a) than the others, and these were identified as Paleolithic dogs (versus wolves). Some caution is advised about data of this nature. Multiple measurements taken from the same structure (mandible) directly imply non-independence among the measurements, and increase the likelihood that the traits will be correlated based on size. Cluster analysis followed by significance tests attempting to detect subgroups in such a dataset is actually a circular way of statistical thinking. If one takes a sample of morphological measurements, then looks for the highest differences on the basis of a cluster analysis, and then statistically tests if these two most different groups are significantly different, almost any dataset would yield significant p-values. It is thus very likely to find such differences even when the sample comes from a homogeneous group of specimens (as in our test in of German shepherds, see below).

Different contact points of the skull on a horizontal plane, with wolves resting on bullae and canine teeth and dogs on bullae and P4 (Benecke, 1987; Zeuner, 1963) Zeuner (1963) explained the difference as due to the larger canines in wolves (Fig. 2).

Different position of the caudal border of the hard palate, being more caudal in dogs, so that the line connecting caudal M2, lies cranial to that border in dogs, caudal in wolves (Benecke, 1987; Iljin, 1941). However, not all agree (Mihelic et al., 2013; Schwabenlander et al., 2015) (Fig. 3).

Smaller stature and size and skull size are accepted widely as domestication signals documenting dogs (Table 6) (Benecke, 1994; Boudadi-Maligne et al., 2012; Clutton-Brock, 1962; Clutton-Brock, 2012; Dayan, 1994b; Degerbøl, 1961; Mertens, 1936; Napierala and Uerpmann, 2012; Pitulko and Kasparov, 2017; Pluskowski, 2006; Rütimeyer, 1861; Wolfgram, 1894; Zeuner, 1963).

Smaller skull size (mainly lengths were measured) has been reported widely as a feature of early domesticated dogs (Aaris-Sørensen, 1977; Andersone and Ozolins, 2000; Benecke, 1987; Chaix, 2000; Degerbøl, 1961; Dimitrijević, 2006; Germonpré et al., 2009; Harrison, 1973; Jolicoeur, 1959; Lüpz, 1974; Mertens, 1936; Morey, 2014; Nehring, 1888; Okarma and Buchalczyk, 1993; Ovodov et al., 2011; Pidoplichko et al., 2001; Rütimeyer, 1861; Sablin and Khlopachev, 2002; Studer, 1901; Sumiński, 1975; Wolfgram, 1894), and is reported to be isometrically to stature and size reduction (Losey et al., 2015; Wayne, 1986).

Reduced skull length (TL) of >25% was reported in three first-generation wolves (TL: 161–175 mm) born in a zoo, to wild-caught parents (TL: 225–272 mm) (Nehring, 1888; Wolfgram, 1894), but not in wild wolves, caught as sub-adults and then maintained in a zoo (Wolfgram, 1894). This was considered an indication for size reduction as a result of domestication if pups were part of the anthropogenic environment from early on.

However as skull size (and stature-size) are influenced by naturally-occurring prey preferences in wolves, as well as by geographical and climatological influences (Geffen et al., 2004; Leonard et al., 2002; Muñoz-Fuentes et al., 2009; Perri, 2016; Perri and Sazelova, 2016; Pilot et al., 2006; Pilot et al., 2010; Pilot et al., 2012), relatively large differences in skull length do not necessarily point to wolf-dog distinctions.

Carnassial size reduction (maxillary P4 and mandibular M1) (Table 8, Table 9) has been reported by many authors as a concise signal of domestication (Clutton-Brock, 1962; Lawrence and Reed, 1983; Morey, 1992; Morey, 1994; Morey, 2010; Tchernov and Horwitz, 1991; Tchernov and Valla, 1997). Yet, in the earliest dogs, this phenomenon is allometric when compared to skull size reduction (Clutton-Brock, 1962; Huxley, 1880; Lawrence and Bossert, 1967; Lawrence and Reed, 1983; Morey, 1992; Zeuner, 1963). Early reports report skull size reductions of 25%, with carnassial length reductions of only 5% (Huxley, 1880). Nehring (1888) reported a 25% skull size reduction in three Eurasian wolf siblings, born in the Berlin Zoo, compared to their wild-caught parents, while their carnassials were reduced <18%. The reason for this disproportion is that tooth size is 80% determined by genetic factors contrary to body and skull size (Dempsey and Townsend, 2001; Howe et al., 1983) and thus more conservative. A delayed tooth size reduction, compared to other cranial elements, is likely to complicate early wolf/dog distinctions, as seen with M1 (See below) (Table 9).

Assignment of a specimen to the dog or wolf group may depend on minute differences in the anatomy of a single tooth, with small differences in shape and size of meta-, hypo-, proto-, paracone, protostyle and cingulum (Colyer, 1990). Particularly, the shape of the protocone has been used as a distinctive landmark (Camarós et al., 2016; Napierala and Uerpmann, 2012) (Fig. 6). Based on the presence of a distinctive protocone on P4 in dogs and its (presumed) near-absence in wolves, a new Paleolithic dog from the Aurignacian (Hohle Fels) recently was proposed (Camarós et al., 2016). This specimen shows a prominent protocone on P4 (Fig. 5).

Shorter and wider snouts are the most frequently reported characteristics used to argue for the identity of a dog specimen (Clutton-Brock, 1995; Crockford, 2005; Drake, 2011; Horard-Herbin et al., 2014; Huxley, 1880; Koler-Matznick, 2002; Morey, 1992; Napierala and Uerpmann, 2012; Nehring, 1888; Studer, 1901; Mertens, 1936; Iljin, 1941; Degerbøl, 1961; Germonpré et al., 2009; Stockhaus, 1965; Lawrence and Bossert, 1967; Olsen, 1985; Olsen and Olsen, 1977; Sablin and Khlopachev, 2002; Wayne, 1986). However, not all studies agree and there exists no concise overview to compare the results from all published studies (Table 10).

Shorter snouts in dogs have been subjective measures and co-existing tooth crowding has been attributed to the consequences of shortening (Benecke, 1987; Benecke, 1994; Bökönyi, 1975; Clutton-Brock, 1995; Crockford, 2005; Dayan, 1994a; Degerbøl, 1961; Drake, 2011; Iljin, 1941; Germonpré et al., 2009; Horard-Herbin et al., 2014; Huxley, 1880; Iljin, 1941; Koler-Matznick, 2002; Lawrence and Bossert, 1967; Mertens, 1936; Morey, 1992; Napierala and Uerpmann, 2012; Nehring, 1888; Olsen, 1985; Olsen and Olsen, 1977; Ovodov et al., 2011; Pidoplichko, 1998; Rütimeyer, 1861; Sablin and Khlopachev, 2002; Stockhaus, 1965; Studer, 1901; Zeuner, 1963). Snout length likely is in part genetically determined, but also can be influenced by environment and diet composition (Aron, 1911; Beecher and Corrucini, 1981; McCance and Ford, 1961).

The snout length ratio/index is derived by dividing a possible snout length measure (Von den Driesch (1976) measurements 8, 9, 12, or 13), by skull length (TL, CbL, BL; Von den Driesch, 1976 measurement 1, 2, 3) (Table 12). Most studies use the ratio 8/2. Several authors have calculated relative snout length indexes (Table 10), in total on 1034 modern dogs, 68 archaeological dogs, and 442 modern wolves. Eleven studies measured skull length and/or snout width and length; two of these were GM studies (Rizk, 2012; Schmitt and Wallace, 2012) and nine were classical morphometric studies (Arensburg, 2004; Benecke, 1994; Lüpz, 1974; Morey, 1994; Nehring, 1884; Stockhaus, 1965; Studer, 1901; Tchernov and Valla, 1997; Wayne, 1986). Among existing reports, only Tchernov and Valla (1997) observed shortening, and only in the pre-maxillary region. Lack of snout shortening in dogs also is reported from GM studies (Drake et al., 2015; Rizk, 2012; Schmitt and Wallace, 2012) that suggest partial length reduction, compensated by partial length increase (Table 10).

Snout width is the greatest breadth of the palate (GPB) (Von den Driesch, 1976, measurement 34) divided by skull length. Studies that measured snout width ratio (Table 10) (Morey, 1994; Stockhaus, 1965; Studer, 1901; Tchernov and Valla, 1997; Wayne, 1986) reported almost all wider snouts (Table 10). Only Tchernov and Valla (1997) reported narrower snouts, but used Von den Driesch (1976) measurement 35 (the smallest snout width), not 34. Wider snouts also are supported by GM studies (Rizk, 2012; Schmitt and Wallace, 2012).

In a recent study of Early Holocene, Neolithic canids from Zhokov (Russia), Pitulko and Kasparov (2017) proposed two new skull height ratios to differentiate dogs from wolves (Fig. 6). Both ratios are calculated by dividing a specific nasal or skull height by basal length (BL, measure 3 from Von den Driesch, 1976). The snout height (DD1) is measured at maxillary P1–2. The skull/cranium height (DD2) is measured at the post-orbital constriction at the level of the zygomatic processes of the frontal bone (Fig. 8). The mean DD1 and DD2 reported in the original study were measured in two archaeological dog specimens (DD1 mean 0.19; DD2 0.37). In wolves from the same Neolithic site (n = 24), the measurements were DD1 mean 0.17 (range 0.16–0.18); DD2 0.34 (range 0.32–0.36).

In 2009, dog domestication was proposed to have occurred about 20,000 years earlier than previously assumed, an assertion based primarily on one large canid from Goyet (34 kya) in Belgium (Germonpré et al., 2009). Previous reports also had proposed earlier Paleolithic dog specimens, but these publications had not gained much traction. The latter reports included one specimen from Mezhirich (Mezhyrich) (c. 14.5 kya) (Pidoplichko et al., 2001; Pidoplichko, 1998) and two from Eliseevichi (c. 17 kya) (Sablin and Khlopachev, 2002).

More recently, 6 more insipient or proto-dogs were reported, bringing the total to 10: Hohle Fels (Aurignacian) (Camarós et al., 2016), Goyet, Mezin 5–490 and Mezhirich 4493/24 (Germonpré et al., 2009), Razboinichya (Ovodov et al., 2011), “Předmostí 1069” (also numbered 1063 or 3), “Předmostí 1060”, “Předmostí–” (Germonpré et al., 2012) and Ulakhan Sular (Germonpré et al., 2017) (Table 6, Table 12, Table 13). Another Razboinichya specimen was classified as a Paleolithic dog based on a caudally oriented coronoid process, a shorter and wider snout, P4 length less than M1 + M2, and shorter skull length (TL 211 mm). Several concerns about these specimens have been published, adding emphasis to our concerns.

We propose that these generally accepted criteria require reanalysis based on larger datasets. Here, we address this need for critical re-evaluation by combining new and prior data to create an updated reference framework including several of the older publications, often not in English, that have been poorly consulted by the English scientific community. We explain questions of robustness, our measurements, and the outcomes, and illustrate our findings by reconsidering earlier reports of large Pleistocene canids (Camarós et al., 2016; Germonpré et al., 2009; Germonpré et al., 2012; Germonpré et al., 2015; Germonpré et al., 2017; Ovodov et al., 2011; Pidoplichko et al., 2001; Sablin and Khlopachev, 2002).