Hedgehogs (Erinaceidae, Lipotyphla) from the Miocene of Pakistan, with description of a new species of Galerix

Hedgehogs (erinaceid insectivores) are a common element in Miocene small mammal faunas of Pakistan, but little material has been formally described. Here, we report on extensive collections from numerous localities across Pakistan, most from the Potwar Plateau, Punjab, and the Sehwan area in Sindh. The dominant erinaceid is Galerix, which is also known from Europe, Turkey, and East Africa. We document a new early species of Galerix, Galerix wesselsae, in sites from Sehwan, the Zinda Pir Dome, the Potwar Plateau, and Banda Daud Shah ranging in age from about 19 to 14 Ma. This species briefly co-occurs with and is then replaced by the smaller Galerix rutlandae, previously based on a few teeth from Daud Khel. G. rutlandae is a common element of Chinji Formation faunas through about 11.5 Ma. Scanty material of an apparently new species of Schizogalerix succeeds Galerix. After 10.2 Ma, no galericines occur in Pakistan, but rare erinaceines are later faunal components.


Introduction
on M1 than does G. symeonidisi and the protoconule on M3 is weaker. It differs from G. uenayae in having M1 clearly longer than P4; m2 clearly longer than p4; and the metaconid on p4 well-developed. Galerix wesselsae differs from G. rutlandae in the stronger, more labially directed metastyle crest on M1; less well-developed protoconehypocone connection on M2; more isolated paraconid on p4; and the absence of the anterior cuspule on p2 and p3. p4 (n=9).-Teeth about triangular in shape in occlusal view. The protoconid, by far the largest cusp, lies labial to the middle of the tooth. A small metaconid is present, slightly posterior to the protoconid in two specimens and directly lingual to it in three. The paraconid, located anterior and slightly lingual to the protoconid, is conical and isolated. It is lower than the metaconid in four specimens and as high in two, but more voluminous. The short talonid bears a posterior cingulum, strongest lingually. A metacristid connects the metaconid to the lingual end of the posterior cingulum in four of eight specimens. There is no median ridge. There are two roots. m1 (n=5).-Most specimens are incomplete. The trigonid is about as long and broad as the talonid. The metaconid and protoconid are equally high and are connected by a crest. The metaconid is slightly anterior to the protoconid. The paralophid descends from the anterolabial corner of the protoconid as a narrow, almost transversely-oriented crest.
This crest quickly changes direction and forms a broader, diagonal crest that extends to the anterior edge of the trigonid. The anterior cingulum extends from a point labial to the anterior edge of the tooth to a point anterior to the protoconid; in one specimen, it extends further posteriad towards the hypoconid. Both the hypoconid and the slightly higher entoconid are well-developed, though lower than the trigonid cusps. The cristid obliqua connects the hypoconid to the protoconid and slants slightly lingually. The posterior arm of the hypoconid joins the entoconid. A posterior cingulum connects to the posterior arm of the entoconid (observed in three specimens only). There are two roots. m2 (n=10).-In terms of morphology and root number, m1 and m2 generally resemble each other. In m2, however, the trigonid is slightly shorter than the talonid, the tooth is smaller overall, and the configuration of the paralophid is different. In m2, the paralophid descends from the anterolabial corner of the protoconid as a low crest.
The labial portion is broad and oriented diagonally until the anterior edge of the trigonid. There, it turns posteriad as a narrower crest that ends anterolingual to the metaconid. The metacristid is not fully preserved in any m1, but in m2 it runs from the lingual side of the base of the metaconid to the entoconid and is notched slightly posterior to its anterior end. The posterior cingulum is connected to the entoconid in three of the six specimens that preserve enough of the talonid to observe this character. m1/2 (n=9).-Among nine fragmentary specimens not securely identified as either m1 or m2, the posterior cingulum is connected to the entoconid in four. m3 (n=5).-Of the five specimens, much smaller than m1 or m2, two are worn (one missing part of the paraconid) and two only preserve the trigonid. The trigonid is about as long as the talonid, but broader. The metaconid is the highest cusp in the trigonid and is slightly anterior to the protoconid. The metaconid and protoconid are connected by a low crest. The paralophid descends from the anterolabial corner of the protoconid and forms a broad curve at the anterior margin of the trigonid. The paralophid turns slightly posterior at the anterolingual edge of the trigonid, but does not reach the metaconid. An anterior cingulum is present at the labial side of the trigonid. Both the hypoconid and entoconid are well-developed, though lower than the trigonid cusps. The hypoconid is closer to the trigonid than the entoconid is. The entoconid is higher than the hypoconid. The nearly longitudinal cristid oblique connects hypoconid to protoconid. The metacristid runs from the lingual side of the base of the metaconid to the entoconid. A crest connects the entoconid and hypoconid at the posterior margin of the tooth. There is no posterior cingulum or hypoconulid. There are two roots. C1 (n=2).-High-crowned and with two roots, the tall central cusp is anterior to the middle of the tooth and slightly recurved. The posterior side of the tooth bears a flattening with a small cuspule. P1 (n=2).-The tear-shaped tooth consists of a single large cusp with a posterior flattening; there are two roots.
P2 (n=5).-P2 is larger than P1 and while the labial side is rounded, the lingual face is almost straight. The main cusp is anterior to the middle of the tooth and a flattening lies posterior to it. There are two roots.
P3 (n=7).-Four complete teeth are identified as P3 of Galerix wesselsae, and three fragments preserve only the labial side of the tooth. The complete teeth are generally rectangular in shape and clearly wider than long, except for one relatively longer specimen. A large paracone is present near the anterolabial corner. A metastyle crest extends posteriorly from this cusp and gently turns labially, ending in the posterolabial corner of the tooth. A cingulum extends across the posterior face of the tooth. Two low but distinct cusps, the protocone and hypocone, are present on the lingual side, separated from the paracone by a broad basin. No parastyle is apparent in four of seven specimens.
Two specimens are worn, so that the protocone and hypocone are indistinct. P3 has three roots.
P4 (n=14).-Despite the large number of specimens, only one P4 is complete. One specimen is missing part of the metastyle crest; two lack the parastyle; six preserve only the labial side; and nine only the lingual side. P4 is relatively broad labiolingually and short anterodistally. The paracone is a large cusp, located slightly anterior to the middle of the labial side of the tooth. A parastyle region is present as a small bulge on the anterolabial side of the tooth. In three specimens, this is only a small flattened area; in three others, a low crest is present at the anterior margin of the parastylar area. A strong metastyle crest descends posteriorly from the paracone. It turns sharply labially and ends in the posterolabial corner of the tooth. A sloping, slightly flattened area is present labial and anterior to the metastyle crest. The lingual side of the tooth bears both a protocone and hypocone. The protocone is larger than the hypocone, but both are much smaller than the paracone. A posterior cingulum extends across the posterior margin of the tooth from the posterior end of the metastyle crest; in one specimen, it continues along the lingual margin to a point between the protocone and hypocone, but in ten it stops anterior to the protocone. There are three roots. (n=9).-This tooth is rectangular in overall shape, with a pronounced metastyle crest directed posterolabially.

M1
The four main cusps are arranged in a square. The protocone and metacone are the largest cusps and the hypocone is the smallest main cusp. A crest descends labially from the protocone towards the paracone. No protoconule is apparent on this crest, but a notch is present slightly lingual to the paracone. The paracone is connected to the labial side of the metacone by a low but continuous mesostyle. Anterior and labial cingula are present. Because the posterolabial corner is not well preserved in any specimen, the condition of the metastyle is unclear. The anterior cingulum ends lingually at the anterior side of the protocone. The posterior arm of the protocone splits into two branches, one leading to the hypocone and the other to the metaconule. The former is stronger in six out of seven specimens. The hypocone is a conical cusp, without a posterior arm. The metaconule is a small, triangular cusp with a short posterior arm that ends at about the level of the bend in the metastyle crest. The metaconule is connected to the metacone by a weak crest. The well-developed metastyle first extends posteriad and then sharply bends labially, ending at the posterolabial corner of the tooth. The labial portion of the metastyle crest points towards the hypocone.
The posterior cingulum is present and runs undivided from the posterolabial corner of the tooth to a point posterior to the hypocone. There are three roots, the lingual root deeply grooved on its lingual side.
M2 (n=10).-M2 is generally similar to M1, but the metastyle crest is less well-developed and the metacone is shifted a little lingually, so that the main cusps do not form a perfect square. The anterolabial area is preserved in several specimens; the anterior and labial cingula intersect at the anterolabial corner and a crest connects the paracone to the anterior cingulum. In one specimen the posterior arm of the metaconule continues to the posterolabial corner of the tooth and the posterior cingulum; in seven the metaconule has a short posterior arm, as on M1. The protoconemetacone connection is stronger than the protocone-hypocone connection; one specimen has no protocone-hypocone connection.
M3 (n=12).-M3 is triangular, with protocone, paracone, and metacone, near the apices of the triangle. A crest descends from the protocone to the paracone. It does not reach the paracone, ending anterolingual to it in two specimens, but does reach the paracone in nine others. The crest is slightly thickened near its labial end in four specimens, but there is no distinct protoconule. Low crests do connect the metacone to both the paracone and the protocone. A low cingulum is present on the posterolingual side of the tooth in five specimens. A strong parastyle at the anterolabial corner of the tooth is connected to the paracone by a low crest. An anterior cingulum descends from the parastyle to a point anterior to the protocone. A root supports each of the main cusps.
Discussion.-Most of the galericine materials from the Zinda Pir Dome area and the lower strata of the Potwar Plateau and Sehwan deposits consists of specimens that are at or beyond the upper end of the size range in our large samples of Galerix rutlandae from younger sites in the Potwar Plateau (Fig. 2). We have detected several morphological differences between G. rutlandae and the larger form represented in the older sites, and we believe these differences, together with the difference in size, are sufficient to recognize a new species. The differences between the two species are perceptible across relatively large samples, but subtle enough that the assignment of individual teeth is often debatable (compare our discussion of apparently mixed samples like Seh 82.24, below). This is no different from the situation in Europe, where it is often difficult to assign small samples to any one wellestablished Galerix species.
The assignment of the three complete specimens that we identify as P3s to Galerix wesselsae merits discussion, as these teeth are unlike most Galerix P3s. However, they are not so different from the P3s of some Parasorex and Schizogalerix (e.g., Engesser 1980), and P3 in general is a rather variable tooth in galericines (Engesser 2009;Van den Hoek Ostende 1992). These teeth are smaller than P4s that we assign to G. wesselsae and lack the strong notch seen in the metastyle crest of P4. The only sympatric mammal these teeth could conceivably belong to would be a second species of erinaceid, but the only such species found in any of the localities the P3s come from are Galerix rutlandae (which is known to have a morphologically different P3) and one indeterminate Erinaceidae from Seh 84.26 (see below). The P3s would be quite small for this latter species, and it would be unlikely that the three P3s all represent this rare taxon (known from a single tooth), instead of the more common Galerix wesselsae.
We identify a number of isolated specimens as lower anterior premolars (p1-p3) of Galerix wesselsae on the basis of similarities to their homologues in other Galerix species, such as G. rutlandae andG. uenayae. If these teeth are identified correctly and representative of their tooth positions, p2 was larger than p3 in G. wesselsae, as in most Galerix species and unlike in G. rutlandae. Although the hypocone on P3 and the frequent presence of a connection between the entoconid and the posterior cingulum on m1-m2 are unusual for Galerix, both occur in some species (see G. rutlandae below for discussion; G. symeonidisi and G. iliensis are the only Galerix species that consistently have a hypocone on P3). Because the new species also lacks essential diagnostic features of Parasorex and Schizogalerix (as listed under Galerix rutlandae, below), its assignment to Galerix is secure.

Referred material.-See
Revised diagnosis.-A species of Galerix with the following combination of characters: size very small (comparable to G. saratji); hypocone absent on P3; metastyle crest on M1 weak and relatively posteriorly directed; protocone and metaconule on M1 and M2 almost always connected; connection between protocone and hypocone usually strong; metaconule on M1 and M2 usually with a short posterior arm that does not reach the posterolabial corner of the tooth; p1 two-rooted; p2 and p3 with distinct anterior cuspule; p3 larger than p2; p4 with poorly developed metaconid and with the paraconid pressed closely to the protoconid; entoconid always connected to the posterior cingulum in m1 and usually in m2.
Differential diagnosis.-Galerix rutlandae is smaller than the European and African species G. africanus, G. aurelianensis, G. exilis, G. stehlini, and G. remmerti and differs from all in the invariable presence of a connection between the entoconid and the posterior cingulum of m1. Galerix symeonidisi, G. iliensis, and G. wesselsae are larger and possess a strong hypocone on P3. Further differences with Galerix wesselsae are discussed under that species. The Early Miocene Turkish species G. saratji and G. uenayae are most similar to G. rutlandae, but in both the p2 is larger than the p3 and both of these teeth lack an anterior cuspule. In addition, G. uenayae is larger, with enlarged P4 and p4, and the metaconid on the p4 is weak or absent. In G. saratji, the trigonid of p4 is better developed with more separated protoconid and metaconid.
Distribution.-Galerix rutlandae occurs in Potwar sites of age 14.3 Ma (Y709) to 11.6 Ma (Y504). It is preceded by Galerix wesselsae, with which it co-occurs at a few sites, and followed by Schizogalerix sp. A. Galerix rutlandae is known from its type locality, Daud Khel (Munthe and West 1980), from Dhok Tahlian (Cheema et al. 1996), and from the Banda Daud Shah, Sehwan, and Gaj areas. Cheema et al. (2000) conferred a molar fragment from Jalalpur (ca. 10-11 Ma) to G. rutlandae, but as they note, its assignment is tenuous.
Description.-Mandible (n=7).-The lower jaw of Galerix rutlandae is known only from fragments. Three show that the mental foramen is below the posterior root of p3 and two show that p3 is larger than p2. We refer a fragment preserving a piece of the mandible posterior to m3 to G. rutlandae. It is similar to G. exilis and distinct from Schizogalerix in that the ascending ramus does not rise steeply. The ventral border of the bone turns dorsad below the posterior root of m3. p1 (n=8).-The first lower premolar is an anteroposteriorly elongate tooth. The anterior and posterior ends tend to be lingually displaced. The main cusp is anterior and labial of the middle of the tooth. Anterior to this cusp is a small flattening, which bears a cuspule in two of six specimens, and posteriorly is a larger flattening, bearing a posterior cingulum that is very weak in most specimens, but quite pronounced in one. There are two roots. p4 (n=30).-The fourth lower premolar is larger and more molariform, with a longer talonid and more complex trigonid ( Figure 4). The trigonid is larger than the talonid and dominated by the protoconid. A paraconid is present. It is elongate in shape and diagonal to almost anteroposterior in orientation. The metaconid is small, often appearing as little more than a raising in the lingual flange of the protoconid, but is often about as high as or higher than the paraconid. It is usually directly lingual to the protoconid, but slightly posterior to it in one of 27 specimens. The talonid is present as a low ridge at the back of the tooth, strongest at the lingual side. A low ridge usually connects the metaconid to the talonid at the lingual margin of the tooth. In 9 of 22 specimens, a very low anteroposterior ridge runs anteriad from the posterior cingulum at the center of the talonid. m2 (n=32).-Similar in most respects to m1, m2 is slightly smaller, the talonid and trigonid are about equally broad, and the configuration of the trigonid is different. The paralophid descends from the anterolabial corner of the protoconid as a low crest. The labial portion is broad and oriented diagonally toward the anterior edge of the trigonid.
There, it turns posteriad as a narrower crest that ends anterolingual to the metaconid. The labial cingulum continues posterior to the protoconid in 9 of 19 specimens. The posterior cingulum is connected to the entoconid in 12 of 18 specimens. m1/2 (n=9).-In eight of these isolated talonids of m1 or m2, the entoconid is connected to the posterior cingulum. m3 (n=31).-The m3 is smaller and narrower than the m1 and m2. The trigonid is about as long as the talonid, but broader. The metaconid is the highest cusp in the trigonid. The metaconid and protoconid are connected by a low crest. The paralophid descends from the anterolabial corner of the protoconid and forms a broad curve at the anterior margin of the trigonid. The paraconid is not individually distinguishable on the paralophid. The paralophid turns slightly posterior at the anterolingual edge of the trigonid, but does not reach the metaconid. An anterior cingulum is present at the labial side of the trigonid. Both the hypoconid and higher entoconid are well-developed, though lower than the trigonid cusps. The hypoconid is slightly closer to the trigonid than the entoconid is. The cristid obliqua connects the hypoconid to the protoconid and is almost longitudinally oriented. The metastylid runs from the lingual side of the base of the metaconid to the entoconid. A crest connects the entoconid and hypoconid at the posterior margin of the tooth, but there is no posterior cingulum or hypoconulid. There is one anterior and one posterior root. M2 (n=68).-This molar is similar to the first, but the metastyle crest is much weaker and the metacone is slightly lingual to the paracone, so that the main cusps are arranged in a trapezoid instead of a square. As on M1, the protoconule may or may not be present. The protocone-metaconule connection is present in 51 of 53 specimens. All 51 specimens in which the structure can be observed have a protocone-hypocone connection, but the connection is weak in 10. A short posterior metaconule arm does not reach the posterior cingulum in 36 of 42 specimens, and 6 lack the posterior arm completely. As on M1, the lingual root is deeply grooved.
M3 (n=43).-Triangular in overall shape, with cusps (protocone, paracone, and metacone) at each corner. The posterolabial side tends to be shorter than the posterolingual side; the anterior side is longest. The protocone tends to be the largest cusp. Along the margins of the molar, the cusps are connected by low crests; the crest connecting the paracone and protocone tends to be higher. Anterolabial to the paracone, a small parastyle is present, which forms the labial end of an anterior cingulum. The paracone and parastyle are usually connected by a low crest. A posterolingual cingulum is present in 16 of 27 specimens. A root supports each of the three cusps.
Discussion.-The original description (Munthe and West 1980) of Galerix rutlandae and all subsequent records were based on scanty material. The samples described here are consistent in morphology and size with the hypodigm of G. rutlandae, sharing such features as absence of a hypocone on P3, presence of a protocone-metaconule connection on M1 and M2, a short, relatively posteriorly directed metastyle crest on M1, and a relatively weak metaconid on p4. Munthe and West (1980) show an M2 of Galerix rutlandae with the posterior arm of the metaconule bisecting the posterior cingulum, but re-examination of a cast of this specimen shows that the posterior cingulum is in fact continuous, and the posterior arm of the metaconule ends short of the cingulum. Some measurements given by Munthe and West (1980) are outside the range of our samples. Our measurements of casts of the M1 and M2 from the type sample fall well within the range of variation in our samples, possibly reflecting a difference in measuring techniques (unspecified measurement protocol). Munthe and West (1980) gave the length and width of the one M3 they referred to G. rutlandae as 1.40x1.28 mm, but Galerix M3s are always wider than long. Their figure shows greater width than length, and a cast of this specimen is 1.12x1.50 mm. The 1.32 mm width of an m3 trigonid given by Munthe and West (1980) is well outside the range of variation of our samples and even of the larger Galerix wesselsae. Unfortunately, this specimen was not figured and we do not have access to a cast.
Here, we expand the known material of the species by more than an order of magnitude. For this reason, we characterize the species in detail and provide a revised diagnosis. Although Munthe and West (1980)  For example, the posterior cingulum of m1 and m2 is usually connected to the entoconid in G. rutlandae.
Although Van den Hoek Ostende (2001) listed the usual presence of this trait as diagnostic of Schizogalerix, he noted that it is also seen in some specimens of G. symeonidisi, G. saratji, and G. remmerti (Doukas 1986;Ziegler and Fahlbusch 1986;Van den Hoek Ostende 1992. Interestingly, the connection is more frequently present in m1 than in m2 in G. saratji and G. remmerti, matching the condition in the Pakistani Galerix. Possibly this trait was variable in early species of Galerix and became fixed in later galericines. In G. rutlandae, p3 was larger than p2, as shown by the preserved alveoli of several jaw fragments as well as measurements of isolated teeth, although Van den Hoek Ostende (2001) listed a p2 that is larger than or as large as p3 as a diagnostic character of Galerix. However, p3 is also larger than p2 in Galerix exilis from the type locality of Sansan, France (Engesser 2009). Prieto et al. (2012) questioned the generic assignment of G. rutlandae on the basis of the presence of a paralophid on p4. Although the condition of the p4 in this species is distinct from the more typical Galerix condition seen in G. wesselsae, our material shows that its variable paralophid is not similar to that of Parasorex.
Other features of Galerix rutlandae are characteristic of Galerix and distinguish it from Parasorex, Schizogalerix, or both: the relatively small width/length ratio of the upper molars, the almost invariable presence of a protoconemetaconule connection and a short posterior arm of the metaconule, the lack of anterior displacement of the lingual cusps of m1 and m2, and the lack of a hypocone on P3. Thus, we continue to assign this species to Galerix.

Galerix, sp. indet.
Several assemblages (Seh 82.24 and 81.14, Y709 and Y733, CH O) include specimens with diagnostic features or size that match either G. rutlandae or G. wesselsae. We interpret these assemblages as sampling sympatric populations of both species, rather than a transitional form between G. wesselsae and G. rutlandae, because there is an abnormally large amount of variation in size at several sites. For example, site Seh 82.24 contains both one of the largest m1s (specimen 82.24-14) and one of the smallest M2s (specimen 82.24-56) in our entire collection. Also, the older G. wesselsae appears to be more derived than G. rutlandae in some morphological features, rendering an ancestor-descendant relationship unlikely. Both species occur at localities Y709 and 733, 14.3-14.0 Ma, and we suspect that all of these sites are of similar age.
The two Pakistani species of Galerix are similar enough that it is not possible to securely identify all material from these sites. The smallest specimens are likely G. rutlandae, the largest can be referred to G. wesselsae, and some specimens have features that enable referral to a species, but others are in a size range of overlap for either or belong to tooth positions in which the species do not differ detectably.
Discussion.-Seh 82.24.-This richest site in our sample, with over 60 erinaceid teeth, includes both specimens unambiguously referable to G. wesselsae (e.g., P3 with hypocone) and some with features diagnostic of G. rutlandae (p3s with anterior cuspule). Overall, G. rutlandae appears to be the more common species in this site by a factor of two to one, but a large proportion of the material cannot be identified to species.
We refer the single p1 from this site to G. wesselsae because of large size. Among three p3s, we refer two to G. rutlandae and one to G. wesselsae on the basis of the presence of an anterior cuspule. The latter is also the largest of them. Among the five p4s from the site, one is small and shows the elongate paraconid characteristic of G. rutlandae, while two larger ones have a paraconid more reminiscent of G. wesselsae. Among six m1, we allocate a virtually complete tooth and an isolated trigonid, both quite large, to G. wesselsae and two small trigonids to G. rutlandae. The site has yielded seven m2s, none completely preserved. Two small teeth are G. rutlandae, a larger is G. wesselsae, and four others (three with trigonid only) are G. sp. Six specimens represent m3, and the size difference between a complete tooth referable to G. wesselsae and another we identify as G. rutlandae is striking.
The single P3 from Seh 82.24 bears a hypocone and is referable to G. wesselsae. Two of 8 complete P4 have the labiolingually broader shape characteristic of G. wesselsae and two small ones are similar to G. rutlandae; identification of the others is left open. The lone M1 fragment from Seh 82.24 is not identifiable to species level, and a small M1 or M2 fragment is almost certainly G. rutlandae. M2 is represented by an impressive 14 specimens. While morphological differences of G. rutlandae and G. wesselsae M2s are too subtle to be of much help, size differences between the two species are pronounced in this tooth position, and we identify 12 specimens as G. rutlandae. Of five M3s from Seh 82.24, two are noticeably smaller and relatively broader, and the paracone is narrower when observed in anterior view, so we refer these to G. rutlandae and the others to G. wesselsae.
CH O.-Four Galerix teeth include a small lingual fragment of P3 bearing a single cusp and a posterior cingulum (CH O 269), which is definitely not referable to G. wesselsae, but also one of the largest M3s found anywhere in Pakistan (CH O 268), which is likely to be G. wesselsae. Of the other two specimens, a large m1 trigonid (CH O 270) is consistent with G. wesselsae, but a lingual P4 fragment (CH O 263) bearing two worn cusps is considered G. rutlandae.
Y709.-This relatively rich site, with about 30 teeth referable to G. rutlandae on the basis of size, includes four quite large teeth likely representing G. wesselsae. A large P4 is like the P4 of G. wesselsae in shape. A complete M2 is too large for G. rutlandae and an M2 fragment appears to be even larger. The single large m1 recovered from Y709 also probably represents G. wesselsae.
Y733.-Most of the 20 Galerix teeth from this site appear to be G. rutlandae on the basis of size and morphology as at Y709, but a very large m1 probably represents G. wesselsae.
Seh 81.14.-This site is likely older (Wessels 2009) than the others, and the material is predominantly G. wesselsae. One worn and abraded M2, 81.14-4718, stands out because it is substantially smaller than two betterpreserved M2s from the same locality, 81.14-4715 and -4716. The protocone is connected to both the metaconule and the hypocone, and the posterior arm of the metaconule is short. The cingula are weak and the parastyle is missing, possibly due to abrasion. We believe this specimen provides evidence for a second smaller species of Galerix here.
Description.-p3 (YGSP 24491).-This tooth is dominated by a large central cusp. A small knob is pressed against the anterolingual face of the central cusp, but there is no true anterior cuspule. A short talonid is closed labially by a low crest. There are two roots.
p4 (YGSP 24490).-This is a wide tooth, especially posteriorly. The trigonid is dominated by a triangular protoconid. The small metaconid is tightly apressed on the anterolingual side of protoconid. The paraconid is elongate and relatively anteroposteriorly oriented. A paralophid connects the protoconid to the paraconid. The talonid is short, and the posterior cingulum is reduced to a short, low crest posterior to the metaconid. Discussion.-The M2 YGSP 24492 can be assigned to Schizogalerix on the basis of the following traits (Van den Hoek Ostende 2001;Engesser 1980): molar short and wide, with width/length ratio of 1.51; no connection between protocone and metaconule; posterior arm of M2 metaconule extends to posterolabial corner of tooth; mesostyle displaced labially ("Schlaufenbildung"). The well-developed protoconule also appears to be characteristic of Schizogalerix. The specimen is unfortunately missing part of its anterolabial corner, rendering the condition of the mesostyle uncertain, but it appears that although displaced almost to the labial margin of the tooth, the mesostyle was weakly divided, similar to the situation in the primitive Schizogalerix species S. evae, S. pasalarensis, and S. voesendorfensis (Engesser 1980;De Bruijn et al. 2006). The posterior portion of the mesostyle is developed into a cuspule, which distinguishes the tooth from S. voesendorfensis, in which the posterior part of the mesostyle is a curved crest and the metacone is closer to the labial margin of the tooth (Engesser 1980;Kälin and Engesser 2001;Prieto et al. 2010). Unlike in S. pasalarensis, there is no cingulum labial to the metacone and the posterior part of the mesostyle reaches the labial margin of the tooth (Engesser, 1980; M2 is unknown in the more primitive S. evae). The M2 is medium in size for the genus, comparable to S. voesendorfensis among other species (measurements in Engesser 1980;Rzebik-Kowalska and Lungu 2009).
Although incomplete, YGSP 24492 shows traits distinct from those of all other species of Schizogalerix; the closest similarities appear to be with the Central European S. voesendorfensis. In view of the small available sample from Y259 and similarly aged sites, the attributions of other teeth to the same species as YGSP 24492 must remain tentative (see below). If these teeth are identified correctly, they provide additional characters that are diagnostic of the Pakistani species of Schizogalerix, but the samples are too small to exclude the presence of additional galericine species in these sites. In view of the scanty material, this species of Schizogalerix is not named here, and we provisionally refer to it as Schizogalerix sp. A.
YGSP 24490, a p4, differs from p4s and other lower premolars of G. rutlandae in being relatively short and wide, with a talonid that is much wider than the trigonid, and in lacking a distinct metaconid. Although its length is similar to the average for G. rutlandae, it is relatively wide. Its general shape resembles that of the p4 of Schizogalerix, but the metaconid is usually well-developed in that genus. However, an indistinct metaconid is also seen in several apparently unrelated species of Galerix (Van den Hoek Ostende 2001) and even in Schizogalerix voesendorfensis (Engesser 1980, fig. 3d). We tentatively refer the tooth to Schizogalerix sp. A.
The premolar YGSP 24491 resembles p2s of Galerix rutlandae, but is longer and broader and has a less distinct anterior cuspule. It is morphologically similar to anterior premolars of Schizogalerix, particularly p3 (Engesser 1980), and is probably referable to S. sp. A. Another premolar from Y259, YGSP 39486, resembles the P2 of G. rutlandae, but is somewhat broader; it may be a P2 ofSchizogalerix sp. A. A similar though substantially larger tooth from site Y450, YGSP 24459, may be another Schizogalerix P2.
In talonid YGSP 39487, the entoconid is connected to the posterior cingulum. This trait is generally diagnostic of Schizogalerix according to Van den Hoek Ostende (2001), but is also seen in some Galerix symeonidisi, G. remmerti, G. rutlandae, and G. wesselsae (see discussion above). The specimen is similar in size to specimens of G. rutlandae, but differs in a smaller angle between the anterior and posterior arms of the hypoconid, which are almost perpendicular in G. rutlandae, but form an acute angle in YGSP 39487. Some species of Schizogalerix show a similar morphology, especially in m2 (e.g., S. pasalarensis ;Engesser 1980). With a width of only 1.48 mm, the specimen is among the narrowest known Schizogalerix m1s or m2s (Engesser 1980;De Bruijn et al. 2006;Rzebik-Kowalska and Lungu 2009;Kälin and Engesser 2001). Unlike derived Schizogalerix species such as S. zapfei, S. sinapensis, and S. macedonicus, YGSP 39487 does not show a free-ending posterior arm of the hypoconid (Van den Hoek Ostende 2001); nor does the posterior arm extend into the posterolingual corner of the tooth, as in S. duolebulejinensis (see Bi et al. 1999). Since the narrowest m2s of S. voesendorfensis are similar in size to YGSP 39487 (Kälin and Engesser 2001;Prieto et al. 2010) and the specimen is morphologically similar to Schizogalerix, this specimen is most likely referable to Schizogalerix sp. A. Because m2 is appreciably narrower than m1 in Schizogalerix (Engesser 1980; Rzebik-Kowalska and Lungu 2009), the specimen is in that case likely an m2.
YGSP 39496 appears to represent the lingual side of the M1 or M2 of an erinaceid. Though fragmentary and apparently abraded, its morphology and size generally match YGSP 24492.
YGSP 49004 and 34946, labial fragments of P4, preserve few distinctive characters, but their non-concave labial faces appear to resemble S. voesendorfensis and S. duolebulejinensis (see Engesser 1980;Bi et al. 1999). Both teeth are larger than the largest P4s of Galerix rutlandae and more robust than more comparably sized P4s of Galerix wesselsae. They differ from both in that the posterolabial bulge is narrower; the posterior and labial margins meet each other at an acute angle, instead of being about perpendicular as in Galerix.
YGSP 21833 appears to be a galericine D3. To our knowledge, no D3 of Schizogalerix has been described or figured before, although Prieto et al. (2010) listed measurements for a Schizogalerix voesendorfensis D3 from Gratkorn, Austria. YGSP 21833 is from site Y76, which is intermediate in age between the last occurrence of Galerix rutlandae at site Y504 (11.6 Ma) and the possible occurrence of Schizogalerix at site Y791 (11.2 Ma). However, it is

Previous studies
There are few published records of fossil insectivores from Pakistan. Munthe and West (1980) identified a handful of teeth from Daud Khel in the Chinji Formation as Galerix rutlandae, Amphechinus kreuzae, cf. Echinosorex sp., an indeterminate soricid, and an indeterminate talpid. The "talpid", a fragment of an upper molar, likely represents a treeshrew and we see little resemblance of the three premolars called cf. Echinosorex sp. to living Echinosorex. Munthe and West (1980) included four teeth (P3, M1, i2, m3) in the hypodigm of Amphechinus kreuzae. The labial cingulum, narrow lingual lobe, centrally placed paracone, and weak protocone of P3 suggest a treeshrew. The other specimens, while representing an erinaceine, are questionably allocated as only cranial characters distinguish Amphechinus (Ziegler 2005a). Engesser (1980) considered the presence of a metaconule on M1-2 (seen in Amphechinus kreuzae) to be characteristic of Mioechinus, but this character is inconsistent according to Ziegler (2005a). For now the Daud Khel erinaceine could be referred to as "Amphechinus" kreuzae. Cheema et al. (1996) described a few teeth of Galerix rutlandae from locality PMNH 8608, also in the Chinji Formation. They saw similarity to Galerix symeonidisi Doukas 1986 from the Early Miocene of Greece, but these resemblances are genus level diagnostic characters of Galerix. The teeth match Galerix rutlandae, except for an unusually broad m3.

Age
The samples from the Potwar Plateau, which represent the majority of the material studied here, have been securely dated on the paleomagnetic timescale of Gradstein et al. (2004) and range in age from 17.8 Ma (Y721, Y747) to 7.3 Ma (Y931). The Zinda Pir Dome area localities Z122, Z124, and Z167 yield erinaceids and occur low in or just below the Vihowa Formation (~19 Ma, Lindsay et al. 2005). Ages of other studied localities are not as well However, Galerix symeonidisi may have preferred a wetter habitat (Van den Hoek Ostende and Doukas 2003) and opinions differ on the ecological preferences of Parasorex socialis (Ziegler 2005a;Prieto 2007). Thus, it is prudent to regard the habitat preferences of Pakistani galericines as unknown.
The appearance of Galerix rutlandae at about 14.3 Ma coincided with an increase in squirrel diversity and appearance of dormice (Flynn and Wessels 2013), perhaps suggesting that G. rutlandae preferred closed, moist habitats. At the time of the final disappearance of galericines in Pakistan, after around 10.2 Ma, C4 grasses first appeared in the Siwaliks, and with a more seasonal climate (Barry et al. 2002). However, other forest-adapted taxa, including treeshrews, lorises, and rhizomyines like Miorhizomys, survived another two million years (Flynn 2003).