PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, NY 10024 Number 3548, 22 pp., 12 figures, 12 tables January 15, 2007 Geographic Variation of Idiurus (Rodentia: Anomaluridae) with Emphasis on Skull Morphometry ANJA C. SCHUNKE1 AND RAINER HUTTERER2 ABSTRACT The geographic variation of skull size and shape among populations of flying squirrels of the genus Idiurus was analyzed with multivariate techniques using up to 57 craniometric characters. The variability of body size and fur coloration was also studied. The results support the present division of the genus into two species, Idiurus macrotis and I. zenkeri. Further division, into subspecies, does not seem to be warranted at present, although distinguishable geographical populationsexist in West, Central, andEast Africa. INTRODUCTION Seven species of Anomaluridae are current- ly recognized (Dieterlen, 1993, 2005) and Anomaluridae are small to large rodents placed in three or four genera: Anomalurus restricted to equatorial Africa between lati- (which occasionally contains Anomalurops), tudes 15uN and 15uS. All except one species, Zenkerella, and Idiurus. Zenkerellainsignis,areabletoperformgliding Most current authors agree that Idiurus flight. All anomalurids live in rain forest, consistsoftwospecies,I.macrotisandI.zenkeri montane forest, or gallery forest. High trees (Misonne, 1971; Rahm, 1988; Nowak, 1991; provide takeoff and landing sites for glides Dieterlen, 1993, 2005). They are the smallest and provide hollows for the daytime rest. As members of the family, with head and body anomalures are nocturnal and seldom main- lengths ranging from about 95 mm in I. tained in captivity, very little is known about macrotis to about 75 mm in I. zenkeri. These their biology (Durrell, 1952; Rahm, 1969; animalshaveaglidingmembranesupportedby Schlitter 1989; Julliot et al., 1998). a cartilaginous rod extending from the elbow 1ZoologischesForschungsmuseumAlexanderKoenig,53113Bonn([email protected]). 2Division of Vertebrate Zoology (Mammalogy), American Museum of Natural History; and Zoologisches ForschungsmuseumAlexanderKoenig,53113Bonn([email protected]). CopyrightEAmericanMuseumofNaturalHistory2007 ISSN0003-0082 2 AMERICAN MUSEUMNOVITATES NO. 3548 andalong,sparselyhairedtailthatconsiderably distribution of the data was tested with exceedstheheadandbodylength. a Kolmogorov-Smirnov test independently A preliminary analysis of the distributions forbothspecies.Forthistest,specimensfrom of both species (Schunke and Hutterer, 2000) the northeastern part of the Democratic revealed that large continuous distributions Republic of Congo were used because the across the Central African forest belt, as largest numbers of specimens were available postulated by some authors (e.g. Kingdon, from this area. The same subsets were also 1997), were not supported by specimen-based testedforsexualdimorphismwitht-tests(with distribution maps, which in contrast suggest the exclusion of specimen of unknown sex). patchy distributions for both species. For the main analysis, the remaining data Geographicalisolationmaycausegeneticdrift were submitted to a principal component andresultinspeciationofpopulationsofthese analysis (PCA). Two PCAs were conducted. arboreal rodents. To test this hypothesis, we The first was based on the ln-transformed have analyzed the morphological variation absolute measurements and thus included among geographical populations of Idiurus variation in size. In the second analysis, linear species across their entire range from Sierra regression lines of individual ln-transformed Leone in the west to Tanzania in the east. measurements on ln-transformed palatilar length (PL, fig. 1) were calculated, and resi- SPECIMENS AND METHODS duals of the individual measurements on this line were used instead of original variables Specimens in 14 museum collections form (Thorpe and Leamy, 1983) to compare only the basis of our report. In total, we examined shape differences. Palatilar length was chosen 116 Idiurus macrotis and 83 I. zenkeri (appen- asreferencebecauseofitsrelativelyhighvalues dix 1).Ofthese,craniometricdatatakenfrom (and thus relatively low measuring error) and 63 skulls of I. macrotis and 51 skulls of I. foritsavailabilityfrommostspecimens. zenkeri were used for the calculations. Only In addition, a PCA was performed for skulls with all molars completely emerged standard body measurements (collectors’ field were used in order to exclude juveniles. From data), including total length, tail length, ear each skull, up to 57 measurements were taken length,andhind-footlength(appendix 2).Total with an electronic caliper to the nearest length was taken as combined tail length and 0.01 mm (fig. 1, table 1). head and body length, where necessary. Resi- Variationincolorwasassessedinasampleof dualswerecalculatedforthehind-footlength. 55skinsofI.macrotisand41skinsofI.zenkeri; Aftergroupingthepopulationsaccordingto colorslidesshowingtheskinsalongwithacolor the results of the PCA (see later), a discrimi- referencewereusedforcomparison. nant analysis was conducted to specify differ- The specimens originated from the entire ences between groups. ranges of the two species (figs. 2–3). We As Idiurus skulls are very fragile, it was defined six geographical clusters and used rarely possible to take all measurements from them as operational units: a single skull. Therefore, the PCA was conducted first with all characters and then 1. WA: West Africa (Sierra Leone, Ivory repeated after a stepwise exclusion of mea- Coast, and Ghana; I. macrotis only). surements that were unavailable for several 2. NWC: Northwestern Cameroon. specimens. While the number of specimens 3. CEG: Southern Cameroon, Equatorial increased,thenumberofcharactersdecreased. Guinea, and Gabon. By comparing the results, it was possible to 4. CAR: Central African Republic (I. test their robustness. zenkeri only). Examples of typical results with intermedi- 5. DRC: Democratic Republic of Congo. ate numbers of characters and specimens are 6. TAN: Tanzania (I. macrotis only). shown in scatter plots for the first and second The two species were analyzed separately. principal component, or in box plots. Measurements were converted to logarithms Calculationswereprocessedwiththeprogram before entering the calculations. The normal SPSS (version 10.0). 2007 SCHUNKE AND HUTTERER: GEOGRAPHIC VARIATION OFIDIURUS 3 Additional statistical details are given in Within I. macrotis, four clusters were formed appendix 3, tables 8–12. bythefirstandsecondprincipalcomponentfor logarithmized absolute data (fig. 4, table 8). INSTITUTIONAL ABBREVIATIONS Thefirstfactorcodesmainlyforsizedifferences. ThesmallestspecimenscamefromWestAfrica AMNH American Museum of Natural andfromnorthwesternCameroon,whilespeci- History,New York BMNH British Museum of Natural History, mens from southern Cameroon, Equatorial London Guinea, and Gabon were considerably larger. FMNH Field Museum of Natural History, Animals from northeastern Democratic Chicago Republic of Congo were intermediate in size, MNHN Muse´umNationald’HistoireNaturelle, withaslightoverlapinoneorbothdirections.A Paris singlespecimenfromTanzania(NMW31089)is MRAC Museum Royal d’Afrique Centrale, larger than average values of series from Tervuren southern Cameroon to Gabon but does not NHMB Naturhistorisches Museum, Basel exceedtherangeofvariationinmostcharacters. NMK NationalMuseumsofKenya,Nairobi NMNH NationalMuseumofNaturalHistory, Unfortunately, no other skulls from the Washington,DC Tanzanianpopulationareavailable. NMW Naturhistorisches Museum, Vienna The second principal component demon- NRM Naturhistoriska Riksmuseet, Stock- strates differences in shape rather than size. It holm separates specimens from West Africa and RMNH Nationaal Natuurhistorisch Museum northwestern Cameroon, on one hand, and (Naturalis),Leiden animals from the Democratic Republic of SMF ForschungsinsitutSenckenberg,Frank- Congo andsouthernCameroontoGabon,on furt the other. SMNS StaatlichesMuseumfu¨rNaturkunde, Stuttgart Calculationsusingtheresidualsforpalatilar ZMB Museum fu¨rNaturkunde, Berlin length gave basically the same pattern but were less pronounced than those obtained RESULTS from logarithmized absolute measurements. Here the single skull from Tanzania clustered NORMAL DISTRIBUTION in most plots within specimens from southern Cameroon to Gabon. Allmeasurementswerenormallydistributed Results for I. zenkeri are less pronounced in both species (p . 0.05). (fig. 5, table 9). No size differences between specimens from different localities could be SEXUAL DIMORPHISM detected with the first principal component. With the second principal component, a sepa- The possibility of sexual dimorphism was checked independently for logarithmized mea- ration (with some overlap) was possible surementsandforresidualsforpalatilarlength. between individuals from southern Came- Thirteencharactersdemonstratedsignificant(t- roonandEquatorialGuineaandnortheastern test,p#0.05)intersexualdifferencesforatleast Democratic RepublicofCongo. Theholotype one of the species. Although preliminary tests of zenkeri from Cameroon (BMNH 48.885) withallcharactersyieldedsimilarresults,those has a slightly smaller skull than most other characters were not used in the final analyses. conspecifics from other localities. Further Their exclusion made it possible to use all reduction of the number of characters dem- specimens in a single analysis and to include onstrated that a second specimen (BMNH skulls of specimens with unknown sex to 48.886)fromCameroonwasevensmaller(not maximizethesamplesize. shown in fig. 5). From the Central African Republic only two skulls were available, one slightly and the other considerably damaged. PRINCIPAL COMPONENT ANALYSES Withastronglyreducednumberofcharacters, After the removal of 13 characters, the they clustered either with the Cameroon and remaining 44 characters underwent a PCA. Equatorial Guinea population or held an 4 AMERICAN MUSEUMNOVITATES NO. 3548 TABLE1 Cranial andDentalMeasurements Takenfrom SkullsofIdiurus macrotisandI. zenkeri 0: characternotusedin final PCA;1 and2: characterused in analyses;2: characterused in figs. 4to 7. Abbreviationsas in fig.1. Code Measurement I.macrotis I.zenkeri Bothspecies TL Totallength 0 0 0 CIL Condylo-incisivelength 1 1 1 BL Basallength 2 1 1 ZB Zygomaticbreadth 1 1 1 RB Breadthofrostrum 2 1 1 NB Breadthofnasalbones 0 0 0 IOB Interorbitalbreadth 2 2 2 BCB Breadthofbraincase 2 1 1 PL Palatilarlength 2 2 2 PPBL Postpalatinebasallength 1 1 1 DL Lengthofdiastema 0 0 0 FIL Lengthofincisiveforamen 1 2 1 FIB Widthofincisiveforamen 2 2 2 BP4 GumwidthatP4 2 2 2 BM1 GumwidthatM1 2 2 2 BM3 GumwidthatM3 0 0 0 ChB Widthofchoana 2 2 2 BullaL Lengthofbulla 2 2 2 BullaB Widthofbulla 2 2 2 RH Heigthofrostrum 2 2 2 ZH Heightofzygomaticarc 0 0 0 PFZH Heightofprocessusfrontaliszygomatici 0 0 0 SkH Heightofskull 0 0 0 ZP Positionofzygomaticarc 0 0 0 BCH Heightofbraincase 0 0 0 PZB Widthacrossposteriorendsofzygomaticarcs 2 2 2 FIOH Heightofinfraorbitalforamen 2 2 2 FIOB Widthofinfraorbitalforamen 2 2 2 RUL Upperlengthofrostrum 2 2 2 RLL Lowerlengthofrostrum 2 2 2 ML Lengthofmandible 2 2 2 MHA Heightofmandibleatarticularprocess 2 2 2 MHD Heightofmandibleatdiastema 0 0 0 PAA Distancebetweenarticularandangularprocess 2 1 1 PCH Heightofcoronoidprocess 2 2 2 UCL Uppercrownlengthofmolars 1 1 1 UIL Lengthofupperincisor 2 2 2 UIB Widthofupperincisor 2 2 2 UP4L LengthofupperP4 1 1 1 UP4B WidthofupperP4 1 1 1 UM1L LengthofupperM1 0 0 0 UM1B WidthofupperM1 1 1 1 UM2L LengthofupperM2 1 1 1 UM2B WidthofupperM2 1 1 1 UM3L LengthofupperM3 1 1 1 UM3B WidthofupperM3 1 1 1 LCL Lowercrownlengthofmolars 2 1 1 LIL Lengthoflowerincisor 2 2 2 LIB Widthoflowerincisor 2 2 2 LP4L LengthoflowerP4 1 1 1 LP4B WidthoflowerP4 1 1 1 2007 SCHUNKE AND HUTTERER: GEOGRAPHIC VARIATION OFIDIURUS 5 TABLE1 (Continued) Code Measurement I.macrotis I.zenkeri Bothspecies LM1L LengthoflowerM1 1 1 1 LM1B WidthoflowerM1 1 1 1 LM2L LengthoflowerM2 0 0 0 LM2B WidthoflowerM2 1 1 1 LM3L LengthoflowerM3 0 0 0 LM3B WidthoflowerM3 1 1 1 intermediate position, depending on the num- length of mandible, and distance between ber of characters and individuals included in articular and angular process) yielded 94.2% the respective analyses. correct classifications (tables 2, 11). However, Calculationswithresidualsgaveremarkably the results for specimens from northwestern similar results to the ones obtained with Cameroonmustbetakenwithcaution,asonly logarithmized absolute measurements. five skulls were available from this area. Size differences between I. macrotis and I. The discriminant analysis gave similar zenkeri were pronounced, with a considerable results for I. zenkeri populations, with 94.9% gap in the first principal component (fig. 6, correctly classified individuals (tables 3, 12). table 10);theywerestronglysupported bythe The characters used were height of rostrum, majority of characters. width of bulla, and upper and lower rostrum An interesting result was obtained from length. Unfortunately, only specimens from a combined analysis for I. macrotis and I. southern Cameroon plus Equatorial Guinea zenkeri using residuals for palatilar length and the Democratic Republic of Congo (fig. 7, table 10). While the first principal yielded a sufficient number of measurements component (not shown) mainly separated I. and could be used in this analysis. macrotis from northwestern Cameroon and I. macrotis and I. zenkeri may be separated the southern Cameroon to Gabon region, but from each other by a single character, such as less clearly animals from West Africa and the height of rostrum, basal length, or palatilar Democratic Republic of Congo, the analysis length. Each of these characters correctly yielded no information as to I. zenkeri. classified 100% of the specimens in a discrim- However, the second principal component inant analysis. divided the specimens in individuals from Tables 4and5providemeansandrangesof Democratic Republic of Congo and animals five selected skull measurements of the differ- from southern Cameroon to Gabon, with ent geographical populations of each species, a considerable overlap between I. macrotis and figures 8–10 show examples of crania of and I. zenkeri from the respective areas. The both species. Skulls of all holotype specimens charactersdemonstratinghighloadingsatthis are figured, with the exception of kivuensis principal axis were length and width of the Lo¨nnberg, 1917, the skull of which is lost. bulla, upper length of rostrum, and two incisor characters. BODY SIZEAND SKIN CHARACTERS DISCRIMINANT ANALYSIS A PCA for total length, tail length, ear length, and hind-foot length (field measure- Most individuals of the clusters obtained ments) yielded no separation between I. through the PCA analysis could be identified macrotis and I. zenkeri apart from size with only few characters. More than 90% of differences (result not shown). The statement specimenswereclassifiedcorrectly.Inpopula- of Kingdon (1997: 179) that the tail of I. tions of I. macrotis, a combination of four macrotis ‘‘is proportionately shorter’’ than characters (height of rostrum, length of bulla, that of I. zenkeri was not supported by the 6 AMERICAN MUSEUMNOVITATES NO. 3548 Fig.1. Definitionoflinearmeasurementstakenfromskulls,asexemplifiedbyaskullofIdiurusmacrotis (ZMB 22885). Greatest length and width of single teeth (not marked in figure) were also taken. For abbreviations,seetable 1. Thescale for the dentition represents1 mm andfor the skull, 10mm. datausedinouranalysis(fig. 11).Thetailsof from all areas strongly overlapped. I. zenkeri bothspeciesdifferinabsolutelength(tables 6, was not represented by sufficient geographic 7), but the relative length is about the same. samples, so no conclusions were obtained for Whenlogarithmizedabsolutemeasurements this species. ofI.macrotisaswellasresidualsforhind-foot Noweightdatawereavailableforanyofthe lengthwere used forthe PCA, then specimens 80specimens ofI.zenkeri examined. Kingdon 2007 SCHUNKE AND HUTTERER: GEOGRAPHIC VARIATION OFIDIURUS 7 Fig.2. DistributionofI.macrotisbasedonspecimensexaminedandliteraturerecords.Closedsymbols, specimensexamined;opensymbols,datatakenfromliterature(Sanderson,1940;Kuhn,1965;Rahm,1966; Adam,et al.1970; Aellen et al.,1970; Julliotet al.,1998). Scalerepresents1,000 km. Fig.3. DistributionofI.zenkeribasedonspecimensexaminedandliteraturerecords.Closedsymbols, specimens examined; open symbols, data taken from literature (Rahm, 1966; Jones, 1971; Delany, 1975). Scalerepresents1,000 km. 8 AMERICAN MUSEUMNOVITATES NO. 3548 Fig. 4. Geographic variation of I. macrotis according to the first and second principal components as obtainedfromlogarithmizedabsoluteskullmeasurements.Forcharacters,seetable 1.Opensymbols,males; closedsymbols,females;symbolswithacentraldot,sexunknown;diamonds,WestAfrica;uprighttriangles, northwesternCameroon;trianglespointingupsidedown,southernCameroontoGabon;circles,Democratic Republicof Congo;star,Tanzania. (1997) gives a range of 14–17.5 g for this DISCUSSION speciesand arangeof 25–35 gforI.macrotis. Weight data for 14 museum specimens of I. The taxonomic history of the genus Idiurus macrotis (table 6) range from 23 g to 40 g. is brief. Matschie (1894) described the genus The coloration of the studied skins ap- and species Idiurus zenkeri based on a single pearedverysimilarinbothspecies,incontrast specimen from southern Cameroon (fig. 5). to other anomalurid species (Schunke and Subsequently,Miller(1898)namedI.macrotis Hutterer,2005).Inbothspecies,thedorsalfur from the same region based on differences in is grayish brown to dark brown, and the sizeandcoloration(‘‘MuchlargerthanIdiurus ventral fur is slightly paler. The skin of the zenkeri Matschie; tail and ears relatively limbs and the tail is lighter colored than the longer; color apparently darker; skull larger; dorsal fur. I. macrotis specimens from Ghana bony palate narrower; second lower molar seem to be lighter and more yellowish, but distinctly larger than first’’). Lo¨nnberg (1917) individuals from Ivory Coast are darker and named Idiurus zenkeri kivuensis from Congo more brownish, as are most specimens from (‘‘These specimens agree much more in their Cameroon and the Congo. In I. zenkeri no general characteristics with Idiurus zenkeri differences could be found between popula- Matschie than with I. macrotis Miller. tions, except for the holotype of I. zenkeri, Especially the cranial dimensions differ from which has an entirely blackish skin and dark those of I. macrotis’’). The taxon was sub- brown fur. sequently given species status by Hayman 2007 SCHUNKE AND HUTTERER: GEOGRAPHIC VARIATION OFIDIURUS 9 Fig.5. VariationofI.zenkeriaccordingtothefirstandsecondprincipalcomponentsasobtainedfrom the logarithmized absolute skull measurements. For characters, see table 1. Open symbols, males; closed symbols, females; symbols with a central dot, sex unknown; upright triangles, northwestern Cameroon; triangles pointing upside down, southern Cameroon and Equatorial Guinea; square, Central African Republic; circles,Democratic Republicof Congo. (1946) or considered a synonym of I. zenkeri paler throughout, including the basal fur’’. (Verheyen,1963),whileDieterlen(1993)listed Hayman (1946), in a review of the genus, it under I. macrotis, a conclusion that we accepted I. macrotis and I. zenkeri as species, share. A second, much smaller specimen but also considered I. kivuensis a species (‘‘In reported by Lo¨nnberg was considered a juve- my view the conclusion must be drawn that nile but may well be a specimen of I. zenkeri. there are only three species of Idiurus, namely Unfortunately the skull of the holotype of zenkeri,macrotisandaspeciesintermediatein kivuensisislost(Verheyen,1963),andtheskull size between these two, for which the name of the smaller specimen is still inside the skin kivuensis is available’’). He also named a new and not accessible. However, the general size subspecies I. kivuensis cansdalei (‘‘Differing of the holotype and the single skull measure- from I. k. kivuensis in the pale general color ment provided by Lo¨nnberg (1917) led us to above and below, and from I. k. panga in the the conclusion that kivuensis is rather a syno- general color of the hair tips above being nym of macrotis than of zenkeri. Allen (1922) between avellaneous and wood brown of described I. langi and I. panga as new species Ridgway instead of light drab’’). Finally, fromthenortheasternDemocraticRepublicof Verheyen (1963) named Idiurus zenkeri hay- Congo.I.langiwasdiagnosedas‘‘[s]izeofand mani from northwestern Cameroon: ‘‘Cette proportions nearly as in Idiurus macrotis nouvelle sous-espe`ce se distingue de la forme Miller, but very different in coloration’’, and typique par la longueur totale du corps, la I. panga as ‘‘[s]imilar to Idiurus macrotis longeur de l’oreille, la plus grande longeur du Miller, but much smaller and considerably craˆne et par la largeur bizygomatique nette- 10 AMERICAN MUSEUMNOVITATES NO. 3548 Fig.6. SizevariationofI.macrotisandI.zenkerifromthefirstprincipalcomponentasobtainedfrom logarithmized absolute skull measurements. Forcharacters, seetable 1. Fig. 7. Shape variation of I. macrotis and I. zenkeri from the second principal component as obtained from residuals calculatedfor palatilarlength. Forcharacters, seetable 1.