FULL PAPER Anatomy Proportion and Cluster Analyses of the Skull in Various Species of the Tree Shrews Hideki ENDO1), Tsutomu HIKIDA2), Loke Ming CHOU3), Katsuhiro FUKUTA4) and Brian J. STAFFORD5,6) 1)Department of Zoology, National Science Museum, Tokyo, 3–23–1 Hyakunin-cho, Shinjuku-ku, Tokyo 169–0073, 2)Department of Zoology, Faculty of Science, Kyoto University, Kyoto 606–8501, Japan, 3)Raffles Museum of Biodiversity Research, National University of Singapore, Singapore, 4)Laboratory of Animal Morphology and Function, Graduate School of Bioagricultural Sciences, Nagoya University, Nagoya, Aichi 464–8601, Japan, 5)Mammal Division, National Museum of Natural History, Smithsonian Institution, Washington DC and 6)Department of Anatomy, Howard University College of Medicine, Washington DC, U.S.A. (Received 18 April 2003/Accepted 20 August 2003) ABSTRACT. The skull adaptation was functional-morphologically examined in 14 species of the tree shrews. From the data of the propor- tion indices, the similarities were confirmed between T. minor and T. gracilis, T. tana and T. dorsalis, and T. longipes and T. glis. We demonstrated that the splanchnocranium was elongated in terrestrial T. tana and T. dorsalis and shortened in arboreal T. minor and T. gracilis from the proportion data. In both dendrogram from the matrix of the Q-mode correlation coefficients and scattergram from the canonical discriminant analysis, the morphological similarities in the skull shape suggested the terrestrial-insectivorous adaptation of T. tana and T. dorsalis, and the arboreal adaptation of T. minor and T. gracilis. Since the osteometrical skull similarities were indicated among the three species of Tupaia by cluster and canonical discriminant analyses, the arbo-terrestrial behavior and its functional-mor- phological adaptation may be commonly established in T. montana, T. longipes and T. glis. KEY WORDS: canonical discriminant analysis, proportion, Q-mode, skull, tree shrew. J. Vet. Med. Sci. 66(1): 1–7, 2004 The arboreal and terrestrial tendencies and the patterns of sia). Since T. mulleri has been considered as a subspecies of behavior have been shown in various species of the tree T. splendidula [10], we included T. mulleri to T. splendidula shrews [1, 2, 4, 5, 9, 11]. Since the skulls have been func- in the statistical analyses. Sex determination was dependent tional-morphologically adapted to each style of locomotion, on the biological data of specimens. We used adult skull we previously examined the 4 species of Tupaia, and sug- specimens with fully erupted molars. gested that T. tana, T. javanica and T. minor have evolved Species composition, origin, and sex are shown in Table different behaviors [7]. Using the skull specimens of 14 1. Skull measurements were obtained with vernier calipers species of Scandentia, we compared the skull size and shape to the nearest 0.05 mm. Measurements are defined in Table among the species to confirm the relationships between the 2, and were based on Driesch [3]. In the present analysis we skull characteristics and the locomotion strategies [8]. used the same osteometrical data as those in the previous Although we compared the measurement data among the study [8]. The mean measurement values were shown in species and applied the principal component analysis to the each species [8]. The skull proportion indices were obtained osteometrical data to discuss the adaptational patterns of as quotients of each measurement value divided by the geo- skull form [8], the skull proportion indices should be actu- metric mean of all measurement values. The significant dif- ally obtained and the similarities of skull shape among tree ferences of proportion indices were examined among shrews should be quantitatively examined to elucidate the various species by nonparametric U-test using software Sta- morphological adaptation in various types of skull form in tistica (Statsoft Inc., Tokyo, Japan). these tree shrews. The clustering was performed by UPGMA method for the distance matrix converted from the matrix of the Q-mode MATERIALS AND METHODS correlation coefficients to avoid loss of information [12]. We did not use the measurement raw data but the proportion We examined 337 skulls of 14 species of the tree shrew indices in this cluster analysis. The distance matrix and the (Tupaia montana, T. picta, T. splendidula, T. mulleri, T. lon- dendrogram between species were obtained using the soft- gipes, T. glis, T. javanica, T. minor, T. gracilis, T. dorsalis, ware Statistica. The canonical discriminant analysis was T. tana, Dendrogale melanura, D. murina, and Ptilocercus also undertaken in each species using the proportion indices. lowii) [8]. The specimens have been stored in Muséum The first two canonical discriminant functions were calcu- National d'Histoire Naturelle (Paris, France), the Mammal lated and their scores were plotted using the software Statis- Division of the Smithsonian Institution (Washington DC, tica. We also applied another canonical discriminant U.S.A.), The University Museum, Kyoto University (Kyoto, analysis to the osteometrical data of the species excluding Japan), the Raffles Museum of Biodiversity Research, D. melanura, D. murina, P. lowii, T. tana and T. dorsalis to National University of Singapore (Singapore), the Depart- detail the morphological similarities among the typical ment of Wildlife and National Parks (Kuala Lumpur, arbo-terrestrial Tupaia species. List of measurements and Malaysia), and Bogor Zoological Museum (Bogor, Indone- their abbreviations are summarized in Table 2. 2 H. ENDO ET AL. Table 1. Behavioral character, origin and sex composition of the specimens in each species Species Behavior Origin of Specimes Male Female Tupaia montana Arbo-terrestrial Borneo 52 54 Tupaia picta Arbo-terrestrial Borneo 3 0 Tupaia splendidula Arbo-terrestrial Borneo 2 2 Tupaia mulleri Arbo-terrestrial Borneo 1 1 Tupaia longipes Arbo-terrestrial Borneo 9 3 Tupaia glis Arbo-terrestrial Malayan Peninsula 22 38 Tupaia javanica Arbo-terrestrial Java 9 6 Tupaia minor Arboreal Borneo, Sumatra 5 4 Tupaia gracilis Arboreal Borneo 14 6 Tupaia tana Terrestrial Borneo, Sumatra 24 26 Tupaia dorsalis Terrestrial Borneo 2 3 Dendrogale melanura Arboreal Borneo 10 5 Dendrogale murina Arboreal Thailand, Vietnam 2 4 Ptilocercus lowii Arboreal Malayan Peninsula 13 17 Total 168 169 The index of LBO was larger in T. minor, T. gracilis, D. Table 2. List of measurements and their abbreviations melanura and D. murina, and smaller in T. tana and T. dor- 1.Cranium salis. Profile length PL The dendrogram and the distance matrix are shown (Fig. Condylobasal length CL 2, Table 5). The dendrogram of male was fundamentally Short lateral facial length SL Zygomatic width ZW similar to that of female. We confirmed the two large clus- Least breadth between the orbits LBO ters as follows: 1) T. montana, T. picta, T. longipes, T. glis, Greatest neurocranium breadth BNB T. tana and T. dorsalis, and 2) T. javanica, T. minor, T. gra- Median palatal length MPL cilis, D. melanura, D. murina, and P. lowii. Only T. splen- Length from Basion to Staphylion LBS didula was present in the different cluster between sexes. Dental length DL The results of the discriminant analysis were visualized Greatest palatal breadth GPB (Fig. 3, Table 6). The pattern of plots was not demonstrably Greatest mastoid breadth GMB Height from Akrokranion to Basion HAB different between sexes. The plots of T. tana and T. dorsalis 2.Mandible and those of T. minor and T. gracilis were concentrated. P. Length from the condyle LC lowii was separated from the other species in plot distribu- Length from the angle LA tion. Length from Infradentale LIA We also showed the plots of the individual discriminant to aboral border of the alveolus scores and the discriminant functions using the osteometri- Height of the vertical ramus HR cal data in each species except D. melanura, D. murina, P. Height of the mandible at M1 HM lowii, T. tana and T. dorsalis (Fig. 4, Table 7), since we The measurement items were based on Driesch [3]. could not clarify the plot distribution in the typical arbo-ter- restrial species at the chart scale of Fig. 3. The plot distribu- RESULTS tion of T. montana, T. glis and T. longipes showed the species-specific tendency at the higher scale of chart (Fig. The proportion indices are arranged in Table 3, and the 4), although the plot areas were adjacently located among relationships of the proportion indices between PL and SL the three species. The plots of T. minor, T. gracilis and T. are shown in Fig. 1. The significant differences are shown javanica were concentrated into the three groups separated among species in Table 4. The significant differences were from the other species (Fig. 4), although the plot areas of the not found in many measurements between T. minor and T. three species were partially overlapped in male. gracilis, T. tana and T. dorsalis, and T. longipes and T. glis (Tables 3 and 4). The indices in the two species of Dendro- DISCUSSION gale were similar in many measurements. The index of SL clearly demonstrated that the face and splanchnocranium We pointed out that the five groups of skull adaptation were rostro-caudally elongated in terrestrial T. tana and T. pattern have been established in these tree shrews [8]. In dorsalis and shortened in arboreal T. minor, T. gracilis and addition, the separation of eight groups was obviously visu- P. lowii (Fig. 1). The PL and CL in the two arboreal species, alized in the principal component charts [8],