Kaszab Zoltán (szerk.): A Magyar Természettudományi Múzeum évkönyve 76. (Budapest 1984)

Demeter, A. ; Lázár, P.: Morphometric analysis of field mice Apodemus: character selection for routine identification (Mammalia)

they had in their hand from the geographical region under study, or by such goals as identification from owl pellets. Several authors have documented the overlap in various external and cranial measurements of the Apodemus species (e. g. FELTEN 1952; KRATOCHVIL & ZEJDA 1962; NIETHAMMER 1969a; TVRTKOVIC & DJULHS 1977). It has generally been thought more useful to plot the individual specimens in a bivariate abstract space, in the form of a scatter diagram delimited by orthogonal axes of such measurements as the length of the hind foot versus length of the upper molar row (AMTMANN 1965 ; ENGLANDER & AMTMANN 1963; NIETHAMMER 1969a; WITTE 1964), or in general, the correlation of various measurements with (and sometimes regression upon) such "size-representing" measurements as head & body length, condylo-basal length of the skull (e. g. HAMAR et al. 1966). The cloud of points representing the specimens are segregated into one or more clusters depending on the number of species and/or populations studied, the degree of segregation being very much variable. More specifically for owl pellet studies, plots of depth of the incisor versus length of upper molar row (FIELDING 1966), breadth of M

1 versus length of upper molar row (HAITLINGER & RUPRECHT 1967), length of diastema versus length of palatal foramen (TVRTKOVKÍ & DZUKIC 1977) and the distance from the incisor to the third upper molar versus the length of the palatal foramen (TVRTKOVIC 1979) have been suggested to be useful. The problem with the frequency distribution diagrams and bivariate scattergrams is that often it is difficult to decide whether a certain point should be referred to cluster A and not to cluster B, or where to draw the boundary between the two. This is when the science of statistics was called upon, first to test whether the differences in measurements were significant (examples too numerous to mention), later to actually aid identification. The latter aim calls for methods which fall into the realm of multivariate analysis, especially discriminant analysis, which was first used to study variation in the long-tailed field mouse by DELANY & HEALY (1964). Later AMTMANN (1965) and ENGLÄNDER & AMTMANN (1963) used this method in an attempt to show introgressive hybridization between A. sylvaticus and A. flavicollis, which was defeated on statistical grounds by REMPE (1965) who incorporated an allometric standardization procedure into the discriminant model, and refuted in the light of biological evidence by NIETHAMMER (1969a). More recently, VAN DER STRAETEN & VAN DER STRAETEN-HARRIE (1977) have used discriminant analysis to facilitate identification of A. sylvaticus and A. flavicollis in Belgium, not leaving out of consideration the special constraints of owl pellet studies. What also struck us when we searched the literature on the taxonomy and identification of Apodemus was that there is no consensus on the choice and number of characters employed in the studies. It is customary to take four external measurements plus possibly the weight of the animal ; in a few studies measure (DELANY 1965; NIETHAMMER 1969a; ROOD 1965) or score (FULLAGAR 1967; LARINA 1958; URSIN 1956) the size of the pectoral spot, and obtain some measure of the pelage colouration (Munsell value and chroma (DELANY 1964), or subjective comparison with the Ridgeway atlas (FELTEN 1952)). A survey of twenty-one studies has shown that of the cranial measurements condylo-basal length and length of upper molar row, thought to represent overall size, are the clear winners, other ones often taken include greatest length of skull, length of palatal foramen, length of diastema, width of braincase, least interorbital constriction and depth of incisor. Unless specifically intended to show the alleged usefulness of a character or a limited combination of characters, it is customary to take about six to twenty-five measurements of the specimens. It is, however, this apparently haphazard choice of characters that intrigued us and prompted us to ask: what should we measure? Whatever is customary, thought to be useful, or convenient to take? And how many of those? As many as we feel like, think to be presentable, or available time, effort and money permits? We have taken a different course and have attempted to answer these questions as objectively as possible. At the very time we asked ourselves the above questions, we encountered the work of DAVIS (1981), which was instrumental in the formulation of our analysis. Material For each species, samples of specimens from lowland as well as hilly localities were selected. The pilot study included a total of 156 specimens : Apodemus sylvaticus (43 specimens): inventory numbers 66. 98.4-7,9-10, 12-13 and 20-23. 8 rfrf +4 $ 9, Gyula, 20 October, 1963, leg. G. Topái. I. Szabó, Z. Kaszab and B. Hüttler; 65.107.121, 23-27,10 çfçf

+16 9 9 , Bakonyszentlászló, 4 October, 1960, leg. A. Kötél. Apodemus flavicollis (70 specimens) : inventory numbers 65. 98. 1, 2, 8, 11, 14-19, 4 ç?c? +799. 65.99.1-33, 16 çfcf +17 9 9, Gyula, 20 October, 1963, leg. G. Topái, I. Szabó. Z. Kaszab and B.