Canidae, Fischer, 1817

Gaubert, Philippe, Veron, Géraldine & Tranier, Michel, 2002, Genets and ‘ genet-like’ taxa (Carnivora, Viverrinae): phylogenetic analysis, systematics and biogeographic implications, Zoological Journal of the Linnean Society 134 (3), pp. 317-334 : 320-322

Family Canidae

Canis adustus

Vulpes vulpes between glass slides and epoxy resin inclusion for study of cross sections.

Finally, we took into account some coat characters, such as patterns and contrasts, but only for the intrageneric analysis. In fact, coat patterns seem to be liable to many convergence phenomena, even at the family level (see Werdelin & Olsson, 1997 for the Felidae ). Coat colour, which remains difficult to objectively characterize ( Endler, 1990), depends on seasonal changes ( Stangl & Grimes, 1987), habitat ( Hadley, 1972; Ortolani & Caro, 1996) and (which is of concern here) on how the specimens have been prepared. Moreover, the behavioural significance of patterns and contrasts ( Roeder, 1984) imposes phylogenetic constraints, which is not the case with regard to hair colouration, as genets can not perceive colours ( Grzimek, 1990). Thus, we did not consider this last character in our phylogenetic analysis.

PHYLOGENETIC ANALYSIS

Two cladistic analyses were carried out: one of genera within the Viverrinae (intergeneric analysis), the other tackling the question of phylogenetic relationships within the genus Genetta (intrageneric analysis). Indeed, the Genetta spp. polytomy observed in the intergeneric analysis required a second analysis including coat characters in order to try to resolve intrageneric relationships. The two data matrices (Appendices I and II) were assembled in MacClade version 3.07 ( Maddison & Maddison, 1997) and then analysed using PAUP version 3.1 ( Swofford, 1993). The PAUP settings used for the analyses consisted of a heuristic search, option ‘stepwise addition: closest’.

In the intergeneric analysis, four different runs using different outgroups (1/ Viverridae ; 2/ Herpestidae ; 3/ Felidae ; 4/ Felidae + Canidae ) were carried out in order to establish the ingroup topology. On the basis of the results of the intergeneric analysis, two civets ( Civettictis civetta and Viverricula indica ) were used as outgroups in the intrageneric analysis.

These two levels of analysis require two different character sets (Appendix III) which differ in the included characters and in a few character codings (character coding dependent on taxonomic sampling). Moreover, characters used in the intergeneric analysis which did not vary in the taxa incorporated in the intrageneric analysis were omitted there. Some characters, inapplicable to the outgroups used in the intergeneric analysis, were added in the intrageneric analysis (see Table 3 for a summary of data matrices and character coding).

Although most phylogenetic studies avoid using polymorphic characters (as defined by Mabee & Humphries, 1993), we chose to take intraspecific variability into account. Indeed, several papers dealing with the ubiquity of polymorphism ( Smouse et al., 1991; Rannala, 1995; Wiens, 1995, 1998) advocate its coding with the aim of increasing the accuracy of phylogenetic reconstruction. Rather than splitting every included taxon into monomorphic units ( Nixon & Davis, 1991) or coding polymorphism using frequency scales ( Rannala, 1995; Wiens, 1995, 1998), we opted for the unscaled coding ( Wiens, 1995). In this method, a polymorphism is given its own character state (for example, 0 for character state a, 1 for character states a and b, 2 for character state b). This kind of coding seems more judicious to us in a cladistic analysis, as it considers each polymorphism to be a dual expression of the genes coding for the characters concerned. However, we avoided using characters that were polymorphic in a large number of taxa, as they might only be incorrectly defined characters.

Ingroup monophyly was tested by permutations of the outgroup taxa listed in the data matrices ( Barriel & Tassy, 1998). In each analysis, we carried out successive ingroup taxa deletions in order to determine which ones stabilized or destabilized the tree topology. This form of manipulation allows us to identify which taxa (or clades) share a high number of character states with their respective sister groups.

In order to provide an estimation of the robustness of each node, the Decay Indice (DI) is shown for each cladogram node ( Bremer, 1988). Consistency Index (CI) and Retention Index (RI) values, which reflect homoplasy levels ( Farris, 1989) are shown. In addition, phylogenetic signal values (g1) for 100 000 trees ( Hillis, 1991) are indicated for each topology.