Page 3 - CrociduraWagler_Sara

INSULAR VARIATION IN CENTRAL MEDITERRANEAN CROCIDURA 285

TABLE II - Discriminant coefficients usedfrom CWA to display the OTU's on to the three-dimensional plane as in Figure 1, and to computethe new
canonical variates to obtain the misclassification rates as shown in Table III.

C. leucodon C. suaveolens C. russula C. cf russula
Italy Italy Sardinia Pantelleria

Fl 1.53 - 3.38 1.13 2.27

F2 3.43 2.50 0.69 - 0.61

F3 - 1.42 0.62 0.27 0.22

C. sicula C. sicula C. cf sicula C. cf sicula
Sicily Gozo Marettimo Ustica

Fl - 0.37 - 2.02 - 1.97 - 0.93

F2 - 0.51 - 0.47 - 0.43 0.47

F3 - 0.15 0.06 - 0.32 1.01

C. cf sicula C. cf sicula
Favignana Levan2o

Fl - 3.01 - 1.79

Downloaded by [183.219.240.165] at 10:10 22 March 2014 F2 0.25 - 0.95

F3 1.10 1.01

significant (Wilks' test) used in computation of the 2) that shows the final relationship pattern among the

canonical variates, the variance rates of factorial axes OTU's. The Ustica OTU joins Sicily, a second group of

and their canonical correlation coefficients are reported association is shown by the Egadi archipelago and

in Appendix 1. Gozo; these two groups associate together before to

Figure 1 shows the-OTU's centroid distribution on join .C.suaveolens and finally Pantelleria joins C. russula.

the canonical variate space (F1-F3), the axes F1-F2 Morphotypes of the large upper premolar and ofthe

display 89.11of total variance, showing either separa- posterior surface of the mandible condylar process, re-

tion of some OTU's ( C suaveolens, C. leucodon, C. russu- ported in Figures 3 and 4, stress some other osteologi-

la and Pantelleria) or a quite distinct grouping. The cal differences that could be useful in taxonomic diag-

first group is formed by most of the Sicilian centroids nosis, together with the taxonomic key proposed by

(D, E, ... P), the second by the Egadis, Ustica and Vogel et al., (1989). Once again the general relation-

G020. Sicilian centroids (A, B and C) are intermediate ships among the OTU's were confirmed; Sicily has its

between these two groups, and only the centroid Q is own morphotypes quite similar to Gozo, Ustica and

distinctly separated from the others. Marettimo (expecially large upper premolars, while the

By means of the discriminant equations and the new condilar processes have a similar shape but smaller

canonical variates it was also possible to compute the size) and Pantelleria is quite similar to C. russula.

misclassification rates (or error probability of classifica- Finally, Appendix 2 shows some external measure-

tion) of the specimens, as reported in Table III.For ments and the body weight of shrews trapped in Sicily,

example, of the 265 Sicilian mandibles, more than half as compared with specimens from the Egadi archipela-

(57.362) are correctly re-classificated as Sicily OTU's. go and the island of Gozo. Sicilian shrews show a not

Only 10.94? have morphometric variate relationships significant sexual dimorphism, but are significantly

similar to those of the Sardinian C. russula, none of greater than the Egadis specimens.

them could be attributed to C. suaveolens or to C. leuco-

don, and 31.691 could be attributed to OTU's present

in islands surrounding Sicily. Table III shows also that

the mainland OTU's C. suaveolens and C. leucodon have aDISCUSSION

low misclassification rate, while the insular ones have

higher rates, generally showing a small scattered pat- The results of the PCA highlight the presence of a

tern. This pattern of misclassification raises mostly be- single taxon on each island considered. Only one spe-

cause some of the specimens in a sample set out again cies seems therefore to live in Sicily, and this finding

to OTU's present in different islands, but they could ought to answer the question, posed by the authors,

indeed belong to the same taxon (see discussion). regarding the presence of up to three splioli, 1879;

Mahalanobis' generalized distances between the Pasa, 1959; Toschi & Lanza, 1959; Niethammer,

OTU's are reported in Table IV; this distance matrix 1962; Vesmanis, 1976; Jenkins, 1976; Vesmanis &

allowed us to make the association dendrogram (Fig. Vesmanis, 1982; Contoli etal., 1989).