Page 2 - CrociduraWagler_Sara

284 M. SARA, M. LO VALVO, L. ZANCA

Downloaded by [183.219.240.165] at 10:10 22 March 2014 were used as controls to evaluate the taxonomical distance among CVA procedures. CVA seeks linear combinations of transformed
the other insular taxa. morphometric variates that have greatest between-group variation
relative to their within-group variability. It is thus possible to trans-
The skulls found inside the pellets have their posterior part very late the starting data on to a three-dimensional plane (canonical
often damaged. Of the 484 skulls formerly selected, it was possible variate space) that best separates the OUT's centroids.
to measure five variates only in 10.33? of the whole sample, four
variates in 30.161 of the cases, three variates in 29.96Z, and two The distance between the centroids was calculated by Mahala-
variates in 98.551. Moreover, in such remains the skull and the nobis' generalized distances (D2). These are akin to phenetic dis-
mandibles were rarely found joined together. Since multivariate tances but take some account of intraspecific variation and correla-
analysis requires the use of a complete set of data for each speci- tion of characters (Jenkins, 1976). Significance levels of Mahalano-
men, the analysis of skull characters had to be excluded due to the bis' distances were first expressed by Fisher's F test, later on by T2
limited amount of information available. Thus, we selected for the Hotelling's transformation.
analysis only 480 right mandibles (see Table I for sample sizes of
each location). The following morphometric variates were mea- Mahalanobis' distance matrix of mandible finally allowed the
sured on the mandible: total length excluding incisor (UKL), association dendrogram of the OTU's considered to be displayed.
height at the coronoid process (COH), length of molars (MI-M3), The method followed to construct the dendograms was derived
breadth of articular condylum (E). from the sum of square groupings (Orloci, 1978).

The first step was to run principal component analysis (PCA); Furthermore, we report the morphotypes graphic comparison of
this multivariate analysis is based on the correlation matrix corre- the skull large upper premolar tooth (in occlusal view) and of the
sponding to the variance-covariance matrix of the standardized posterior surface of the mandible condylar process. These char-
variates (Chessel & Auda, 1986). acters have already proved useful in morphological discrimination
(Jenkins, 1976; Hutterer, 1986;.Molina & Hutterer, 1989).
The analysis seeks the interrelationships among the variates mea-
sured on each single specimen and allows the clustering of all the RESULTS
specimens showing a similar pattern of interrelationship. The pur-
pose of the method is to give evidence of one or more clusters The principal component analysis of mandible va-
(taxa) on each insular and peninsular sample. riates raised a single cluster for each insular and penin-
sular sample. In Table I, sample sizes, mean values,
To emphasize this aspect, the 265 mandibles coming from Sicily and standard deviations of the morphometric variates
were not grouped together, but each location of pellet collection considered are reported.
was considered to represent a separate «island». The samples from
Favignana and Levanzo were excluded from the PCA, due to their Table II shows the discriminant coefficients of the
small size. canonical equations used to display each OTU on the
canonical space. The canonical coefficients statistically
The second step was to perform canonical variate analysis
(CVA), that maximizes separation between taxonomical units (Le-
gendre & Legendre, 1979).

In the study the group of specimens from each island is referred
to by the term operational taxonomic unit (OTU), according to the

TABLE I - Mandible biometry (mm) of the Crocidura OTU's from central Mediterranean islands and mainland Italy. Mean ± SD of length
mandible excluding incisor (UKL,), length ofmandible molars (M1-M3), breadth ofarticular condylum (E), andheight ofmandible at the
process (COH).

UKL C. leucodon C. suaveolens C. russula C. cl russula
M1-M3 Sardinia Pantelleria
E Italy Italy » = 99
» = 12 » = 11 » = 42
COH 10.55 ± 0.28
10.43 ± 0.33 9.33 ± 0.36 3.93 ± 0.11 10.93 ± 0.33
UKL 4.07 ± 0.09 3.55 ± 0.15 0.98 ± 0.06 4.06 ± 0.14
M1-M3 1.21 ± 0.09 0.96 ± 0.04 4.84 ± 0.14 0.92 ± 0.08
4.88 ± 0.18 4.20 ± 0.14
E C. cf sicula 4.98 ± 0.18
COH C. sicula C. sicula Marettimo
Sicily Gozo C. cf sicula
UKL »=17 » = 16 Ustica
M1-M3 » = 265 »=13
E 9.94 ± 0.26 10.03 ± 0.27
10.47 ± 0.29 3.73 ± 0.11 10.15 ± 0.35
COH 3.87 ± 0.14 3.74 ± 0.09 3.69 ± 0.09
0.90 ± 0.05 0.91 ± 0.07
0.90 ± 0.09 4.36 ± 0.11 0.93 ± 0.04 4.56 ± 0.13
4.59 ± 0.15 4.37 ± 0.13
C. cf sicula
Favignana C. cf sicula
Levanzo
9.67
3.57 9.77
3.68
0.83 0.88
4.24 4.41