Table 1.
Sample localities and size.
Figure 1.
From Viscosi et al., 2009, modified.
Figure 2.
Shape variation including all observations and replicas.
Scatterplots of PC1 vs PC2 (a), and PC3 vs PC4 (b), which overall explain 88.3% of variance. As an example, the first and second replicas of the specimen ‘Campobasso 14a’ are visualized using outline drawings magnified 10 times (c); full shapes are shown in the visualization and square and diamond symbols are used to indicate the position of the replicas of this specimen in the shape sub-spaces of the first PCs.
Table 2.
Centroid size variation.
Table 3.
Shape variation.
Figure 3.
Size variation after averaging leaves within trees.
Box plots (drawn in PAST): median, 25–75% quartiles, minimum and maximum.
Figure 4.
Shape variation after averaging leaves within trees.
Scatterplots of PC1 vs PC2, which together explain 74.2% of variance. Shapes are visualized for the positive extremes of these axes using outline drawings; there is no magnification and square/diamond symbols are used show the positions of visualized shapes in the PCA scatterplot.
Figure 5.
Shape variation after averaging leaves within trees.
Scatterplots of PC3vs PC4, which together with the first two PCs (Fig. 3) explain 90.3% of variance. Shape are visualized using the same conventions as in Figure 4.
Table 4.
Differences between means of populations after averaging leaves within trees.
Figure 6.
Discriminant analysis of geographic populations using leaf shape after averaging within trees.
Frequencies of discriminant scores predicted by a jacknife (leave-one-out) cross-validation are shown using histogram bars; population mean shapes are visualized using outline drawings magnified 10 times.
Table 5.
Differences between populations after averaging leaves within trees.
Figure 7.
Regression of shape onto size pooling within populations.
Scatterplot of regression scores (i.e., the projection of shapes in the direction of the vector of regression coefficients, Drake and Klingenberg, 2008) vs centroid size; shapes at the opposite extremes of the range of allometric variation are shown using leaf outlines with no magnification.
Figure 8.
Discriminant analysis of geographic populations using leaf shape after averaging within trees.
Same differences as in Figure 6 (printed black and white version) visualized in PAST using TPS deformation grids and colour coded Jacobian expansion factors which measure the degree of local expansion or contraction of the grid: yellow to orange red for factors >1, indicating expansions; light to dark blue for factors between 0 and 1, indicating contractions).
Table 6.
Allometric shape variation after averaging leaves within trees.
Figure 9.
Example of the effect of different superimpositions on the interpretation of results.
A set of 10 random triangles (raw data) was superimposed either using Bookstein baseline superimposition (a1) or Procrustes (b1). Shape coordinates were subjected to PCAs whose results were illustrated using biplots (a2, b2) showing both the scatterplot of the specimens (filled circles) and the loadings (dotted lines) used to weight the matrix (X1, Y1, etc.) of shape coordinates. Shape variation at the positive extreme of PC1 was visualized magnified four times using either displacement vectors (a3, b3) or TPS grids (a4, b4).