Appendix: A Note on the Multivariate
Representation of Dissociation

Experimental studies of individual characters have provided most recorded cases of dissociability, but through the multivariate study of form separate patterns of variation in size and shape have been detected as an almost universal property of covariation within natural populations. The loadings of original variates upon orthogonal factor axes usually cluster very strongly. Measures of size tend to associate with the first principal component, while loadings upon subsequent axes reflect differences in shape.* Since axes are orthogonal, the clusters of size and shape are "independent" in the most technical sense. This is a constraint of the method; it must not be equated with biological reality.

Fig. 42. A Cerion shell from Bimini. Markers of development
are indicated: (a) the adult lip; (b) the suture of the last pre-
adult whorl and (c) the overlap of the adult aperture onto the
last pre-adult whorl (the extent of overlap, c minus b measured
parallel to the axis of coiling, is a measure of development).

* This arises from the unenlightening fact that most parts are larger in bigger animals. This primary correlation nearly always subsumes the greatest amount of variance and provides the first principal component. Jolicoeur (1963) has derived a general test for isometry using this axis; this is important for my definition of size differences as geometrically similar.