A species is scored for its full range of variation in leaf size. Exclude any leaf that is obviously immature and not fully expanded. These will usually be near the shoot tip. Leaf size is measured in categories using the template available here.

Note that if a lobed leaf is dissected more than half the distance from the apex of the lobe to the base of the lamina or to the midrib, the unit area scored is an individual lobe (Figure Z).

Z) In lobed leaves the individual size unit is the lobe.	AA) Measure only the leaf lamina and not the petiole.

Leaf size in CLAMP is best measured using a template that can be downloaded here. Print this out on a transparent sheet to the correct size using the scale on the template. This template can then be overlain on both living and fossil leaves and size estimated quickly.

The leaf size (or area) categories are defined differently from the categories proposed by Raunkiaer (1934), as modified by a Webb (1959), although the division between the Microphyll III and Mesophyll I categories approximately corresponds to the division between Webb's (1959) Notophyll and Mesophyll categories. Raunkiaer's (1934) original categories were defined by him based on examination of a large number of herbarium specimens.

In sizing leaves early in the development of CLAMP, Wolfe observed that many species tend to have leaves concentrated in certain sizes the represent categories intermediate between two of Raunkiaer's categories, and thus Wolfe developed the system proposed here. Note that if a lobed leaf is dissected more than half the distance from the apex of the lobe to the base of the lamina or to the midrib, the unit area scored is an individual lobe.

During development of CLAMP, three additional size categories were scored. One was smaller than the Leptophyll I category and two were larger than the Mesophyll II category. In all three instances, the multivariate analysis produced character-state scores almost the same as the scores for the adjacent leaf size character-states. The similarities in scores indicated that no significant information was being added, and deletion from the database of the duplicate character states produced somewhat higher eigenvalues and had little effect on samples scores.