Because almost all characters are correlated, either positively or negatively, through a complex interplay between the genome and developmental programmes, it makes no sense trying to correlate individual characters with individual climate variables. A change in one character, perhaps as a result of environnmental selection by one or more other climate variables, will lead to a change in another character that may be primarily, but not exclusively, related to one or more different climate variables. Instead we have to look at climate-leaf relationships 'in the round'. We need to assess the relationships of the combined character suite (something we call the sample physiognomic spectrum) with several climate variables simultaneously, and try to extract the primary relationships between leaf form and climate. This also means that a preservational or collecting loss of one or more characters does not prevent the extraction of meaningful palaeoclimate information from fossil leaves.

To do this analysis of multiple variables simultaneously CLAMP uses a statistical ordination method known as Canonical Correspondence Analysis (ter Braak, 1986). This is called a 'direct' ordination method because environmental variables are explicitly related to the samples by combining two files, one with leaf physiognomic data (shape, size etc.) and one with climate data. This combination generates a cloud of modern vegetation samples in what we call physiognomic space. Samples are positioned relative to one another based on the leaf physiognomic traits found in vegetation represented by those samples. Within that cloud climate trends (vectors) are positioned based on the climate measured at each of those sites.

Leaves in fossil assemblages are scored the same way as the modern leaves, but because the fossils have no known climate data associated with them they enter the analysis in a 'passive' way. This means they find their own place in the cloud of modern samples based on the similarity between the physiognomic trait spectrum of the fossils to those of the modern samples that form the calibration space. Because they are passive, the introduction of the fossil samples does not disturb the structure of calibration space. The position of the fossils along the various climate vectors provides the palaeoclimate estimates