Four methods for testing whether a characteristic is the result of selection or constraint:

1. The use of adaptive prediction.

If a theory of shell adaptation predicted accurately and successfully the relationship between shell form and environment – which forms should be present, and which absent, in various conditions – then, in the absence of an equally exact embryologic theory, that would count in favour of adaptation and against developmental constraint.

2. A direct measure of selection.

This requires that non-existent phenotypes are somehow made experimentally, and then tested to see how selection worked on them. We then find out by observation whether there is negative selection against these forms.

3. A measure of the character's heritability.

If a constraint is preventing mutation in a character, it should not be genetically variable. Genetic variablility can be measured, and the constraint hypothesis will be refuted for any character that shows significant heritability.

4. Comparative evidence from many species.

When more than one character is measured, and the values for the two characters in different organisms are plotted against each other, a relationship is nearly always found. This has been done most often for body size, and the relationships are then called allometric. For example, graphs of tooth size against body size in different species of primate show positive correlations. Julian Huxley (pictured opposite in 1918) was an influential early student of allometry, and he liked to explain allometric relationships by the hypothesis of constraint:

"whenever we find allometric relationships we are justified in concluding that the relative size of the horn, mandible, or other organ is automatically determined as a secondary result of a single common growth mechanism, and therefore is not of adaptive significance. This provides us with a large new list of non-adaptive specific and generic characters.”