process only in really exceptional situations, whenever it is justified to treat them apart rather than to integrate them in the general mold.

2.1.5

Iterative logic

The study of above mechanisms was based implicitly on a matching between a description composed of sub-descriptions (or local descriptions) and a descriptive model composed of sub-descriptive models. The description is concerned with the observed facts whereas the model is concerned with observable facts.

It often occurs that, in a description, several characters, although they are not rigorously identical, are of the same "kind" and follow the same descriptive sub-model. Consider the example of mammal teeth. If we were to describe the human set of teeth (if we are afraid of the dog's one), we well see that there are several kinds of teeth, let say 3 or 4 kinds depending on our perspicacity. Those of us who are well informed will name them directly : incisors, canins, premolars and molars; but it is not necessary at to know all their names to describe them correctly. It is sufficient to follow a common sub-model of teeth description, and apply it iteratively as many times as necessary (here 3 or 4 times according to the ability of the descriptor to see the difference of nature between premolars and molars).

We pointed out that we had to respect the fundamental homology principle. If we had to compare in detail the set of teeth of the cat and of the dog, we must be sure to compare canins (or "fangs") of one with canins of other; otherwise we get lost. One must be aware of the interpretation risk (of being subjective) arising when venturing in "local identifications"; the descriptor who is not aware of the limits of his knowledge in the domain would make a mistake if he names canins the tusks of an elephant and the tusks of a morse; the consequence of this mistake is that objects to be compared are not homologous : the elephant tusks are modified incisors, whereas the morse ones do are canins, though of an exceptional size. It is right that it is difficult to only describe without searching to understand and to learn; but paradoxically, a good description should not call for intelligence because we are biased by our mental model and anybody may make a mistake.

Another situation may occur when describing. Suppose that we proceed to a local description of a plant inflorescence, and that the corresponding descriptive sub-model gives as a list of possible colours white, yellow and red, and that several answers are allowed (multiple choice). If we answer together white and yellow, that means that the colour is white or yellow, it is a lack of precision (why not an intermediary shade as white-yellowish ?). To express that we observe effectively the colours white and yellow simultaneously, it is necessary to make two successive local descriptions, one for describing flowers with white colour only, the other for flowers with yellow colour; in fact, there is a high probability to find other characters to differenciate the two types of flowers, as for instance their localisation in the inflorescence or also their sex, and that these flowers have not the same organical signification.

Remark: we need to distinguish this last case from the description of associations (of colours for instance) which are referenced under names like streaks, mosaic, etc. The fact that a leaf is variegated with green and yellow must not be expressed by the choice of green and yellow simultaneously, but by the single choice of the association green+yellow duly indexed. This can be represented in the descriptive model by a hierarchy of classified values like this :

leaf colour

variegated^unicoloured green+yellowgreen+purplegreenpurplesilver

Each time we have to express co-existing facts (noted simultaneously), the iteration process is