This study centers on sequences of consonnants and tries to show how their repartition is not a random repartition. They are limited in number as welle as in length and by the types of consonnants used in the sequences. The analysis is conducted from a latin corpus because in that field, latin has clearcut structures. All the same its results have a wider scope. Already, at bundaries between two words, that is in speech (fr. langage), not in the language (fr. langue), the sequences of consonnants have a lesser frequency than should be expected if we consider their number at the end and beginning of words. In morphology, there appears no sequences of consonnants at the articulation of morphemes. At last, Sequences may be reduced according to the places and manners of articulation. Lastly we leave to the care of experimental phonetics the explanation of the facts that we brought to light out of te intricate contexts that hide them.