Humans seem to be genetically endowed with a mathematical aptitude known as subitizing. It refers to the ability to correctly recognize very limited quantities (up to 3/4 items). Conversely, higher cardinalities require a different, non-innate process: counting. In keeping with the hypothesis that cognition plays a major role in structuring the linguistic complexity (cf. Lakoff and Johnson, 1980; Heine, 1997), here I suggest regarding subitizing as a significant shaping force in language. Our cognitive limitations may be reflected in the organization of linguistic structures, especially in the field of numerals. There are a number of clues which lead us to believe that certain fundamental linguistic features have autonomy and priority compared to others. These include: the diffusion of “1-2-(3)-many” numeral systems, the morpho-syntax of the lowest number words, the suppletion of the first ordinals, etc. Evidence of this in the IE. family was collected in §4. The split between the set of the first few elements and the rest is sometimes regarded as accidental or due to a chronological gap in historical evolution. However, I think that subitizing, as an innate and universal cognitive pattern, can provide a more satisfactory explanation. Since the first quantities are perceived more clearly and immediatly, the structure of some linguistic domains (e.g. numerals) reflects this cognitive hierarchy.