Grammaticalization of a sign or construction takes place in a given language system. It may be fully described with reference to this system, to the variation that it allows and to the discourse that it underlies. Beyond that, one may ask for the factors that lead its speech community to implement a certain grammaticalization. Many grammaticalization paths are universally available, so an appropriate item or construction may be recruited for grammaticalization and fed into such a path at any time. There remains, nevertheless, the possibility that there are factors outside the linguistic system which favor such a variation. One of these factors is language contact.

The grammaticalization of the numeral ‘one’ to an indefinite article and of the distal or neutral demonstrative ‘that’ to a definite article both constitute universally available paths (Givón 1981). An indefinite article has developed along these lines in Turkish. A definite article has developed along these lines in Irish and Indonesian. Both paths have been followed by Germanic and Romance languages.

Now the Germanic and Romance languages have been in contact for roughly two millennia. This contact was especially intense between French and English just at the time in which they grammaticalized their articles. This may be a factor in the parallel and simultaneous development. Although we are here talking about a documented historical period, there are not too many data relevant to this problem because the language of written communication was mostly Latin whereas the grammaticalization of the articles was in all probability a process pushed in oral communication.

There are other languages in the world which developed an article system in historical times. Yucatec Maya grammaticalized a demonstrative to a definite article and the numeral one to an indefinite article according to this globally available model. In this particular case, however, the development happened in the past half millennium, which is precisely the time that Yucatec Maya has been in constant contact with Spanish, a language which offered such an article system as a model and had created it by the very same grammaticalization process several centuries earlier. In such a case, it is highly probable – although, again, hard to prove – that the grammaticalization process in the Mayan language was set into motion by the contact with Spanish.

In the specific case of the articles, no special framing conditions are known which a language system must meet if it is to develop an article system. The sources of the two grammaticalization paths are universally available, and the introduction of articles does not seem to conflict with anything else in the language system. In other words, pretty much any language might develop an article system either by itself or if it came into close contact with an article language like Spanish. Even if the language had worked well without articles, their introduction would, at any rate, heighten the redundancy in messages. The languages in question – Germanic, Romance and Yucatec Maya – have forfeited other inherited categories at the same time so that the overall measure of grammatical structure and, thus, of redundancy may have remained constant.