The diversity of clinical pictures and different theoretical approaches may explain the blurred boundaries of the concept of autism. Since Kanner, autism has been explained either as a social/affective deficit or a cognitive one. In the fruitless debate of the 1970/80s, some authors argued that the primary deficit was social/affective while others that it was in the cognitive area of language. Today the cognitive and developmental approaches predominate. The latter revives the social/affective approach to autism and dissolves the social x language opposition. One of its epistemological assumptions is to view language as a social activity because its precursors can be found in the mother-infant nonverbal communications. The purpose of this article is to begin a discussion about the importance of the epistemological assumptions of the different theoretical approaches to autism for the search of its etiology and intervention.