This paper examines the question of whether two labels should be used for the units traditionally called "sentence" and "clause" or whether the same label should be used and the units distinguished in some other way. Proponents of a two-level analysis have traditionally argued that sentences and clauses can have different structures. A related argument is that sentences and clauses have different distributions. It is proposed that a sentence and a clause have the same deep structure, and that they have the same potential distribution, and therefore can be assigned the same categorical label. An approximate theory of levels is proposed which considers words and sentences as borderline categories, with words defined as minimal phrases and sentences as minimal utterances. Four grammars are outlined as well as a sentence taxonomy which is compatible with each alternative. A translation into Finnish and Swedish of the terminology contained in the taxonomy is proposed. It is concluded that one category, "sentence," should be used for the traditional categories "sentence" and "clause." This category can be subdivided according to functional criteria into "independent,""conjunct," and "hypotactic" sentences, and according to structural criteria into "compound sentence" and "clause." (AM)