The Sememic Approach to Structural Semantics1

Abstract

AS PART of the justification for their existence, languages have an intimate but only partially understood relationship to a vague entity which we call meaning. There is no general agreement as to just how initimate this relationship is: that is, as to whether meaning is or is not included within language, partly or wholly, in one of its aspects or levels or more than one (if indeed there are more than one). But many linguists have thought it would be very desirable, and perhaps also possible, to set up or discover some structural units which would either be units of meaning itself or in some other way would help us to systematize our treatment of that vague entity. A name for such a unit, Sememe, has been available for a long time. It was provided by Bloomfield (1926). We are not constrained, of course, to use his definition in order to use the term. Just as with the morpheme and other concepts of his and others, we may refine our understanding of the unit without adding a new name to an already formidable collection of terminology. For Bloomfield, the sememe was the meaning of a morpheme. This concept would seem to put sememes and morphemes in a one-to-one correspondence with each other-that is, a sememe for every morpheme and a morpheme for every sememe. Such a view of the relationship ought to make it unnecessary for there to be two kinds of units, and one is led to wonder why, in this case, Bloomfield bothered to have sememes in his system at all. But the reason becomes clear when we recall that for Bloomfield a morpheme was composed of phonemes. Certainly, sequences of phonemes could not themselves be units of meaning as well; so it was necessary to adopt an alternative, namely that these sequences, the morphemes, have meanings; and these meanings were the sememes. Nowadays, on the other hand, many linguists view morphemes not as composed of phonemes but as represented by phonemes and combinations of phonemes. For those of us who hold this view, morphemes are not units of expression so the notion of a one-to-one correspondence between morphemes and sememes would amount to the same thing as making the sememes quite useless, except for those who take pleasure in multiplying entities praeter necessitatem. Now, of course, the fact that a name exists, even one given by Bloomfield, does not compel us to go out and find something to use it for. Nor should we be unduly influenced by its suffix, so that we suppose we must find an emic type of entity in the area of content, having those same important properties as other emes we have come to know. Of course, there is no objection to anyone's setting out to look for such an entity, provided he maintains the proper degree of caution. On the other hand, this was not the approach that I used. It