Polysemy and the lexicon
Compromise #1: Going half-way
As little polysemy as possible
Rampant polysemy might be reduced to the extent that small differences in meaning might be attributable to the meanings of surrounding words, in association with general conceptual knowledge of how the world works; the meaning differences would therefore not need to be enshrined in the mental lexicon as distinct polysemes. Consider, as an example, the properties of the path followed by the trajector when you “walk over a field”, “fly over the ocean”, or “jump over the wall”, with respect to such matters as whether the trajector is in contact with the landmark entity, or whether the path has a (roughly) rectilinear or a curved shape. It could be argued that these aspects follow from what we know about the situations involved. Walking involves contact with a surface, flying involves lack of contact, jumping involves an up-down path. It would follow, therefore, that we need to recognize, for these examples, only one sense of over, namely, as designating a path which goes from one side of a landmark to the other, in a configuration which is appropriate for the kind of trajector, its manner of motion, and the kind of landmark entity, in question.
Take, as another example, the variable reference of the word window, cited above. This might be handled in terms of the active zone phenomenon (Langacker 1990). Whenever an entity is involved in a predication, typically only some facet, or facets, of the entity will participate in the predication, whereby the relevant active zone can be inferred on the basis of general conceptual knowledge of the entity and of the kind of predication that it participates in. Given what we know about windows, and about what it means to paint something, we can reasonably infer that “paint a window” implicates only the “paintable” facets of the window, namely, its wooden frame. Likewise, when we “brick up a window”, we can infer that the reference of window has got to be a vacant gap in a wall, not a wooden frame, or a pane of glass.
Appeal to the active zone phenomenon begs the question of what the basic, or “real” meaning of a word is. What kind of thing does window before the active zone phenomenon kicks in– refer to? It cannot be the sum of its active zones– wood, glass, and a hole in a wall do not add up to a coherent category. Rather, we want to say that the word window provides access to a body of conceptual knowledge, concerning the typical location, the typical function, and the typical manner of construction of the things in question. It would follow that window, as a decontextualized word, does not actually refer to a thing at all, but rather stands at the hub of a conceptual network; for contextualized uses, the relevant concept would be conjured up in accordance with the specifications of a particular scene (cf. Barsalou 1987). Our impression that the word does have a determinate reference would be due to the fact that certain contextualized uses– such as open the window– frequently re-occur, thereby entrenching the relevant conceptualization.
As Langacker noted, the active zone phenomenon gradually shades into the more general process of metonymy (see also Taylor 2003b: 127). The named entity functions as a reference point for accessing the intended referent. The intended referent may be a part of the named entity, or it may be something that is closely associated with it within a given knowledge configuration. Interestingly, processes of metonymic extension typically apply across the board to words of a certain semantic category. A well-known example concerns the names of institutions. University can be used to refer to the institution, to the building which houses the institution, and even to a person who acts on behalf of the institution. Compare: The University was founded in the 19th century, I parked near the University, and The University just telephoned. Similar metonymies can be observed with institution-type words in general: school, museum, institute, etc. For another example, consider the fact that path prepositions in English can in general be used to designate a place at the end-point of a real or imaginary path. Compare: walk round the corner vs. live round the corner; run up the hill vs. be standing up the hill (for further polysemization patterns of the English prepositions, see Taylor 1993). To the extent that we can identify processes which apply across the board to sets of words, the different senses that are thereby generated may not need to be separately listed. This kind of “generative” approach to polysemy has been pursued by, amongst others, Pustejovsky (1991), as well as by Bierwisch and the “two-level” theorists (Bierwisch 1981; Wunderlich 1993).1
1. The “two-level” approach is of interest, in that it attempted to address some of the problems raised above in connection with Searle’s proposal for unitary word meanings. The two-level theorists did make proposals for unitary “semantic representations” for a selected number of words, above all, the prepositions, and for a few other examples, such as institution-type words. In addition, mechanisms were proposed for explicitly deriving “conceptual” (that is, contextualized) interpretations from these semantic representations. I have critically reviewed the two-level approach elsewhere (Taylor 1994, 1995, 2003b). One of my criticisms was that the mechanisms for conceptual elaboration often failed to predict the sometimes quite specific properties of the words in context. These specific properties could only be derived by appeal to ad hoc principles of conceptual interpretation. Appeal to such ad hoc principles, however, would be tantamount to recognizing the distinctiveness and independence of the different readings of the words in question, that is, that the words were indeed polysemous.
