and its strong linkage to perception. At the same time, Lakoff and Johnson's insight about the pervasiveness of
metaphor still stands, if in a more limited fashion.
These crossfield parallels raise an interesting question of polysemy. Doeskeep, for instance, have a single sense that is
neutral as towhatfieldit appears in? Or does ithavefour relatedsenses? I a minclined totake thelatter route,because
of the syntactic and lexical peculiarities in eachfield. For instance, in three out of the fourfields above,keeprequires a
direct objectand a further complement; but in the possessivefield,keeptakes only a single argument, the directobject.
Goappears in three of the fourfields, but we cannot say *The meeting went from Tuesday to Mondayin the scheduling
sense—we can only say was changed/moved.Changeappears with the simplefront-tosyntax only in ascription and
scheduling; in the spatialfield we have to saychange position/locationand in the possessionalfield we have to saychange
hands. We also have to know that in English, we sayon Mondayrather thanin Monday, the latter certainly a logical
possibility. All these littledetails haveto be learned; they cannotbe part of the general mapping thatrelates thesefields
to each other. This means that each word must specify in whichfields it appears and what peculiarities it has in each.
Thus we are dealing with semiproductive alternation in the sense of Chapter 6 again. If this be polysemy, so be it.
I have described these as senses related by feature variation. This differs fro mthe usual view in cognitive linguistics
that(14)–(16)are derivedfrom(13),thespatialfield, bymetaphoror image-schematransformation(Lakoff1987). The
two views are compared in (17).
(17) a.Conceptual semantics
toin (13) =TOSpatial
toin (14) =TOPoss
TOis a path-function that isfield-neutral, and the subscripts specialize it for afield
b. Cognitive linguistics
toin (13) =TO
toin (14) = FPoss(TO)
TOis a spatial path-schema, and F is a function that maps thefieldof spatial images into possessional imagesThat is, cognitive linguistics tends to view cross-field parallelisms as derivational. Hence the polysemy ofkeepis a
multiply branching chain, rather likesmoke(section 11.3). By contrast, I view the mas parallel instantiations of a more
abstract schema (Jackendoff 1976, 1983: ch. 10, 1992a; Jackendoff and Aaron