When the Path is a Source–Goal pattern(from W toZ), the second argument ofGOhas its own Function–argument
structure—which is why Path-prepositions take syntactic objects. We can notate the composite as (18), where Z is the
Source and W the Goal.
The characteristic inference fro m(18) can be stated infor mally as (19).^186
(19) At the beginning of (18),BE(X,W)
At the end of (18),BE(X,Z)(19) is independent of semanticfield. However, the consequent clauses of (19) lead to the characteristicfield-specific
inferences for BE described above, which add further aspects of interpretation.
Inmuchoftheliterature, Paths are recognized only inthecontext ofmotionalong a path, and therefore Pathhas been
thought of as a partly temporal notion (e.g. Hinrichs 1985 and Verkuyl 1993). (18) recognizes Path as a distinct onto-
logical type, and motion or change over time is expressed solely by the functionGO. Paths themselves are atemporal,
and can appear as arguments of other functions, including the two kinds of state-function shown in (20).
(20) a.EXT(x,Path)
Non-temporal extension:The road goes across the river.
b. ORIENT(x,Path)
Orientation:The sign points across the river.The inferencesfro mthese functionsare quitedifferentthan fro mGO. In a motionevent(GO), X is located at different
parts of the Path at different times. In a state of extension like (20a), different parts of X occupy different parts of the
Path, all at once.^187 In a state of orientation like (20b), the sign is oriented along the Path, but it neither travels along it
nor occupies it. Without an independent category of Path, it would be impossible to unify these disparate uses,
showing why they can all be expressed by the same preposition.
362 SEMANTIC AND CONCEPTUAL FOUNDATIONS
(^186) Making it more formal requires characterizing‘beginning’and‘end’, for instance as in Jackendoff (1991; 1996b).
(^187) It is often contended (e.g. Talmy 1996a) that extension is“metaphorical”or“fictive”motion, giving the sense of an observer scanning the extended object. Although this
has some intuitive appeal, I just don't see how it can account for the difference in inference patterns. Jackendoff (1996c) offers a formal decomposition of GO and EXT
that brings out their similarities and differences.