task) which then is used to formulate transient connections (i.e. the actual responses to stimuli in the experiment).
More generally, any sort of episodic memory (in the sense of Tulving 1972 raises this problem: an episodic memory is
byhypothesissomethingthatisremembered onthebasisofoneoccurrence,and itusuallyinvolvesobjects, places, and
people with which one is familiar. One cannot just encode an episodic memory by gradually strengthening the
associations among the involved characters: the precise relations among the characters are crucial, and they are
established at once.
I will offer no speculations on how this transfer fro mworking me mory to structured long-ter m me mory is
accomplished,giventhat onecannolonger speak simplyof“shipping information offtosomeotherarea ofmemory.”
As withtheother threecases of thissection, I leave itas a challenge for neuroscience.And as withtheother cases, itis
not something that can simply be disregarded by ignoring language.
To su mup, a theory of how language is instantiated in the brain must grapple with four proble ms that arise fro mthe
combinatoriality of language: the massiveness of binding in linguistic structure, the problem of multiple instances of a
known unit in a novel structure, the necessity for encoding and instantiating typed variables, and the relation between
long-term and short-term memory encodings of structure. These problems are not exclusive to language, but they
certainly come to the fore in dealing with the linguistic phenomena that linguists deal with every day.
A furtherissuearisesfromcombinatoriality: thatof learningtheprinciplesthatgovernit.This requiresa wholechapter
of its own.