Cognition and FDG 751976); and the permanent store of our knowledge would correspond to the
second kind. A further distinction within long-term memory, i.e. that be-
tween episodic memory (in which we recall particular episodes or events)
and semantic memory (in which we integrate all our knowledge about the
world; Tulving 1972) does not add much, since both kinds may be useful
for interacting with the interpersonal, representational and expression lev-
els.
Long-term knowledge makes itself very evident in vocabulary interpre-
tation, when we resort to all the encyclopaedic information stored in
individual lexical items. Descriptions of how this encyclopaedic informa-
tion deep in our minds can be used and linked to single words appear in the
work of some cognitively oriented authors. Langacker (1987: 162), a cog-
nitive linguist, suggests that lexical units are merely points of access, and
that therefore “concepts are simply entrenched cognitive routines”. More
precisely:
The entity designated by a symbolic unit can (...) be thought of as a point of
access to a network. The semantic value of a symbolic unit is given by the
open-ended set of relations ⎯ simple and complex, direct and indirect ⎯ in
which this access node participates. Each of these relations is a cognitive
routine, and because they share at least one component the activation of one
routine facilitates (...) the activation of another. (Langacker 1987:163)Langacker does not suggest this description out of the blue. It is based
on proposals by psychologists and computer scientists for models that in-
volve semantic networking. The main hurdle is how to determine, for every
access node, the extent of the corresponding conceptual network. Since it is
possible to think of a network in terms of an unending chain of associations
between nodes, there must be some mechanism, congruent with current
psychological evidence, that helps determine which nodes get a place in the
network, in other words which nodes are ‘relevant’. The keyword here is
activation. This is a term used in connectionism to designate the degree of
strength that the different nodes in a neural network may have.
Models of cognition that incorporate connectionist ideas use this term
profusely. Anderson (1983), for instance, in his ACT model, distinguishes
declarative memory from production memory and working memory. To re-
trieve information from declarative memory so that we can use it in our
working memory, a certain amount of energy is needed in the form of acti-
vation. The various modes of activation for concepts in a network explain
the link between words and encyclopaedic knowledge dynamically.