The Evolution
46
"orange juice" and the relationship that "drink" has with "orange juice."
Thus they might say that "eggs" operates by spreading activation whereas
"drink" tends to yield "orange juice" by a kind of restriction; but if
one is going to do this then why not go all the way and postulate the
existence of two types of links? After all, to introduce a type of
restriction for use with spreading-activation is only an implicit way
of saying what should be said explicitly, namely that the most parsimonious
explanation of the data involves using two types of links. This would
allow the abandonment of their "interlinking" explanation of the order
effect, an explanation which required that there be spreading-activations
emanating from concepts such as "red," but that these spreading-activations
be inefficient because of a lack of interlinking, an argument that was
somewhat clumsy to begin with.
It is a clumsy argument in that it requires that the concept
"striped" in the intersection "African striped" activate not merely
"striped" concepts such as "zebra" and "barber pole," etc., but also
indirectly such concepts as "haircut" and "scissors" (both linked to "barber
pole"). To assume such a spreading-activation is not only counter-productive
in that it implants in memory expectations that are gratuitous and
misleading; it is also unnecessary. A simple restriction of some kind
would do as well without creating all of the unjustified expectations.
It would be interesting to see if a letter string such as
STRIPED can facilitate lexical decisions on letter strings such as
HAIRCUT and SCISSORS (Meyer & Schvaneveldt, 1971), If an extended