214 CHAPTER 6
competition suggests not only that it is possible for a feature characteristic of
both dogs and cats to dominate but that in fact it will occur. As a result, we can
predict that in some instances, on a probabilistic or statistical basis, a person will
call a dog a cat and vice versa: “The connecting strengths of the association be-
tween the string of phonemes characterized as ‘dog’ and the features characteriz-
ing ‘dog-ness’are insufficient to provide consistency” (Williams, 1993, p. 559).
We can apply the same principles to grammar and usage errors. For example,
we may instruct students on the ungrammaticality ofThe reason is becauseand
I feel badly, providing them with the correct forms:The reason is thatandI feel
bad. The lesson will be stored in the brain as associated structural patterns that
we can symbolically represent in the following diagrams:
In both cases, there are two potential targets activated by the trigger. After
our instruction, it is possible that students will select the correct target. How-
ever, because they are immersed in a language environment where the over-
whelming majority of people, regardless of education, produce the erroneous
strings, the connection strength for the incorrect form will be far greater than
the connection strength for the correct form. On this basis, we can predict with
confidence that instruction will have a limited effect on performance unless it is
reinforced consistently both in and out of school—and unless our students are
motivated to change their language.
Cognitive grammar proposes that language acquisition is intimately allied
with experiences and internal representations of reality. Sentence production and
grammaticality, indeed language as a whole, are “tied to associations between
various patterns of regularity generalized through interaction with the environ-
ment” (Williams, 1993, p. 561). From this perspective, the act of producing an ut-
FIG. 6.1. Graphic Representation of Connection Strengths.