Running Head Word Association Spaces 1 Word Association Spaces Semantic Spaces based on Free Association that Predict Memory Performance Send correspondence to
asymmetries in the word associations cannot be predicted. Therefore, in our first measure for
associative strength we take:
(1)Sij=Aij+Aji
S(1)ij is equivalent to adding forward strength to backward strength. This measure is of
course symmetric so that S(1)ij = S(1)ji. This measure is indexed by (1) because it based on only
the direct association between i and j and involves only one associative step going from i to j. In
the norms of Nelson et al. (1998), subjects were only allowed to give the first response that came
to mind. The second strongest response in one subjects’ mind might be elicited by another
subject or it might not be elicited at all if the first response is a strong associate. Therefore, the
S(1) measure might be underestimating the associative strength between two words especially in
cases where the measure is zero (Nelson et al., 1998). In the second measure for associative
strength, we take:
(2)(1)(1)(1)Sij=Sij+∑SikSkj
k
This equals the forward plus backward plus mediated strength through other associates.
Note that this measure involves the direct strength between i and j as well as the indirect strength
by summing over all paths from i to k to j, the product of the symmetric associative strengths
between i and k, and k and j. These indirect associative strengths involve the two step
probabilities of going from i to j and vice versa and hence the index (2). Research has shown that
the indirect associative strengths play a role in cued recall (Nelson, Bennet, & Leibert, 1997;
Nelson & Zhang, 2000) and recognition (Nelson, Zhang, & McKinney, submitted). For
example, Nelson & Zhang (2000) found that including the indirect associative strengths in a
measure for associative strength significantly increases the explained variance in the extra-list
cued recall task.
热门文档