Thinking about knowledge learned from instruction and experience: Two tests of a connectionist model

Abstract Precise characterizations of thinking processes were tested quantitatively in 5 experiments. Newly learned bodies of knowledge were shown to be subject to thinking processes that were simulated by spreading activation through an associative network: In 3 domains, thought patterns of 717 Air Force recruits were successfully predicted by a power algorithm spreading activation process, which always produces the first principal component of the associative network, implying that specific form for the thinking process. Similarly, preexperimentally learned knowledge was shown to be organized into distinct, discrete subrepresentations corresponding to the principal components of the knowledge network's associative matrix: In 120 undergraduates, principal components with large loadings were successfully induced and completed when incomplete, whereas components with small loadings were squelched.