Global co-occurrence memory models, such as HAL and LSA, encode language input (words as symbols). These symbols form the basis of high-dimensional word vectors which have the characteristics of distributed representations: graceful degradation, concepts formed by a large array of elements, and straightforward generalization. An advantage of these models is that they use learning procedures that scale up to real world language problems. We show that the output of a global co-occurrence learning algorithm produces virtually the same output as an SRN and that relatively few trials are required for learning. The global co-occurrence approach suggests a different view of similarity and has implications for the symbol-grounding problem, catastrophic interference, and the relationship between episodic associations and categorical knowledge.