||Three experiments were conducted to explore the nature of information that becomes activated during anaphor resolution when the anaphor is weak. A number of factors have been shown to influence the reactivation of antecedent information during reading, including semantic overlap and contextual support. Memory-based models of text comprehension assume that any information related to the current input could potentially become activated on the basis of semantic overlap. Studies have shown that if the context supporting an unnamed concept from general world knowledge is strong enough, it may even be instantiated into the text representation in place of an explicitly mentioned antecedent. Experiment 1a was a continuation study, which demonstrated that the passage materials had been constructed such that the context was sufficient to activate the target concept-i.e., the explicitly stated antecedent. This experiment also served as an offline measure of activation. Experiment 1b, a reading time study, demonstrated that a very weak anaphoric reference was sufficient to activate a target concept when it was explicitly stated in the passage, and when it was only implied by context. Experiment 2, a naming study, was designed to confirm that the activation of these concepts occurred during reinstatement. However, no significant effects were observed. The conclusion was that while weak anaphoric references seem to be able to facilitate integration of a target concept, they may not provide sufficient activation of the concept itself, but rather may be activating other elements of the context from the passage.