The ability to integrate information from diverse texts and to detect logical implications of the integrated information is fundamental to the understanding process. This paper shows that identifying and configuring relevant facts in order to support hypothesized inferences is extremely difficult unless the facts have been committed to memory. Simply reading relevant texts for familiarization and then referring to them as needed provides an inadequate basis for deductive logic. Further, apprehension of the logical configuration of fads underlying a particular inference (...) can be an essentially automatic process for the reasoner who has structured the facts appropriately in memory. A model of search and memory mechanisms is proposed. It accounts for accuracy and reaction time data, as well as individual differences in inference evaluation. The implications of the superiority of memorization over familiarization‐plus‐referencing as a basis for learning and reasoning are discussed. (shrink)

The ability to integrate information from diverse texts and to detect logical implications of the integrated information is fundamental to the understanding process. This paper shows that identifying and configuring relevant facts in order to support hypothesized inferences is extremely difficult unless the facts have been committed to memory. Simply reading relevant texts for familiarization and then referring to them as needed provides an inadequate basis for deductive logic. Further, apprehension of the logical configuration of fads underlying a particular inference (...) can be an essentially automatic process for the reasoner who has structured the facts appropriately in memory. A model of search and memory mechanisms is proposed. It accounts for accuracy and reaction time data, as well as individual differences in inference evaluation. The implications of the superiority of memorization over familiarization‐plus‐referencing as a basis for learning and reasoning are discussed. (shrink)

The fan effect says that “activation” spreading from a concept is divided among the concepts it spreads to. Because this activation is not a physical entity, but an abstraction of unknown lower‐level processes, the spreading activation model has predictive but not explanatory power. We provide one explanation of the fan effect by showing that distributed neuronal memory networks (specifically, Hopfield networks) reproduce four qualitative aspects of the fan effect: faster recognition of sentences containing lower‐fan words, faster recognition of sentences when (...) more cues are provided, faster acceptance of studied sentences than rejection of probes, and faster recognition of sentences studied more frequently. These are all a natural result of the dynamics of distributed associative memory. (shrink)