Acquiring Generalizations to Organize Human Databases
Abstract
Five experiments are briefly described in this report, and plans for three further experiments are set forth. We are investigating the consequences of people forming concepts or categories after they've been exposed to a collection of instances (stimulus objects, patterns, events) for which certain features are highly inter-correlated. One primary consequence is that once such regularities are discovered, they are exploited to greatly simplify the recording of new instances into memory. In particular, new instances come to be recorded simply in terms of their belonging to a familiar category plus having a few distinctive features. We've found strong evidence for this kind of coding of instances. A second consequence is that once the category (correlated features) of an instance is identified, the person can focus his learning efforts on recording the distinctive features of the instances, resulting in better memory for this information. In a short-term memory experiment, we've found strong evidence for this strategy. A third consequence of people learning consistently- correlated features of stimuli is that it affects the way they judge the similarity of two instances.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1988
- Accession Number
- ADA202964
Entities
People
- Gordon H. Bower
- John Clapper
Organizations
- Stanford University