Semantic Distance and the Verification of Semantic Relations

This experiment is based on the following paper:

Rips, L., Shoben, E., & Smith, E. (1973). Semantic distance and the verification of semantic relations. Journal of Verbal Learning and Verbal Behavior, 12, 1-20. Reprinted by permission of Academic Press.


Early in the study of semantic memory, two different models were proposed: network and set-theory. Network models assumed that words were independently represented in memory and connected by a complex network through which relationships between words were defined. For example, the sentence A robin is a bird is represented by two units (robin and bird), the connection between them (is a). Set-theory proposed that semantic memory is organized into overlapping sets of elements, where the elements could be exemplars or attributes of the concept expressed by the sentence.

When a person recalls semantic information, some memories take more time to recall than others. For example, people can verify that A robin is a bird faster than A robin is an animal. To explain this difference, researchers proposed that robin and bird are semantically more closely related than robin and animal, a phenomenon described as the semantic distance between them.

Each model of semantic memory had a different explanation for semantic distance effects. Network models proposed that the connection between robin and bird is more direct than between robin and animal, and so the time it takes to verify a sentence is proportional to the number of connections required to link the words. Set models proposed that verification requires a search through a set of elements. The size of the set determines how long the process takes. Because the set of birds in semantic memory is smaller than the set of animals, verification times will differ.

Nevertheless, both explanations for semantic distance effects share a common assumption. Each model proposes that semantic memory is organized to reflect the categorical structure of our language. Both models predict that it will take more time to verify a relationship between items that cross categorical boundaries than for items represented at the same structural level. Because a robin is an item in the subset of birds, and bird is an item in the subset of animals, then to verify that a robin is an animal will take longer than to verify that a robin is a bird.

In this study, Rips, Shoben, and Smith reported that semantic distance effects do not always occur when items cross categorical boundaries. In one case when the category was mammal instead of bird the effect was reversed.


The full text describing the experiment is available here.

The experiments are available below. (48.2 KB) (77.2 KB)