ERIC Number: ED130652
Record Type: RIE
Publication Date: 1966-Jan
Reference Count: 0
A Comparison of Two Tree Construction Methods for Obtaining Proximity Measures among Words. Number 47.
The prediction that two different methods of constructing linear, tree graphs will yield the same formal structure of semantic space and measurement of word proximity was tested by comparing the distribution of node degree, the distribution of the number of pairs of nodes connected y times, and the distribution of adjective degree in trees constructed by the alternate methods. Fourteen male college students were each asked to construct two tree graphs from a list of 24 adjectives. For the first tree, the subject selected a pair of similar adjectives from the list and connected them with a line. He then sequentially connected the remaining adjectives to the pair according to their degree of similarity. Using the second method, subjects sequentially added single adjectives to the original pair, as in the first method, or started a new tree with a new similar pair or connected two trees together until all 24 adjectives were connected. Comparison of the formal structural distributions investigated showed no difference between the two methods. The methods did differ, however, with respect to their ability to yield the true ordering of the strength of association between paired adjectives. (KB)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Science Foundation, Washington, DC.; National Inst. of Mental Health (DHEW), Bethesda, MD.
Authoring Institution: North Carolina Univ., Chapel Hill. L.L. Thurstone Psychometric Lab.
Note: Archival document