skip to main content skip to footer

Principal Component Analysis of Ratings of Some Deviant English Sentences

Author(s):
Danks, Joseph H.; Lewis, Charles
Publication Year:
1970
Report Number:
RB-70-12
Source:
ETS Research Bulletin
Document Type:
Report
Page Count:
90
Subject/Key Words:
National Science Foundation (NSF), Office of Naval Research, Comprehension, English, Factor Analysis, Linguistics, Statistical Analysis

Abstract

The comprehension of deviant sentences is dependent on several linguistic variables. Grammaticalness (G), meaningfulness (M), and familiarity (F) are three variables which are potentially such. In order to study the effect of violating those variables upon Ss' responses to deviant sentences, 85 deviant and 15 correct sentences were assigned to eight groups representing all combinations of two values ("correct" or "deviant") on these three variables. The 100 sentences were given to four equal groups of SS (total N=112), who rated each sentence from 0 to 10 on the basis of either grammaticalness (G*), meaningfulness (M*), familiarity (F*), or ordinariness (0*). The data of the first three groups were then combined into an 84 by 100 matrix. A principal components analysis was performed on the crossproduc matrix with a varimax rotation. Four interpretable Factors emerged, accounting for 89 percent of the variability. Factor I was a general comprehensibility factor in the factor loadings, related to changes in all three variables. However, the familiarity Ss scored highest on Factor I. Factors II and III represented G-G* and M-M*, respectively, in both factor loadings and factor scores. Factor IV corresponded to the F variable in the factor loadings, but was uninterpretable for the factor scores. (Author/FWB)

Read More