Semantic patterns of words interact in a syntatlc structure. That is, in encoding, the meaning of any high-order constituent (phrase or sentence) Is created from the serial coding of the word elements involved.In decoding, the process is reversed; codings for higher-order constituents are some resolvable function of the component word codlings. Semantic 2 We are implicitly assuming that there exists a universal set of semantic features for any particular domain, bit that the pattern, or the representation of a given word In terms of these features, may be cultually unique. Thus, it is possible that, in contrast to our own culture, father is given a zero coding on the potentcy feature, In deference to the maternal uncle who receives a plus. While the feature is commonly relevant to both cultures, words may align themselves differently with respect to *t. Tables 3 and 4 respectively. In Table 4, feature codings apply to the speaker, or the left-hand member of the role-pair.
Tables 3 and 4 about here
SubjectsThe subjects for this study were 53 high school students from Thailand, studying in the United States for one year. There were 27 males and 26 females.
The QuestionnaireThe construction of the questionnaire matrix was accomplished by
writing (in Thai) the pronouns as column headings and the role-pairs as row headings. This resulted in a matrix of 60 x 14 cells. The Ss were asked to make a Judgment of each cell and to indicate with a plus, if that combination was appropriate ^congruent), with a minus if it was not appropriate (incongruent), and with a zaro if it was acceptable (permissible),The behavioral context for each combination was described as being that of "talking to," e.g., 'Father in talking to Son will use the pronoun to represent himself." Table 3 Coding Table 5.
AnalysisThe Ss' data matrix was used as the criterion for Judging the adequacy of the predicted matrix and, hence, the adequacy of the a priori analysis.Comparisons between predicted and obtained matrices yielded a residual matrix.If the subjects' data is predicted perfectly, the residual matrix will be a zero matrix. On the other hand, if the data matrix is not predicted perfectly, the residual matrix will contain some "l's," proportional to the number of mistakes made in prediction. Plus l's and "-l's" are considered "patchable" errors, while a "90" arbitrarily designates an "unpatchable" error, A "+l" means that the combination was predicted as a "0" but Ss Judged it as a "+1"; a "-l" means that it was predicted as a "0" whereas Ss Judged it as a "-1"; a '90" means that It In order to compare the two models, the subjects' data matrix was factor analyzed by the Principal Cevpnnents method. Six factors (Varlmax rotation) were found to account for 94% of the total variance. Table 6 shows these six factors, together with their highest loadings.Insert Table 6
