Abstract
M.A.
Intelligence testing in cross-cultural assessment settings where traditional intelligence tests are used, often reveals inappropriate test content and invalid test scores for various cultures, language or socio-economic groups (SES). This moved some test users to conclude that traditional intelligence tests are not culture fair and are not suitably constructed for use across different cultures. As a result, these traditional intelligence tests often show invalid score variability when used across different socio-economic groups and culture. The purpose of this study was to investigate the suitability of the Snijders-Oomen Non-verbal Intelligence Test-Revised (SON-R) on a group of primary school learners from different language, socio-economic status (SES) and gender groups and grade levels in the Johannesburg metropolitan area. The aim of the study was to establish psychometric
properties of the SON-R and to demonstrate that the SON-R can minimise the influence of home language, SES, age, grade level and gender on sub-test scores. Five postulates and four specific hypotheses were formulated for the purpose of the study. The sample of this study consisted of 400 primary school children from schools in Roodepoort and Melville. The test was administered to all subjects individually in their respective home languages within their school premises. Various statistical procedures, e.g. multiple analysis of variance (MANOVA), analysis of variance (ANOVA), Principal Axis Factoring, Pearson Correlation coefficients, Cronbach alpha internal consistency coefficients and post-hoc tests (Scheffe and Dunnett T3) were used
to analyse the data. The findings on item difficulty analysis, in terms of the first and second mistakes committed on each sub-test, shows that items of the SON-R generally correspond to the theory of item difficulty underlying the items of the SON-R. Furthermore, the difficulty levels of the sub-test items appeared to be compatible with the cognitive levels of the subjects. The findings on the relationship of socio-economic status, age, as well as gender, with the sub-test scores were not statistically significant, indicating that sub-test scores of the SON-R
were not influenced by those variables. The relationship between home language as well as grade evel with sub-test scores was statistically significant. The two variables have caused significant differences on sub-test scores of the SON-R. Cronbach alpha internal consistency coefficients were relatively high. In general, the correlation coefficients between sub-test scores and academic performance in school subjects were high, indicating significant predictive ability of the SON-R for scholastic purposes. A single factor solution for the combined group data, indicated one general factor of non-verbal intelligence for all four language groups. The study has raised research questions, which should be considered for future research on the SON-R. In general, the data has successfully supported the postulates and the specific hypotheses of the study.