Resumen de Normative testing and bilingual populations
The problems and procedures of producing test norms for bilingual populations are considered. The difficulty of defining sub‐populations for separate test norm production (e.g. first language and second language populations) causes problems of empirical classification. Problems of classification of population groups by skill level and by functional usage are analysed. Four practical solutions are detailed: school classification, person classification, test score classification and language background classification. Criticisms of these four methods are considered, and a discussion of triangulation procedures using a combination of methods. The construction of a Welsh language spelling test is used as an example of a procedure for population classification using latent class analysis, a form of cluster analysis. The empirical classification, using a language background questionnaire, allocated the 2,262 children into four groups which implicitly questions the usual a priori procedure of L1/L2 classification for test norm purposes.
