Rivista di Psicolinguistica Applicata/Journal of Applied Psycholinguistics volume:15:59-74.
This paper describes the statistical analyses performed to validate the THAM-3, a metalinguistic ability test for late adolescence and adulthood, translated and partially adapted from the original Italian version TAM-3 (Pinto & Iliceto, 2007) into Spanish (Pinto, Titone, & Gonzales Gil, 2000; Lasagabaster, Merino, & Pinto, 2015). The test consists of three subtests: Comprehension, Acceptability and Figurative language, and its scoring system is based on the distinction between Linguistic (L, measuring metalinguistic awareness at the implicit level) and Metalinguistic (ML, measuring metalinguistic awareness at the explicit level) scores. One hundred and fifty students recruited at the University of Valladolid in Soria (Spain), aged 18 to 37 years (mean age = 21.1; SD = 2.7), took the THAM-3 and Raven’s SPM38 (Raven et al., 1998). Skewness and Kurtosis as measures of the shape of the distributions showed that the data were normally distributed. Reliability analyses, measured by Cohen’s Kappa (Cohen, 1960) and Cronbach’s alpha, presented relatively adequate internal consistency estimates. Convergent validity, measured by the correlations between all the THAM-3 measures and the SPM38 scores, showed statistically significant correlations. The results of Confirmatory Factor Analysis (CFA, Brown, 2006) supported the theoretical two-factor structure of the original test, assigning the ML subtests’ scores to the corresponding Metalinguistic latent factor, the L subtests’ scores to the corresponding Linguistic latent factor, and explaining the inter-correlations between the observed variables. The final model provided a good fit to the empirical data, although only the Metalinguistic factor emerged as a unitary construct, whereas the Linguistic factor proved poorer. On the basis of these results, the THAM-3 can be considered as a valid measure of metalinguistic abilities for all subjects having Spanish as their first language.