The status of metalinguistic knowledge in instructed adult L2 learning

Abstract

This study investigated second language metalinguistic knowledge, or explicit knowledge about the second language, in English-speaking university-level learners of German and Spanish. The status of metalinguistic knowledge in relation to the individual difference variables of language-learning aptitude, working memory for language, and participants' language-learning history was identified. Language-learning experience in formal settings was found to be the strongest predictor for levels of metalinguistic knowledge attained by the participants. Moreover, it was found that despite a significant relationship with language-learning aptitude, metalinguistic knowledge is separable and distinct from both aptitude and working memory. In conclusion, suggestions for further research are put forward.

Keywords:

• language-learning aptitude
• language-learning experience
• metalinguistic knowledge
• second language instruction
• second language learning
• working memory

Acknowledgements

This study was supported by the University of Essex Research Promotion Fund. We are grateful to Phil Scholfield for statistical advice. We would also like to thank Charles Alderson and Phil Scholfield for their constructive comments on an earlier version of this paper.

Notes

∗∗significant at the .01 level (two-tailed);

∗ significant at the .05 level (two-tailed).

1. While this is probably the most common understanding of metalinguistic knowledge in the adult L2 learning literature, work primarily concerned with metalinguistic development in children and/or bilinguals may assume slightly different definitions. For instance, Bialystok (1994, 2001; Bialystok & Ryan, 1985) does not equate metalinguistic knowledge with conscious awareness, and Birdsong (1989) does not regard conscious awareness as a defining characteristic of knowledge about language.

2. A principal components analysis is a data-reduction technique, which is used ‘to discover components that underlie performance on a group of variables’ (Hatch & Lazaraton, 1991, p. 491). Put differently, a principal components analysis is aimed at identifying subsets of variables that are highly correlated among themselves, but not correlated with variables in other subsets. If a subset of variables predominantly correlates with each other, either positively or negatively, it is said to load on the same component. It is the task of the researcher to decide how many subsets of mutually correlated variables (or components) to include in the final matrix. Typically, their decision is informed by the eigenvalue of each component, that is, an estimate ‘of the proportion of variance in each observed variable explained’ (Colman & Pulford, 2006, p. 142). An eigenvalue larger than 1 indicates that a component captures useful information and should be taken into account. In summary, then, the aim of a principal components analysis is to identify the smallest number of underlying components that needs to be recognised to enable the greatest amount of variance to be accounted for.

3. Like principal components analysis, multiple regression is also a data-reduction technique. A multiple regression analysis can be used to discover ‘how well we can predict scores on a dependent variable from those of two or more independent variables’ (Hatch & Lazaraton, 1991, p. 480). Thus, a multiple regression analysis includes several independent or predictor variables, but only one dependent variable. The analysis reveals which of the independent variables are related to the dependent variable and which are not, while at the same time discounting relationships among the independent variables themselves. R² shows the percentage of variance in the dependent variable explained by the predictor variables taken into account at that stage of the analysis. A stepwise multiple regression allows the researcher to see which predictor variable has the greatest effect, the second greatest effect, and so forth. Variables are weighed against each other successively, discounting the effect of the previous stronger variables. In the final model of the stepwise multiple regression, R² shows the total percentage of the variance explained when all significant predictor variables are considered.