Latent Variable Mixture Modelling (LVMM) for language learning motivation research

Purpose: Research on language learning motivation has made great of use of questionnaires for data collection and quantitative data analysis both for theory formation and testing. Yet, the quantitative tradition has been rightfully critiqued by a lack of focus on the individual. In this paper, we explore some possibilities offered by Latent Variable Mixture Modelling (LVMM), as a statistically defensible and valuable method for analysing language learning motivation data.

Methodology: As a person-centred methodology, LVMM takes as its starting point the hypothesis that there is heterogeneity in any given population. With respect to motivation research, this permits the researcher employing LVMM analysis to identify sub-groups of participants characterised by distinct motivational profiles. Such sub-groups are known as latent classes since groupings are not defined by observable differences – for example, gender or age group. We apply LVMM to analyse the motivation and proficiency of English learners from Bogotá, Colombia.

Findings: Our findings show the existence of five learner profiles, with differing levels of motivation, proficiency and, as well, differing relationships between motivation and proficiency. The value of taking the LVMM approach is highlighted.

Originality: Unlike variable-based data analysis methods, LVMM enables the researcher to uncover multiple and diverse motivational profiles of language learners and to glean sophisticated insights into the balance of associations between motivational traits in different groups. Subsequently, it is possible to propose how these diverse groups of learners can be more effectively taught, which has substantial ethical implications.