LLAMAs, ALPACAAs and testing aptitude Seminar

Event details

Language learning aptitude has received increasing attention in recent years (Li, 2016; Reiterer, 2018). Most models of aptitude and aptitude tests suggest there are different components that make up language learning aptitude (Carroll & Sapon 1959; Meara, 2005; Linck et al, 2013) and aptitude may be related to working memory (Wen et al, 2017). Since Carroll & Sapon’s early MLAT tests (1959), a number of alternative tests have been constructed (Pimsleur, Hi-Lab, CANAL-FT, LLAMA). One problem facing researchers has been the availability of such tests. Hence, the freely available LLAMA tests (Meara 2005) have been widely used. However, the LLAMA tests were originally designed as exploratory tools for research methods training and have not been validated or standardised (Bylund&Bokander 2019, Granena 2013, Rogers et al, 2017). A number of design flaws have been identified in the original LLAMA tests including difficulties with the interface, different scoring methods across the tests, ceiling/floor effects in LLAMA E in particular, and some errors in items. Using OpenSesame reaction time software (Mathôt et al, 2012), we developed a BETA version of the LLAMA tests, dubbed the ALPACAAs (Applied Linguistic ProgrAmmes for the Computerised Assessment of Aptitude). The ALPACAAs retain the four components of the LLAMA tests: vocabulary (LLAMA_B), sound recognition (LLAMA_D), sound-symbol correspondence (LLAMA_E) and grammatical inferencing (LLAMA_F). We administered these tests to 143 participants including 20 who received a subsequently further revised version of the sound-symbol correspondence (ALPACAA3/ LLAMA E) test. The participants also took a battery of working memory tests. In this paper we will present these ALPACAA and WM results and discuss how they have fed in to the new, revised online LLAMA tests (available through the www.lognostics.co.uk website). Over 700 people have taken the LLAMA B and LLAMA D tests, 389 have taken the new LLAMA E and 279 people the LLAMA F test. We will present initial results of the distributions of these scores to compare with the original LLAMA tests.