The chapter will first provide an overview of the developmental stages of the LLAMA aptitude test battery since its inception in 2005; then, more importantly, the chapter moves on to discuss the potential limitations of the current format in terms of internal and external validity. Towards the end, it will outline some new features and changes that have recently been brought to this increasingly popular language aptitude test battery. Data from the new online LLAMA tests will be reported (from LLAMA to ALPACAA and back again). The new LLAMA tests are now available online on Meara’s lognostics website: http://www.lognostics.co.uk/tools/LLAMA_3/index.htm.

Introduction

The work described in this chapter grows out of a concern to explore the power and appropriacy of one of the most persistent and pervasive of the metaphors used in vocabulary research, that of the lexical network. Our attempts to investigate the network metaphor have led us to apply the principles of graph theory to word association data in order to compare the relative density and structure of L1 and L2 vocabularies. This chapter will briefly describe the background to this research and then report the findings of a new empirical study based on a graph theoretical approach. Our discussion and analysis of this work will focus in particular on one central issue raised by Nation in his overview of vocabulary models and measures in the opening chapter: the question of learner attitude in any form of vocabulary testing.

Any field of study that deals with things that we cannot directly observe is obliged to rely heavily on metaphors to describe and probe those phenomena. Research into the mental lexicon is no exception. Yet very often the ‘metaphors we live by’ in vocabulary research (to borrow Lakoff and Johnson's 1980 formulation) are so much part of the field's furniture that we no longer see them for what they are. We forget that they, and all the implications they bring with them, are only approximations to the phenomena we are seeking to capture and do not in fact translate any empirically verifiable reality.