This chapter examines a series of studies of individual LI, L2 and random vocabulary networks generated by a word association task and analysed according to graph theoretical criteria. It is a development and continuation of Wilks and Meara's preceding chapter in this volume and addresses questions as to the applicability of the metaphor we so commonly use of the lexicon as a web of words. It will consider this question in the light of the findings of three recent large-scale studies in which established corpora were analysed to determine whether semantic networks exhibit properties of a specific subclass of networks – small-world ones.
As noted in the previous chapter, the concept of semantic, vocabulary or lexical ‘networks’ appears quite often in second language acquisition literature as a metaphor which many researchers have used to explain the way words in the mental lexicon hang together (cf. McCarthy, 1990; Aitchison, 1994; Henriksen and Haastrup, 1998; Huckin and Coady, 1999). The metaphor of a web, or network, used by these researchers seems to be intuitively correct. Otherwise it would not be so prevalent, and we would not accept it so easily – to the point that we tend to forget that it is a metaphor, albeit an invaluable one insofar as it provides the impetus to explore the potential of an empirically based model of the structure and function of semantic networks.