Introduction

D_Tools is a program that computes a statistic reflecting the lexical richness of a text. It is based on work by Malvern and Richards (e.g. Malvern et al., 2004). The statistic that the program computes (called D) is essentially the same measure that is computed by Malvern and Richards' vocd program, but whereas vocd requires that data are formatted according to the conventions used by the CHILDES system (MacWhinney, 2000), D_Tools allows the user to work directly with raw text. Given that the conventions used by CHILDES require some considerable learning effort, the version presented here has some practical advantages for ordinary researchers.

What is D?

Malvern and Richards' vocd-D is a complex statistic. It is basically a type/token measure, and as such belongs to a large family of statistics which use the ratio of word types to word tokens (TTR) in a text as a way of analysing the text's lexical richness. These measures have been widely used in lexico-statistics, particularly in the area of stylistics and authorship studies (e.g. Smith & Kelly, 2002) and in the area of clinical and developmental linguistics (e.g. Fairbanks, 1944; Miller, 1981). By the 1980s, as vocabulary was becoming a focus for research in L2, they were beginning to be used in studies of second language acquisition (SLA). The most notable of these studies was perhaps the ESF project on language acquisition by adult immigrants in Europe, particularly Broeder et al. (1988), which meticulously logged changes in the outputs of naturalistic L2 learners, and used a number of TTR-based indices as a way of measuring these changes.

Some researchers had noted that TTRs and the statistics that are based on them are not as straightforward as they appear at first glance (e.g. Richards, 1987). Specifically, all other things being equal, TTRs will tend to fall as the number of words in a text increases. Obviously, the longer a text is, the more likely it is to contain a word token which has already occurred in the text. This happens for two reasons.