International Journal of Corpus Linguistics - Volume 19, Issue 4, 2014

Tests of second language learners’ knowledge of collocation have lacked a principled strategy for item selection, making claims about learners’ knowledge beyond the particular collocations tested difficult to evaluate. Corpus frequency may offer a good basis for item selection, if a reliable relationship can be demonstrated between frequency and learner knowledge. However, such a relationship is difficult to establish satisfactorily, given the small number of items and narrow range of test-takers involved in any individual study. In this study, a meta-analysis is used to determine the correlation between learner knowledge and frequency data across nineteen previously-reported tests. Frequency is shown to correlate moderately with knowledge, but the strength of this correlation varies widely across corpora. Strength of association measures (such as mutual information) do not to correlate with learner knowledge. These findings are discussed in terms of their implications for collocation testing and models of collocation learning.

As repositories of spontaneously realized language, corpora generally have an uncontrolled and unbalanced structure where all variables operate simultaneously. Consequently, a variable’s real effect can be concealed when studied in isolation because of the exclusion of the impact of other potentially confounding variables. Analyzing a variational case study, the alternation between inflected and uninflected attributive adjectives in Dutch, it will be demonstrated how confounding variables alter the impact of explanatory variables on the response variable, resulting in spurious effects in the bivariate analyses. Multiple Correspondence Analysis will be used as a heuristic tool to unveil the association patterns between explanatory variables in the data matrix which induce the spurious effects. Based on these findings, we will argue for a thorough analysis of the database patterns to gain insight in the underlying associations between explanatory variables before modeling their real impact on the response variable in a multivariate model.

Speakers do not always attribute agency straightforwardly when they communicate. While complying with the maxims of explicitness and relevance, they may depict states of affairs headed by an identifiable source. More often than not, however, it seems they leave out this source for a number of reasons and through different mechanisms. This paper is a corpus-based study of one such non-identifying structures, namely the extrapositional have-it-that construction, in examples such as Several hypotheses have it that land-use changes. Drawing on data from the BNC, this paper investigates the use, distribution and functioning of the have-it-that construction. The paper also highlights the usefulness of simple collexeme analysis in revealing systematic co-selection relationships within the construction.

The purpose of this paper is to show how corpus data can contribute to assessing explicit hypotheses about natural language just as experimental protocols can. The particular hypotheses tested concern the source of generalised conversational implicatures with quantifier some. Is the “some and not all” meaning of some a default interpretation of this item or a requirement of certain contexts? The defaultist approach (Levinson 2000, Chierchia 2004) would predict a preponderance of implicatures in the uses of some, whereas the contextualist approach (Sperber & Wilson 1986; Carston 1988, 2002) would predict that the implicature be found only with identifiable contextual triggers. The analysis of attested usage from the Bergen Corpus of London Teenage English (COLT) is shown to invalidate the former and to support the latter hypothesis. The workings of conversational implicatures are argued to be better understandable through corpus investigation than by recourse to decontextualized, self-fabricated, stock examples.

This article presents a corpus study of the variable placement of adverbial satellites in spoken Dutch. It is widely contended that the relative order of satellites is motivated by three general principles: information status, length and the proximity principle. The proximity principle maintains that the placement of satellites is motivated by their semantic relationship with the sentence verb. We investigated the effect of the proximity principle on the relative placement of 8 different satellite classes based on a corpus sample of 202 combinations of two satellites retrieved from the Corpus of Spoken Dutch. The exact binomial test was used to evaluate the statistical significance of the observed orders. Our main results corroborate the hypothesis that the proximity principle influences satellite ordering. We also found, however, that the placement of certain satellite classes appeared very restricted, which suggests that the proximity principle does not play an active role in their placement.