Evidence from corpus and experimental data
12:00 - 14:00
Lina Baldus, University of Trier
Lectures, talks and seminars
Language and Linguistics, Department of
Victoria Mead firstname.lastname@example.org
Constructionist and usage-based approaches to language claim that the frequency with which linguistic input is experienced is a key factor in the process of second language acquisition (Bybee 2008, Ellis 2002; Ellis et al. 2016; Goldberg 2006; Madlener 2015).
One aspect that is often left underspecified, however, is which component parts of linguistic input need to be experienced with sufficient frequency by learners in order to form a native-like schema representation for a certain construction.
The present study addresses this question, using the English catenative verb construction as a testbed phenomenon. The catenative verb construction consists of the ‘catenative verb’ (see bold verbs in (1) and (2) below) and a non-finite complement, the so-called ‘catenative complement’ (Huddleston & Pullum 2002) which can be, for instance, a gerund-participial or a to-infinitival complement as in the following examples:
The construction is of particular interest to research in the study of English as a Second Language because learners often show choices of the complement type different from those of native speakers (Deshors 2015; Gries & Wulff 2009; Martinez‐Garcia & Wulff 2012).
This study explores how different frequency-based variables affect this variation, with the goal of finding out what component parts of the construction need to be experienced with sufficient frequency for the learner to build a native-like schema for this construction.
In order to address these issues, a pseudo-longitudinal corpus study with language data from German learners of different proficiency levels (A1-C2) was carried out using the EF-Cambridge Open Language Database (EFCAMDAT, Geertzen et al. 2013). In addition, two experimental studies, a sentence completion task and an acceptability judgment task with advanced German learners of English (C1 level), were conducted. In all cases, the focus was on a selection of verbs which are distinct for one of the two types of catenative complements mentioned above and which occur with different frequencies in the catenative verb construction, both determined on the basis of the British National Corpus (Davies 2004-).
Each study was analysed with multivariate statistics (cf. Baayen et al. 2008; Levshina 2015) with the frequency of the matrix verb (i.e. the catenative verb in all of its different uses) and the frequency of the verb together with its distinct catenative complement as predictor variables, in order to see to what extent the different factors had an impact on the dependent variable, namely the type of catenative complement preferred by the learner and its target-likeness.
While in the experimental studies with the advanced learners of English a considerable number of non-target-like choices (e.g. *…avoided to use…) could be observed, the majority of catenative verb constructions produced by learners of different proficiency levels in the corpus data showed the opposite, namely a very high number of target-like choices. In all three studies, the frequency with which the catenative verb occurs with the respective complement type made a strong and significant prediction of the target-like choice of the complement type. However, the frequency of the matrix verb in general had an impact on the target-like choice of the complement type only in the corpus study but not in the experimental data.
These findings provide an important insight into how frequency affects the mental representation of the catenative verb construction: it is essential for L2 learners to have experienced the catenative verb together with its target-like complement type rather than being familiar with the matrix verb alone. The catenative verb together with its complement type forms a processing unit (i.e. a construction) which needs to be experienced with sufficient frequency in order for the learner to build a native-like schema representation of the catenative verb construction.