A corpus-based investigation
of cognate object constructions
This paper takes up a corpus-based discussion of so-called Cognate Object Constructions (COC) (as exemplified in He slept a deep and dreamless sleep), carried out within the framework of Cognitive Linguistics (CL). This paradigm has recently become criticized for promoting a usage-based approach to language but falling short of wide-ranging applications of empirical methods such as corpus investigations and the thorough analysis of retrieved data using elaborate statistical methods. Our study, in which we introduce COCs as a family of related constructions, presents and discusses the results of an investigation of the pattern as recorded in the British National Corpus (BNC). More than 3,100 instances of COCs could be extracted, which were carefully examined and interpreted by means of a collostructional analysis, a statistical procedure developed by Stefanowitsch and Gries (2003). Different types of the construction will be described, with an attempt to arrange these in a usage-based network of constructions, incorporating the data obtained from our statistical analysis. The network comprises information of different levels of abstraction, ranging from abstract, high-level schemas to low-level, lexically filled patterns (idioms, set phrases).