The purpose of this paper is two-fold. Firstly, it presents an argument for usage-based inheritance models over complete inheritance models in construction grammar. It is argued that, with the principle of inductive language learning as their foundation, usage-based inheritance models allow for redundancies and incongruities in construction networks which enables linguists to take into account details of language use, which would otherwise not be facilitated in complete inheritance models. Secondly, making use of the method of collostructional analysis, the article offers a corpus-based description of the use of the Danish ihjel-construction which accounts for patterns of attraction of construction-verb attraction, patterns of productivity, and various types of subconstructions, including item- and item-class-based ones and metaphorical extensions. The description of the ihjel-construction should also illustrate the superiority of usage-based inheritance models over complete inheritance models in construction grammar-based language description.