The focus in this chapter is on the possibilities that corpus linguistics offers for the study of unorthodox linguistic expressions which are not easily retrievable through conventional tagging and parsing. A corpus-driven analysis of ‘fragments’, that is, of independent non-clausal constructions with semantic, communicative functionality comparable to that of complete clauses or sentences, is carried out, based on data from the one-million-word British component of the International Corpus of English. The quantitative and qualitative analyses of the data lead to a precise definition of ‘fragment’, couched within (Cognitive) Construction Grammar, and to an inventory of the cognitive mechanisms operating in fragments which allow them to work as alternatives in the competition between fragments and complete clauses in language. The approximately 1,000 fragment tokens in the database are classified into five cognitive types, namely matching, functional, constructional, situational and register support strategies, which evince, respectively, compensation via Information Structure, grammar, Construction-Grammar knowledge, pragmatic competence and text-type entrenchment. As expected, only the so-called situational fragments show statistically significant correlation tendencies with specific text types or registers.