Discontinuous forms are non-canonical options where the choice of the form cannot immediately be related to discourse functions. Rather, the choice between a continuous and a discontinuous variant has been linked to cognitive complexity – speakers will use that form that is easier to process for the hearer. Usually, the non-canonical (i.e., discontinuous) variant is associated with a higher degree of cognitive complexity, which raises the question of why these variants exist in the first place, especially since discourse functions are not as obvious as with other non-canonical word orders. Putting to the test processing-based explanations in two experiments, this chapter investigates the choice between a continuous and a discontinuous particle verb. While the data from a rating task are aligned with previous findings from corpus studies, the data from a reading experiment are not, showing that the continuous variant does not always facilitate reading. This suggests that the non-canonical (i.e., the discontinuous) form might serve functions apart from information-structural ones.

This chapter provides an overview of ways to study, categorise, and analyse non-canonical syntactic patterns in registers of English. It introduces two distinct approaches to studying the role of discourse and register in determining syntactic variation. The first (‘variationist’) approach looks at non-canonical syntax as a case of grammatical variation with register as the predictor. The second (‘text-linguistic’) approach takes register as its proper object of investigation and looks at non-canonical constructions as frequent and pervasive features of a register. We classify non-canonical syntactic constructions according to their form as either reduced, expanded, or re-ordered versions of canonical clauses. Each of these patterns is exemplified in one of the studies that constitute the section of the volume introduced by this chapter (ellipsis as reduced constructions, clefts as expanded constructions, and particle placement as reordering). Comparing these studies, this chapter also elaborates on the role of corpus methods as well as experimental data in shaping research questions regarding the motivation for non-canonical patterns. A final part discusses trends and open questions, such as problems of register classification for text from media and new challenges presented to the field by generative AI tools.