Turns out is not ellipsis? A usage-based construction grammar view on reduced constructions

ABSTRACT In many theoretical approaches, structures such as Turns out , I was wrong are seen as originally being the result of ellipsis, i.e., the deletion of underlying syntactic material (cf. It turns out , I was wrong.). In contrast to this, Usage-based Construction Grammar advocates a surface-oriented view of syntax and, consequently, eschews the postulation of unexpressed, covert syntactic information. In this paper, we will provide a usage-based explanation as to how such reduced constructions can arise in the first place, namely as online constructs in the working memory. Once reduced structures appear in the input, most approaches concede that they can become conventionalized. In the second part of the paper, we test whether there is any empirical evidence for the conventionalization of the constructions at hand. For this, we draw on one of the largest corpora of spoken English (the UCLA NewsScape Library of International Television News corpus), which yields more than 28,000 relevant tokens. Analysing these data for their distributional frequency, syntactic environment as well as their emotive content (using automatic sentiment analysis), we will show that there is synchronic evidence to suggest that the two structures are two individual, yet taxonomically related constructions.