The development of small clauses in English: constructional growth and decline

This study investigates the development of the English Secondary Predicate Construction, against the background of current views on language change. A Secondary Predicate Construction, or SPC, consists of a verb, a noun phrase (NP), and an Xphrase (XP). The NP and XP are in a predicative relation, similar to a subject-predicate relation in a copular clause. In (1a) for instance, consider him handsome involves a verb (consider), an NP (him), and an XP (the adjectival phrase handsome), with handsome predicating a property of him. Similarly, makes it difficult in (1b) contains a Verb-NP-XP configuration and conveys a subject-predicate relation between it and difficult.

  a.           He’s not exactly Douglas Fairbanks, but some might consider him handsome.

(BNC, 1991)

            b.         The massive destructiveness of nuclear weapons makes it very difficult to

justify them as ‘effective’ weapons in this sense. (BNC, 1986)

While the predicative relation need not be signalled by a dedicated function word, as in (1) above, it can also be marked explicitly by an introductory as (2a), to be (2b), or for (2c). Thus, the English SPC comprises four ‘SPC subtypes’, which are referred to as zero-SPCs (with ‘zero’ markers), as-SPCs, to be-SPCs, and for-SPCs respectively.

  a.           They see him as a possibility for the future. (BNC, 1992)

b.           Most of us have rather too much faith in the high tech ability of these smart

cards, and imagine them to be capable of more than they are. (BNC, 1992)

c.           Roirbak had mistaken her for another boy.  (BNC, 1991)

The four SPC subtypes are functionally related constructions that are characterised by partly distinct and partly overlapping distributions. Moreover, the amount of distributional overlap they exhibit has been subject to change over time. This study, then, presents a diachronic, corpus-based account of the SPCs from Old English (OE) to Late Modern English (LModE), documenting their history and establishing the causes of shifts in their different distributions. By contrasting and linking the respective developments in zero-SPCs, as-SPCs, to be-SPCs and for-SPCs, light is also shed on the mechanisms underlying changes within a broader functional domain. This overarching goal splits into three more specific objectives, each feeding into the other.

Firstly, I chart the developments of the individual SPC subtypes and establish their underlying mechanisms. For each SPC, I identify the licensing verbs with which it can occur and map out the changes it shows with respect to this set of licensing verbs (in terms of token frequency and productivity) in different stages of its development. Verb-SPC combinations have not remained constant over time: some have become obsolete (e.g., the zero-SPC with mistake in (3a) or the for-SPC with reckon in (3b)). Conversely, some patterns have been newly introduced, such as as-SPCs with regard in (4).

  a.           Let them (...) mistake me wanton. (1589, OED)

b.           I reckon for nothing the researches of a Coyer. (1764, OED)

                 Many pop stars still regarded a racehorse as the ultimate status symbol.

(2005, OED)

For each of the SPC subtypes, I further describe changes in the verbs’ combinatorial potential with the XP slot (which is characterised by various formal realizations), in the sequence of the formal components Verb-NP-XP, and developments in the semantics of the SPCs. Having established the development and distribution of the different SPC subtypes, I also explore the various mechanisms accounting for shifting distributions. These mechanisms may be restricted to one SPC subtype, but they may also stem from interactions with other SPCs. That is, growth in one SPC subtype may result from construction-internal developments affecting only that particular SPC, but may also be due to the influence of other functionally related SPCs. An example of the former is for instance the case in the zero-SPC, where make has developed a mental meaning (e.g., Whom makest tow þe? ‘Whom do you make yourself?’) through analogy with mental hold (e.g., He held it impossible that …). However, analogy can also operate across SPC subtypes: in Chapter 7, it is shown that the as-SPC has copied parts of the distribution of the zero-SPC on the basis of analogy, adopting a number of cognition verbs such as understand, conceive, approve, or know, which were already well-established with the zero-SPC. Given the functional relatedness of the various SPC subtypes, special attention is given to these construction-external mechanisms, as they are a useful diagnostic for tracing possible interactions between the various SPC subtypes.

            The first research goal feeds into my second aim, which is to show the significance of the results of the descriptive analyses for the broader theoretical discussion on mechanisms of language change. In particular, I challenge the time-honoured view (Bolinger, 1977; Goldberg, 1995; Nuyts & Byloo, 2015) that functionally related constructions entering into competition are bound to result either into one construction replacing the other (‘substitution’), or into both constructions developing their own functional niche (‘functional differentiation’). While such an idea has successfully accounted for some developments in previous studies, I argue that it fails to capture the intricate evolution of the SPC. Instead, I draw on the hypothesis developed in De Smet et al. (submitted) that argues for a scenario of ‘attraction’ between two or more functionally similar constructions. In such an attraction process, the constructions by default develop more, rather than less, overlap, because the language user recognises their commonalities and consequently transfers features from one construction to the other more easily. The remaining challenge is then to account for persisting differences between the constructions, as these cannot be put down to substitution or differentiation. It is shown that such remaining differences can often be explained on the basis of the constructions’ individual histories and relations with their source constructions, which may exert a pulling, or ‘anchoring’ force on the constructions’ later distributions. Since the SPC comprises not only two, but four related constructions, it constitutes an ideal testing ground for these hypotheses and serves to expand our knowledge of language change.

            Finally, this study is couched in the cognitive-functional framework of Construction Grammar. A third aim therefore is to reflect on some of the benefits and drawbacks of Construction Grammar for a study on the SPC. With its simultaneous focus on form and function and its (more recent) emphasis on Diachronic Construction Grammar (Traugott, 2008c; Fried, 2009; Trousdale, 2012; Traugott & Trousdale, 2013; Barðdal & Gildea, 2015), Construction Grammar offers an ideal background to unravel the intricate developments of the four formally and functionally related SPCs. Nonetheless, as this framework was mainly tailored to fairly specific constructions in the past, such as, among others, the let alone-construction (Fillmore et al., 1988), the way-construction (Israel, 1996), or the size noun-construction (Brems, 2010), it poses some challenges to apply it to highly abstract constructions like the SPC, which consists of a variety of formally and functionally related subtypes. In particular, two major aspects of Construction Grammar pose a challenge. The first relates to the question of the level of abstraction that can or should be posited for a construction. According to Construction Grammarians, constructions can be stored at different levels of schematicity (or, abstraction) and they can be stored redundantly. It follows that the linguist needs to select a level of schematicity for investigation, but how this should be done, or which underlying criteria should guide this decision, is not mentioned in the literature. I argue that this issue was not even thought of as problematic until recently, but may have serious repercussions for how the researcher approaches the construction. A second issue concerns the fact that relations within constructions, or ‘vertical/intra-constructional relations’, are often taken to have priority over relations between constructions, or ‘horizontal/inter-constructional relations’. Consequently, possible construction-external sources of influence on a construction’s development may be overlooked, resulting in an incomplete picture of the diachrony of a construction. The findings of the present study contribute to these discussions, and serve to evaluate Construction Grammar as a framework for studying abstract patterns.

This study is structured as follows. In a first part (Chapters 1 to 5), I set out the necessary theoretical and methodological background for the study. The second part (Chapters 6 to 9) details the results of the corpus analyses, which amounts to a description of the historical development of the different types of SPCs and the parallels and differences they exhibit. The third part (Chapter 10 and conclusions) advances further theoretical reflections and sums up the conclusions.

The main conclusion of the study is that the notion of an SPC macro-construction subsuming four SPC subtypes is only correct from a synchronic viewpoint, and is in fact the result of a gradual process of attraction. Originally, the SPC subtypes had little in common and were probably stored separately in the mind of the language user. As the subtypes developed more functional overlap over time, an overarching level of schematicity started to emerge, giving rise to the SPC as we know it today. This conclusion represents an important contribution to the issue of ‘abstraction’ in Construction Grammar, as it shows that the right level of abstraction can sometimes be revealed only from a diachronic perspective. This insight could only be attained by devoting specific attention to the analysis of horizontal relations existing between constructions, and thereby further underscores the importance of considering language change across constructional networks.