3.2.1 Loss of Word Order Flexibility and Emergence of Non-Canonical Patterns

According to Haeberli (Reference Haeberli2018), loss of OV (or V-final) begins in the OE period and reaches completion during the Middle English (ME) period (see also Rissanen Reference Rissanen and Lass1999). Initially, this strengthens V2 word order in late OE and early ME, particularly in areas with language contact between OE and Old Norse (Haeberli Reference Haeberli2018). Loss of V2 then begins in the fifteenth century and reaches (near) completion by the seventeenth century (Nevalainen Reference Nevalainen1997; Fischer & van der Wurff Reference Fischer, van der Wurff, Hogg and Denison2006: 185). Remnants of OV can be found in Early Modern English (EModE) and remnants of V2 into Late Modern English (LModE), as examples (9) and (10) show.Footnote ⁷

PDE also allows for V2 word order, but such instances are clearly non-canonical. Hawkins (Reference Hawkins1986) provides examples of V2 in negative contexts (as in (11)), following so (in (12)), and with directional adverbs or locatives, illustrated in (13) and (14), respectively (subjects and verbs in italics; emphasis added).

1. 1. a. Never have I been so insulted.

   2. b. Under no circumstances can I permit it.

1. 1. a He can drive a car, and so can I.

   2. b. So be it.

   3. c. So slowly did the workmen get on with their work that they were dismissed.

While the chronology of the major change in word order type is relatively uncontested, the factors that contributed to it and that allow us to explain this fundamental typological change in the history of English are more varied and have been put forward within different theoretical frameworks. We will look at them in the next subsection.

With the flexibility of simply promoting almost any constituent to sentence-initial position in OE having been lost, the language had to develop patterns that would allow for discourse-pragmatic requirements to be met. Already existing constructions that allowed non-prototypical subjects, namely incipient be-passives, fully grammaticalised and increased in frequency (see Seoane Reference Seoane, van Kemenade and Los2006). In addition, English developed non-canonical syntactic patterns that allowed for topicalisation of already ‘given’ referents, which previously would simply occur in sentence-initial position. These include prepositional and get-passives (as in (15)), it-clefts (illustrated in (16)), extraposition (see example (8) above), and – if word order is taken to include the mapping of semantic roles onto syntactic function – patterns with non-canonical subjects, notably middles and secondary subject constructions, illustrated in (17) and (18), respectively.

1. 1. a. This problem has already been taken care of.

   2. b. The train got cancelled.

1. 1. a. It was surely a novel, high-performance ChatBot that insulted him in this genius way.

   2. b. It is with sincere regret that we have to inform you of the cancellation of the scheme.

1. 1. a. She scares easily.

   2. b. The book reads well.

   3. c. The straps on the backpack adjust for a comfortable fit.

As Hawkins (Reference Hawkins1986, Reference Hawkins, Nevalainen and Traugott2012) has pointed out, a concomitant change to a rigid SVO word order in English is an increase in the semantic diversity of the S‑slot. In other words, verbs have fewer selection restrictions and broader subcategorisation frames; in addition, English became more flexible with respect to raising and movement constructions. Thus, in the change from a loose-fit (discourse-oriented) to a tight-fit (syntax-oriented) language, English subjects became more semantically diverse: The middles in (17), for instance, have experiencer and patient subjects with no marking for voice on the verb. The flexibility of English middles goes beyond these two semantic roles and even allows for locative subjects as in (19).

1. 1. a. The new trails in the park cycle smoothly, even for beginners.

   2. b. The moonlight pond fishes well in the early morning.

These examples illustrate well how ‘a language that changes its lexical verb position, from SOV to SVO (like English), can see earlier requirements for predicate frame differentiation and argument differentiation weakened or overridden’, as Hawkins (Reference Hawkins, Nevalainen and Traugott2012: 629) points out.

3.2.2 Theoretical Approaches to Word Order Changes

A relatively common explanation for the loss of the more flexible word order and a shift towards SVO makes a connection between loss of case marking and word order, claiming that in the absence of morphological marking for syntactic function, the development of fixed syntactic slots was a way to avoid ambiguities and ensure communication; in other words, the typological shift was a means to compensate for loss of morphological links (e.g., Vennemann Reference Vennemann and Li1975). Since this account makes a connection between the requirements of processing and fixed order, change is seen as occurring in language use, which is a functionalist explanation. The complementary theoretical perspective conceives of word order changes to occur during acquisition and as a result of shifting frequencies in the input. Change, in this view, is seen as occurring in I-language, with parameters of Universal Grammar (UG) being set differently from one generation to the next. This traditionally falls under the umbrella of a ‘formal’ (or generative) approach to change (even though much of the original formalism in generative grammar has been given up).

Fischer and van der Wurff are critical of the view that communication and successful processing rely on morphological marking or fixed word order, pointing to the role of context:

Levshina (Reference Levshina2022) does not deny the possibility of a connection between word order and morphological complexity. However, the existing research (e.g., Koplenig et al. Reference Koplenig, Meyer, Wolfer and Müller-Spitzer2017) that she cites indicates that the causal effect is from fixed word order to loss of inflections rather than in the other direction, or in her words, ‘rigid word order helps to lose distinct forms’ (Levshina Reference Levshina2022: 143). This is, by the way, a point that had already been made by Jespersen at the end of the nineteenth century (Jespersen Reference Jespersen1894: §75): ‘a fixed word order was the prius, or cause, and grammatical simplification, the posterus, or effect’.

Some generative linguists conceive of V2 as a result of movement (from OV to V2 in main clauses: V-to-C movement); with loss of verbal inflection the verb no longer moves to V2 (see, e.g., Fischer et al. Reference Fischer, van Kemenade, Koopman and van der Wurff2000: 135–6). The problem with this account is that loss of inflection in the verb does not necessarily lead to a loss in the flexibility of word order, as the case of Dutch illustrates (Fischer et al. Reference Fischer, van Kemenade, Koopman and van der Wurff2000). In another generative account, OE pronominal subjects are analysed as clitics. According to this approach, de-cliticisation of the pronominal subjects increased SVO in the input and thus brought about the change in the basic word order pattern that children acquired (Lightfoot Reference Lightfoot1999; Taylor & Pintzuk Reference Taylor, Pintzuk, Nevalainen and Traugott2012). In other words, this account assumes that competition was between two different types of grammar (OV and VO). This ‘double base hypothesis’ (Hinterhölzl Reference Hinterhölzl, Los and de Haan2017) can find an answer within the generative framework via change in the input to first language acquisition. There is a problem for the competing grammars account, however, in that variable use often continues for much longer than catastrophic change scenarios allow for: a case in point would be the fact that OV after auxiliary continues to be found into EModE and V2 is occasionally attested beyond the 1700s. Biberauer and Roberts have countered the criticism that has been voiced against catastrophic change scenarios:

Biberauer and Roberts (Reference Biberauer, Roberts, Ledgeway and Roberts2017: 139) also acknowledge that variation between conservative and innovative patterns can also be observed in texts produced by the same individual, who would then, theoretically, have to have access to competing grammars. They therefore settle on an emergentist approach to parameter setting (see below).

A more sociolinguistic angle is provided by linguists who see language contact with Old Norse and (imperfect) large-scale second language acquisition (or disrupted transmission) as the driving factors that not only led to simplification and the loss of inflections but subsequently also to a more fixed word order (see Weerman Reference Weerman1993). McWhorter (Reference McWhorter2002: 266) summarises this factor in the following way: ‘The English timeline was decisively influenced by what Trudgill (2001) has termed, in apt and savory fashion, “the lousy language-learning abilities of the human adult”’. That sociolinguistic factors may affect syntactic change, particularly with respect to the use of non-canonical patterns, is discussed in more detail by Lange (Chapter 5 in this volume).

Sometimes different theoretical accounts overlap in the cause they see behind the shift in word order type. In his largely functional account, Rissanen (Reference Rissanen and Lass1999: 265) makes a connection between the (late) grammaticalisation of the article and the development of a more fixed word order (again, to compensate for loss of morphological marking). Hinterhölzl (Reference Hinterhölzl, Los and de Haan2017) also sees a connection with the development of the article at the end of the OE period and the loss of word order flexibility; from within a generative framework, he provides a more detailed explanation claiming that

In this analysis (Hinterhölzl Reference Hinterhölzl, Los and de Haan2017: 26), OE has a VO-base and variable spell-out influenced by prosody and pragmatics, with the determiner influencing the weight of the noun phrase: ‘Definite DPs will count as prosodically heavy as soon as the definite determiner is fully grammaticalized and reanalyzed as the head of the DP’. Bastiani (Reference Bastiani, Los, Cowie, Honeybone and Trousdale2022) shows that these forces are also at work in early ME.

As pointed out above, generative accounts traditionally view syntactic change occurring from one generation to the next as part of a change in the input and subsequent changes to the grammar that the next generation of speakers acquires. Usage-based accounts such as grammaticalisation theory and usage-based construction grammar conceive of change as emerging incrementally via small changes in patterns and pattern frequencies. More recent generative accounts call for an emergentist view of syntactic change, too:

In other words, theoretical accounts of language change (including word order change) appear to be converging. This also applies to methodological approaches in the field (see Section 3.2.2).

Construction grammar conceives of constructions as ‘conventional, learned form–function pairings at varying levels of complexity and abstraction’ (Goldberg Reference Goldberg, Hoffmann and Trousdale2013: 17). An important difference between this approach to language and generative accounts is that form and meaning are not separate components or modules but intricately linked in the speaker’s grammar or ‘construct-i-con’. The representation of English possessive constructions in Figure 3.1 provides an illustration of the different levels of abstraction or schematicity (first introduced by Trousdale Reference Trousdale2008: 107) that construction grammarians have come to assume.

Figure 3.1 A construction grammar view of levels of abstraction

(after Hundt Reference Hundt, López-Couso, Méndez-Naya, Núñez-Pertejo and Palacios-Martínez2016: 237; reprinted by permission from Brill)

Figure 3.1Long description

There are four sections in the model: from macro-level, meso-level, micro-level, to construct. Each level is explained on the right in a box. The levels and the explanations are as follows:

• Macro-level: Predicative construction: subject-copula-subject-predicate

• Meso-level: P C 1 to P C 2 to P C 3

• Micro-level: X Copula, the or a, N P, professor of subject, at place

• Construction: She is emeritus professor of linguistics at M I T.

Interestingly, these can now also be found in emergentist approaches to parameter setting from within the generative paradigm. Biberauer and Roberts (Reference Biberauer, Roberts, Ledgeway and Roberts2017: 149) distinguish

• macroparameters (all heads of a relevant type);

• mesoparameters (all heads of a natural class, e.g., V);

• microparameters (small, lexically definable subclass of functional heads, e.g., modal auxiliaries, subject clitics); and

• nanoparameters (one or more individual lexical items).

An important difference with the constructionist approach lies in the fact that the different levels of schematicity or abstraction are not limited to the phrasal level but may be found in other areas of the construct-i-con, for instance the transitive schema at the most abstract level, which allows for less abstract (mono- and ditransitive) constructions at a meso-level, benefactive constructions at a micro level, and lexically filled constructs at the lowest level of abstraction.

Kuningas and Leino (Reference Kuningas, Leino, Suominen, Arppe, Airola and Heinämäki2006) claim that construction grammar is particularly suited to the modelling of word order because it makes the link between form and meaning, including discourse-pragmatic requirements. Despite this, research providing more of a bird’s-eye perspective by aiming to account for the typological shift in English from a loose- to a more tight-fit language was a blind spot in diachronic construction grammar for a very long time. This most likely had to do with the fact that construction grammar was initially somewhat biased towards more idiomatic constructions. An early example of diachronic research into English constructions looks at the emergence of the way-construction (Israel Reference Israel and Goldberg1996) and how the range of verbs that enter the construction develops across time. Hilpert (Reference Hilpert2013), one of the first book-length approaches to English diachronic syntax from a constructionist perspective, has a chapter on ‘information packaging constructions’. However, this provides the basics of what information structure is and lists a number of constructions that serve the reordering of arguments (cleft constructions, dislocation, extraposition) rather than giving a formal account of how they could be modelled in a constructionist account.

Word order changes started to be modelled within constructionist approaches below the constituent level. Cappelle (Reference Cappelle2006) and Perek (Reference Perek2012) assume that at a relatively high level of abstraction, the nodes in the construct-i-con may be underspecified constructemes which, at a lower level of abstraction, would allow for allostruction and thus word order variation, for example in particle placement in particle verbs (she took her hat off vs. she took off her hat). Zehentner (Reference Zehentner2019) applies this notion in her diachronic study of the dative alternation, where a ditransitive constructeme would allow for two allostructions: the direct object construction, as in He gave her the book, and the prepositional object construction, as in He gave the book to the first girl he met that day.

Another factor for the relative paucity of word order as object of study in construction grammar approaches may have been that early constructionist research focused on vertical relations, and Figure 3.1 is an example of this emphasis. The notion of allostructions already shifted this emphasis towards horizontal links, as alternating constructions were conceived of as being linked on this level. This was then brought to bear on diachronic construction grammar more systematically (Sommerer & Smirnova Reference Sommerer and Smirnova2020). Horizontal or ‘lateral’ relations take centre stage in the modelling of OV to VO change that Bloom (Reference Bloom2021) provides. She makes a connection between subject relative clauses and main clauses in her study; the notion of syntactic slots that can be variably filled is another concept characteristic of a constructionist approach:

Similarity relations (or ‘analogy’) thus are seen as an important driving force in a major typological shift in English.

The development of the novel syntactic patterns that emerged to compensate for the loss of flexibility in positioning elements in sentence-initial position (see Section 3.2.1) have also been studied from within different theoretical approaches, notably generative grammar (e.g., Los & Dreschler Reference Los, Dreschler, Nevalainen and Traugott2012; Los & Komen Reference Los, Komen, Nevalainen and Traugott2012; Los & van Kemenade Reference Kemenade, Nevalainen and Traugott2012; Dreschler Reference Dreschler2015, Reference Dreschler2020), grammaticalisation theory (e.g., Sun & Traugott Reference Sun, Traugott, Narrog and Heine2011), and diachronic construction grammar (e.g., Hundt Reference Hundt2007).

3.2.3 Methodological Approaches to Word Order Change

Research on word order change in English used to be largely qualitative. With the advent of computer corpora, more quantitative approaches became possible, which were initially pursued by linguists working within the functional paradigm (e.g., Nevalainen Reference Nevalainen1997). Usage-based construction grammar, especially research within the frequentist strand, has regularly relied on quantitative corpus data and statistical modelling. Again, Bloom (Reference Bloom2021) is a good example of this approach being applied to word order change. More recently, generative studies have also increasingly used quantitative methods (e.g., Komen et al. Reference Komen, Hebing, van Kemenade, Los, Bech and Eide2014), including multivariate statistical modelling, to substantiate their claims. A good example is the study by Struik and van Kemenade (Reference Struik and van Kemenade2020), who use data from The York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE) and regression modelling to validate the claim that changes in word order and aspects of information structure are connected:Footnote ⁸

In a similar vein, García García (Reference García García2023) is able to show that a connection exists between the development away from a relatively flexible, discourse-oriented syntax and the increase of labile verbs (i.e., those allowing alternation between transitive and intransitive construal).