4.1 Introduction
Full-verb inversion – also often referred to as locative inversion – is one of the non-canonical syntactic constructions described by Birner and Ward (Reference Birner and Ward1998) in their overview of non-canonical word order in Present-Day English (PDE). It is characterised syntactically by a reversed order of the subject and the full verb phrase, and in the majority of cases occurs with to be, unaccusative, or passivised transitive verbs, following a clause-initial prepositional phrase (PP). Birner and Ward (Reference Birner and Ward1998) describe it as having an information-structuring function: the initial phrase (PP or otherwise) needs to be as discourse-old as or more discourse-old than the inverted subject. Example (1) illustrates this inversion pattern in PDE, with the preposed PP showing a clear link to the previous main clause (a great big tank – in the tank), and the subject introducing new information to the discourse (all these pots).Footnote 1
They have a great big tank in the kitchen, and [in the tank] are sitting all these pots.
The deviation from the rigid subject-verb-object order in PDE clearly makes full-verb inversion non-canonical from a theoretical perspective. From a frequency-based perspective, too, it is non-canonical: it is rather rare; it is also restricted in use, being limited mostly to main clauses (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan2021: 917) and specific contexts – although these contexts are less well-defined than they are for some other types of PDE inversion, such as subject-auxiliary inversion in yes/no-questions or after certain adverbs, such as (not) only (Huddleston & Pullum Reference Huddleston and Pullum2002: 94–7). Indeed, in contrast to subject-auxiliary inversion, full-verb inversion is usually optional in that it is not grammatically required, although, as in (1), an uninverted clause may be ineffective in information-structural or stylistic terms.
In earlier stages of English, especially before AD 1500, inversion of subject and finite verb was a frequent phenomenon. Indeed, in certain contexts, inversion was the only option; in other words, it was clearly not non-canonical, neither from a frequency-based nor from a theory-based perspective. In fact, Old English (OE, until 1150) is commonly analysed as a verb-second (V2) language (see also Hundt, Chapter 3 in this volume), where – as in modern V2 languages such as German and Dutch – the finite verb occurs in second position in the main clause, leading to inversion of subject and finite verb when the clause does not begin with the subject, as illustrated in (2).
[On þam dagum] wæs Alexander geboren on Crecum …
in those days was Alexander born in Greece
‘At that time, Alexander was born in Greece …’
The OE system of inversion, however, was already more variable than that of modern V2 languages, and in the Middle English period (ME, 1150–1500), the system was lost altogether, with the final loss of the pattern usually dated around 1500 (e.g., Fischer et al. Reference Fischer, van Kemenade, Koopman and van der Wurff2000; Haeberli Reference Haeberli, Jan-Wouter Zwart and Abraham2002). In addition to the inversion patterns that characterise V2 (i.e., involving only the finite verb), another and less common subject position has been identified for OE and ME, the so-called late subject position (Warner Reference Warner2007), where the subject follows both the finite and non-finite verb, as illustrated in (3).
[þurh ða wifunge] sind getacnode þæs lichaman lustas
through the wife-taking are signified of-the body lusts
‘The body’s lusts are signified by taking to wife’
The descriptions of this type of inversion in the literature (Warner Reference Warner2007; Biberauer & van Kemenade Reference Biberauer and van Kemenade2011; Dreschler Reference Dreschler, Los and de Haan2017) seem similar to PDE full-verb inversion: late subjects occur after the full verb phrase and mostly with unaccusative and passivised transitive verbs; and they are associated with discourse-newness. However, the (non-)canonical status of late subjects is less clear: inversion overall is more common in OE/ME, and yet late subjects seem to be both infrequent and distinct from the V2-type of inversion.
The presence of and similarities between late subjects and PDE full-verb inversion raise the question to what extent these are the same phenomenon in both periods. However, PDE full-verb inversion and OE/ME late subjects have so far been studied as distinct phenomena, and there is as yet no overview of the pattern of full-verb inversion throughout all periods. Crucially, while late subjects are one of the many types of inversion in OE/ME, full-verb inversion is one of the few types of inversion in PDE; and unlike some other types, full-verb inversion was not lost entirely. One explanation for its continued use may be its discourse function: both PDE inversion and late subjects have been described in terms of their information-structural function, suggesting that they allow speakers or writers to achieve a particular information-structural effect, one that was as useful in OE as it still is in PDE. In order to explore this idea, however, we first need to better understand the syntactic development throughout all periods.
In this chapter, I present a large-scale study of inversion patterns in the history of English – based on four syntactically parsed corpora of historical English, spanning the years from 450 until 1910 – in order to gain insight into two questions: (i) How does full-verb inversion (i.e., inversion of subject and full verb phrase) develop syntactically over time? (ii) To what extent is full-verb inversion non-canonical in different stages of the language? The aim is to gain insight into the syntactic development; however, I will also include in the analysis some syntactic factors known to be related to information status.
This chapter is structured as follows. Section 4.2 defines full-verb inversion in OE/ME and PDE and provides an overview of existing studies. Section 4.3 discusses the methodology for the current study. Section 4.4 first presents the distribution of inversion patterns throughout all periods and then zooms in on each stage of the language. Section 4.5 concludes.
4.2 Background
This section reviews previous research on full-verb inversion in PDE and OE/ME, identifying similarities and differences between the two phenomena, as well as gaps in the literature.
4.2.1 Full-verb Inversion in PDE
Full-verb inversion in PDE involves inversion of the subject and the full verb phrase. It is distinct from inversion that only involves the auxiliary, such as in yes/no-questions and wh-questions, and after certain negative or focusing initial constituents including (not) only, so, thus (Huddleston & Pullum Reference Huddleston and Pullum2002: 94–7). It is known under various names, including ‘locative inversion’ (e.g., Bresnan Reference Bresnan1994; Hoekstra & Mulder Reference Hoekstra and René1990; Broekhuis Reference Broekhuis2005), ‘inversion’ (Birner & Ward Reference Birner and Ward1998), ‘subject-dependent inversion’ (Huddleston & Pullum Reference Huddleston and Pullum2002), and ‘full-verb inversion’ or ‘full inversion’ (Chen Reference Chen2003; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan2021). To include all relevant instances in PDE, not just those starting with a locative PP, and to distinguish between different types of inversion in earlier stages of English, I will use the term ‘PDE full-verb inversion’.
PDE full-verb inversion is often referred to as ‘locative inversion’ because it is especially common after PPs indicating a location, although this observation does not account for all examples. In Birner and Ward’s (Reference Birner and Ward1998) corpus study, the majority of the 1,778 examples involve PPs in initial position (72%), followed by verb phrases (16%) and adjective phrases (6%). Bresnan (Reference Bresnan1994) accounts for non-PP examples by proposing that even in those examples, there is some sense or indication of either location or direction, as in (4), where crashing through indicates movement.
However, in the absence of systematic corpus studies, it is unclear whether this explanation holds for all non-PP examples. Indeed, Huddleston and Pullum (Reference Huddleston and Pullum2002) describe the occurrence of a sentence-initial locative PP as one condition for full-verb inversion, but list a separate condition for non-locative initial elements, which is that the initial element should contain a strong discourse-link, as in (5).
They had every kind of cake imaginable, all lined up in a row. [Adorning the first one] was a monstrous rose sculpted from white chocolate.
The initial element Adorning the first one in the second sentence is inferable from the description in the first sentence.
A second syntactic characteristic is that locative inversion overwhelmingly occurs with to be, intransitive verbs (more specifically unaccusative verbs), and passivised transitive verbs. To be-inversion is treated as a distinct category by some authors, with Birner and Ward (Reference Birner and Ward1998) arguing that it is less restricted both semantically and syntactically, while Chen (Reference Chen2003) argues it is the prototypical type and other verbs represent an extension. Birner and Ward (Reference Birner and Ward1998) focus on the information value, rather than the argument structure, of the non-be verbs, proposing that the verb needs to be predictable from the context. Unaccusative and passivised transitive verbs, in turn, are usually grouped together based on their argument structure: as Levin and Rappoport Hovav (Reference Levin and Hovav1995: 215) explain, these verbs lack an underlying subject with the role of agent. Yet examples of other verbs involved in full-verb inversion do exist, as in (6), which has an unergative intransitive verb (i.e., it does have an agent subject).
[On the third floor] worked two young women called Maryanne Thomson and Ava Brent …
However, exceptions seem to be few: Birner and Ward’s (Reference Birner and Ward1998) corpus study contains only two transitive examples.
The final defining characteristic of PDE full-verb inversion is its pragmatic function, which has been described in many different ways. Birner and Ward (Reference Birner and Ward1998) propose that full-verb inversion is governed by a relative information-structural constraint: in order for the sentence to be felicitous, the preposed element needs to be older than the postposed element (78.2% of their 1,778 examples) or equal in information status (10.7% old–old; 11.1% new–new). Birner and Ward (Reference Birner and Ward1998) specifically phrase their constraint in terms of occurrence in the text (discourse-oldness) rather than familiarity based on world knowledge (hearer-oldness). They also briefly address the factor of definiteness and find that the majority of preposed elements are definite, but postposed constituents are equally as often definite as indefinite, which they explain by pointing out that definiteness only indirectly relates to information status. Birner and Ward’s (Reference Birner and Ward1998) relative information constraint is compatible with an earlier proposal by Bolinger (Reference Bolinger1977), who defined the function of inversion as ‘present[ing] something on the immediate stage’ (Reference Bolinger1977: 93–4). Chen (Reference Chen2003), in turn, defines full-verb inversion as a construction that allows for placing the ground before the figure, going against the standard English order of figure-before-ground. Although the details of these accounts differ, they share the basic notion that the reversal of subject and verb achieves a certain pragmatic effect associated with ‘positioning’ the last element with respect to the first.
4.2.2 Inversion Patterns in OE and ME
OE word order is notoriously variable. With respect to the position of the subject, most attention has gone to identifying the regularities and variation within the V2 system (following van Kemenade’s initial analysis, Reference Kemenade1987), a syntactic organisation of the main clause which involves inversion of subject and finite verb in main clauses that start with a phrase (an unspecified XP) that is not the subject. While in modern V2 languages such as German and Dutch, the finite verb is always in second position in the main clause, whatever the first constituent, V2 in OE is less systematic (for an overview, see Fischer et al. Reference Fischer, van Kemenade, Koopman and van der Wurff2000). Two contexts are regular: pronominal subjects are almost never inverted, while certain clause-initial elements always trigger inversion, most notably þa and þonne (both meaning ‘then’). It is the remaining contexts – nominal subjects in declarative main clauses which start with an XP other than þa and þonne (most frequently adverbs, PPs, and objects) – that show variation between V2 with inverted subjects, as in (7), and V3 with uninverted subjects, as in (8).
[Under Moyses æ] moste se bisceop habban an geæwnod wif, …
under Moses’ law should the bishop have an espoused wife
‘Under Moses’ law, a bishop was to have an espoused wife, …’
[æfter þan] þæt lond wearð nemned Natan leaga
after that that land was named Natan lea
‘After him, that land was called Netley’
In (7), the subject se bisceop follows the finite verb moste, whereas in (8) the subject þæt lond precedes the finite verb wearð. Information structure has been invoked as a guiding principle for the variation in OE and ME, following observations that preverbal V3 subjects tend to be given and postverbal V2 subjects tend to be new (e.g., Bech Reference Bech2001; Hinterhölzl & Petrova Reference Hinterhölzl and Petrova2010; Biberauer & van Kemenade Reference Biberauer and van Kemenade2011). The variation between V2 and V3 continues to exist until near the end of the ME period, with the loss of V2 usually dated as taking place in the fifteenth century (e.g., Fischer et al. Reference Fischer, van Kemenade, Koopman and van der Wurff2000; Haeberli Reference Haeberli, Jan-Wouter Zwart and Abraham2002; van Kemenade Reference Kemenade, Nevalainen and Traugott2012), although there is considerable variation between contexts, such as the type of verb and type of initial elements.
In addition to the V2/V3-variation in OE and ME, late subjects have been identified as a third subject position and seem to represent a distinct phenomenon. Warner (Reference Warner2007) is the first to use the term ‘late subject’, although the pattern was also described by others before him (van Kemenade Reference Kemenade, van Kemenade and Vincent1997; Haeberli Reference Haeberli1999, Reference Haeberli, Pintzuk, Tsoulas and Warner2001; Bech Reference Bech2001). The late subjects in Warner’s database of OE and ME examples do not only follow the finite verb, as in V2 inversion, but also a non-finite verb or a verbal complement (i.e., they represent full-verb inversion).
[Than] entered onto þe castell on Jon Butler
then entered into the castle one John Butler
‘Then one John Butler entered into the castle’
Warner (Reference Warner2007) analyses the verb and subject properties of late subject examples and finds that unaccusative and passive verbs with an indefinite subject favour the late subject position, just as subjects longer than four words do, where he takes indefiniteness and length as proxies for discourse-newness. Biberauer and van Kemenade (Reference Biberauer and van Kemenade2011) predict that late subjects are information-structurally new but do not provide corpus data. Testing Biberauer and van Kemenade’s (Reference Biberauer and van Kemenade2011) prediction for OE, Dreschler (Reference Dreschler, Los and de Haan2017) analyses the information status of 121 late subjects in passive main clauses. Her results show that most late subjects are not hearer-new, but are indeed discourse-new. In addition, analysis of a larger database consisting of all main clauses with passive verbs confirms Warner’s (Reference Warner2007) findings regarding definiteness and length: late subjects are generally long, and definite subjects are slightly less likely to occur in the late subject position. Finally, Dreschler (Reference Dreschler, Los and de Haan2017) shows that the initial elements that trigger V2 inversion do not seem to influence late subject placement, which suggests that late subjects are indeed a phenomenon distinct from V2 inversion.
4.2.3 Interim Summary
Both PDE and OE/ME full-verb inversion (i.e., late subjects) are characterised by the subject occurring after the full verb phrase in clauses with unaccusative verbs and passivised transitive verbs. The two phenomena also seem to share an information-structural function: both PDE and OE/ME full-verb inversion are subject to a relative information-ordering constraint in terms of discourse-familiarity, as demonstrated by Birner and Ward (Reference Birner and Ward1998) for PDE, Dreschler (Reference Dreschler, Los and de Haan2017) for OE, and Warner (Reference Warner2007) for ME. There are also some differences. PDE full-verb inversion is closely associated with clause-initial locative PPs, while no such specific association with clause-initial elements has been demonstrated for OE/ME. Most importantly, however, full-verb inversion occurs as one of only few inversion options in PDE, while various other inversion patterns exist in OE and ME, which may also serve a similar information-structural function. Overall, full-verb inversion seems to be more specialised in PDE than in OE/ME, both in terms of the verbs and initial elements and in terms of the clearly defined information-structural function.
While both syntactic and discourse-related factors have been analysed for OE/ME and PDE, no such studies exist for the intermediate periods, and the OE/ME and PDE data are based on different types of corpora, methods, and theoretical analyses. In the remainder of this chapter, I provide an overview of syntactic properties of full-verb inversion in all periods using a consistent methodology, in order to answer the question whether PDE full-verb inversion represents the emergence or continuation of a non-canonical word order. In addition, I will add to the existing research by analysing certain factors (type of verb, definiteness and length of subject) for those periods where no such data are yet available.
4.3 Methodology
This study uses four syntactically parsed corpora of historical English prose texts: the York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE), the Penn-Helsinki Parsed Corpus of Middle English, second edition (PPCME2), the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME), and the Penn Parsed Corpus of Modern British English (PPCMBE). Corpus sizes range from 950,000 to 1.7 million words and the corpora contain prose texts from various genres. While all the corpora are annotated according to the same principles, each corpus also has its unique characteristics, partly reflecting the differences between the stages of the language, especially between OE/ME and later periods. Despite certain methodological issues (see, e.g., Taylor Reference Taylor2020), these corpora have been widely used in historical syntactic research. To search the corpora, I used the programme CorpusStudio, which builds on CorpusSearch 2.
The database for the study consists of declarative main clauses with a subject and a finite verb: inversion is mostly a main clause phenomenon and the subject and finite verb are the key elements of variable order.Footnote 2 I excluded pronominal subjects because they overwhelmingly occur before the verb, both in OE and PDE.Footnote 3 Likewise, I excluded subject-initial and finite-verb-initial clauses, because they do not allow for variation in the order of subject and finite verb.Footnote 4 The remaining database, then, consists of main clauses with a nominal subject, (at least) a finite verb, and some phrase other than the subject or finite verb in clause-initial position.
Types of inversion are not annotated as such in the corpora; instead, subject patterns were selected based on the surface order of the key constituents: initial element (X), subject (S), and verb (V). The clauses in the database were divided into two groups: clauses with only a finite verb, and clauses with both a finite and a non-finite verb.Footnote 5 For clauses with both a finite and non-finite verb, I first singled out the uninverted subjects – XSVV, as in (10). Then, I distinguished between clauses where the subject follows the finite verb but precedes the non-finite verb – XVSV, as in (11) – and clauses where the subject follows both the finite and the non-finite verb – XVVS, as in (12).Footnote 6 The orders are illustrated here with made-up PDE examples for ease of reading.
XSVV [After ten years,] they were murdered.
XVSV [On that day] was Alexander born in Greece.
XVVS [On the wall] were displayed three pictures.
The presence of the non-finite verb acts as a diagnostic for the position of the subject as either occurring immediately postverbally, which corresponds to V2 inversion (11), or following the full verb phrase, which marks it unambiguously as full-verb inversion (12).
For clauses with a single verb, I likewise first selected the uninverted subjects: XSV, as in (13). For the inverted subjects in clauses with a single verb, there are no straightforward diagnostics to distinguish between V2 inversion and full-verb inversion, especially in earlier periods when different types of inversion existed. To distinguish between different surface patterns, I separated out subjects that immediately follow the finite verb – XVS, as in (14) – from subjects that follow the finite verb and some other phrase – XVxS, as in (15).
XSV [After ten years,] they met again.
XVS [Into the room] stepped a woman in a green coat.
XVxS [At that moment] stepped into the room a man in a blue coat.
This selection does not specify the type of material that occurs between the finite verb and the postverbal subject. These phrases may be verbal complements, one of Warner’s (Reference Warner2007) diagnostics for full-verb inversion, but other phrases, such as pronouns or clitics, may not be reliable diagnostics. In other words, XVxS is likely to represent full-verb inversion, while XVS clauses are a ‘mixed bag’: inversion may be due to known V2-triggers or to factors associated with full-verb inversion.
4.4 Results
This section presents the results of the study by first providing an overview of the distribution of subject patterns throughout all periods and then zooming in on the profile of the different stages of historical English.
4.4.1 General Overview
Figure 4.1 and Table 4.1 present the relative frequencies of the five subject positions in all four corpora, using the abbreviations and subperiods as defined in the corpora, with ‘O’ referring to OE, ‘M’ to ME, ‘E’ to Early Modern English, and ‘B’ to Modern British English. The preverbal subjects (XSV and XSVV) have been combined.
Figure 4.1 Distribution of subject positions in the four corpora of historical English
Figure 4.1Long description
The vertical axis marks percentage, ranging from 0 to 100, in increments of 10. The horizontal axis marks the subject positions as O 1, O 2, O 3, O 4, M 1, M 2, M 3, M 4, E 1, E 2, E 3, B 1, B 2, B 3, and B 4. Five different line styles are marked on the graph. The line styles are as follows:
Solid line with dot – X S V V
Solid line with solid square – X V S V
Line with triangle – X V S
Line with empty square – X V x S
Line with star – X V V S
The trends observed in the line styles are as follows:
The X V S line starts at 60% and drops to 39%, which again rises to 51% and then gradually declines to 8%.
The X S V V line starts at 40%, gradually declines to 25%, and then rises to 90% with fluctuations in between.
The X V S V, X V x S, and X V V S lines start from 0% rises and then show a gradual decrease, and again subside to 0%.
Table 4.1Long description
The table displays the distribution of subject positions in four historical English corpora across 14 periods, from 450–800 to 1840–1914. The periods are labeled O 1, O 2, O 3, O 4, M 1, M 2, M 3, M 4, E 1, E 2, E 3, B 1, B 2, and B 3. The periods are 450 to 800, 850 to 950, 950 to 1050, 1050 to 1150, 1150 to 1250, 1250 to 1350, 1350 to 1420, 1420 to 1500, 1500 to 1569, 1570 to 1639, 1640 to 1710, 1700 to 1769, 1770 to 1839, 1840 to 1914. The periods are grouped into historical ranges, illustrating the linguistic evolution over time. The data from left to right is filled as follows:
For X S V V, the corresponding values are 6, 1267, 2478, 2693, 736, 320, 3137, 3131, 2596, 2837, 1876, 1282, 1530, and 1529.
For percentage, the corresponding values are 40, 35, 25, 36, 33, 47, 62, 63, 67, 81, 87, 93, 88, and 90.
For X V S V, the corresponding values are 0, 449, 1034, 794, 309, 41, 410, 296, 280, 165, 44, 30, 41, and 37.
For percentage, the corresponding values are 0, 13, 10, 11, 14, 6, 8, 6, 7, 5, 2, 2, 2, and 2.
For X V S, the corresponding values are 9, 1369, 5220, 3274, 990, 225, 1289, 1272, 710, 379, 203, 55, 152, and 121.
For percentage, the corresponding values are 60, 38, 52, 43, 44, 38, 26, 26, 18, 11, 9, 4, 9, and 7.
For X V x S, the corresponding values are 0, 357, 845, 573, 163, 41, 82, 136, 99, 36, 24, 5, 6, and 4.
For percentage, the corresponding values are 0, 10, 8, 8, 7, 6, 2, 3, 3, 1, 1, 0, 0, and 0.
For X V V S, the corresponding values are 0, 135, 383, 222, 6, 21, 127, 144, 210, 104, 19, 8, 15, and 14.
For percentage, the corresponding values are 0, 4, 4, 3, 3, 3, 3, 3, 5, 3, 1, 1, 1, and 1.
For the total, the corresponding values are 15, 3577, 9960, 7556, 2262, 678, 5045, 4979, 3895, 3521, 2166, 1380, 1744, and 1705.
For percentage, the corresponding values are 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, and 100.
Figure 4.1 clearly shows the most obvious change in the distribution of these subject positions: preverbal (i.e., uninverted) subjects – XSV(V) – become the dominant pattern over time. While in the earlier periods, preverbal subjects make up between roughly 30% and 40% of the X-initial clauses, there is a first major shift between M1 and M3, when the percentage rises to 62%. The percentage continues to climb steadily, but at a slower pace, and is relatively stable in the Modern British English (MBE) corpus, with percentages of roughly 90% throughout. Not surprisingly, the inversion pattern which is initially most frequent – XVS – shows the opposite development, starting out at 60% and decreasing more or less at a constant rate throughout the periods, except for dips in O2 and B1, eventually reaching 7% in the final subperiod of MBE.
The remaining three inversion patterns are much less frequent overall, and as a result changes in their frequencies take place on a smaller scale. Note that in the literature on PDE, all full-verb inversion types are generally grouped together, but distinguishing between them in earlier periods provides a better insight into the historical development.Footnote 7 The XVSV pattern – V2 inversion – is relatively frequent in OE and ME, except in O1 (which has extremely low numbers compared to other subperiods). Until M2, the percentages range between 11% and 14%. In the second half of the Middle English period, they dip below 10%, and by the final subperiod of Early Modern English (EModE), they are at a mere 1%; throughout MBE the pattern makes up 2% of all main clauses in the database. The XVxS pattern – likely to be full-verb inversion – shows a continuous decline like XVS but starts out at a lower percentage (10%) and also ends up with a lower percentage (0%). Finally, unambiguous full-verb inversion pattern, XVVS, starts out as the smallest category with 4% in O2 and slowly decreases even further over time, ending at 1% in MBE. Interestingly, while it is a very infrequent option to start with, it does not disappear completely.
While the focus in this chapter is on inversion patterns in X-initial clauses, there are two further relevant developments. First, the proportion of subject-initial main clauses increases over time: from 46% in OE, to 57% in ME, 64% in EModE, and 74% in MBE.Footnote 8 In other words, the potential set of inverted subjects becomes smaller, which in turn means that inversion patterns overall become less frequent, and hence more noticeable. Second, among the group of X-initial clauses, the distribution changes of the types of clause-initial elements, each of which triggers inversion to a different degree. Most importantly, adverbs decrease between OE and MBE, from 52% to 27%, while PPs increase from 15% to 38%. NPs, including objects, also decrease, from 15% in OE to 6% in MBE, while subordinate clauses increase, from 9% to 17%.Footnote 9 Overall, then, adverbs and NPs become less prominent in clause-initial position, while PPs become more prominent.
4.4.2 Inversion Patterns per Period
In this section, I discuss the distribution of inversion patterns per period, focusing on frequencies of the five subject positions and the role of clause-initial elements. In the discussion that follows, I will focus on three phrasal categories – adverbs, PPs, and NPs – because these are (i) directly involved in full-verb inversion in all periods, and (ii) robustly present throughout all periods. For the two later periods, I will also analyse verbs (type of verb) and subjects (definiteness and weight) because no such data are yet available for those periods.
OE: Considerable Variation with Inversion as the Dominant Pattern
Subject-initial clauses account for 44% of all main clauses in the entire YCOE.Footnote 10 Among the X-initial clauses, adverb phrases (AdvPs, 48%) are the most frequent initial element, followed by PPs (17%) and NPs (15%).
Table 4.2 presents the overall frequencies of the five subject positions in all X-initial clauses and further specifies the frequencies for AdvP-, PP-, and NP-initial clauses in order to assess whether there is a specific association between clause-initial elements and subject positions. NPs have been divided into ‘NP-Object’ and ‘NP-Other’ because objects in particular are central to the V2-system in OE. The remaining initial elements, such as adjectives, participles, and subordinate clauses, are grouped under ‘XP-Other’.Footnote 11
Table 4.2Long description
The table presents linguistic analysis from the York-Toronto-Helsinki Parsed Corpus of Old English, abbreviated as Y C O E. The table focuses on subject positions and clause-initial elements. It is organized into six main columns with the following headers: Initial element, X S V V, X V S, X V x S, X V V S, and total. Each of these columns is divided into two sub-columns labeled frequency and percentage. The rows categorize different clause-initial elements: A d v P, P P, N P Object, N P Other, and X P other. A final row labeled total sums the data across all initial elements.
The row details from left to right are as follows:
For A d v P, the corresponding values are 3045, 27, 1246, 11, 5623, 50, 908, 8, 345, 3, 11167, and 100.
For P P, the corresponding values are 892, 28, 347, 11, 1607, 50, 165, 5, 197, 6, 3208, and 100.
For N P object other, the corresponding values are 462, 51, 52, 6, 335, 37, 43, 5, 13, 1, 905, 100.
For N P other, the corresponding values are 878, 39, 166, 7, 873, 39, 220, 10, 125, 6, 2262, and 100.
For X P other, the corresponding values are 1161, 33, 466, 13, 1425, 40, 439, 12, 60, 2, 3551, and 100.
For the total, the corresponding values are 6438, 31, 2277, 11, 9863, 47, 1775, 8, 740, 4, 21,093, and 100.
Overall, roughly only one third (32%) of the X-initial clauses are not inverted. It is fair to say, then, that based on frequency, inversion is a majority pattern in OE X-initial clauses. Among the inverted subjects, the largest category is the ‘mixed bag’ category of XVS (47%). This is true for all but the NP-initial clauses, where in fact the uninverted order is the largest (51% for objects; 39% for other NPs). The straightforward V2 subjects, XVSV, account for only 11% overall. Unambiguous full-verb inversion, XVVS, represents a small category, with only 4% overall; likely-to-be full-verb inversion clauses, XVxS, are a little more frequent, at 9%.
With respect to differences between initial elements, XVVS is a little more common with PPs and ‘NP-Other’, while XVxS is a little more frequent with ‘NP-Other’ and ‘XP-Other’. In contrast, all NPs have lower percentages for both XVS and XVSV. However, the differences are small and do not provide evidence for a particular association between a particular initial element and subject position.
In assessing canonical word order patterns in OE from a theory-based perspective, it is important to note that the largest category, XVS, is also the most diffuse category: it includes both subjects that are inverted because of V2 triggers and subjects that are inverted because of factors such as the type of verb or information status. While these factors have not been distinguished here systematically, it is clear that V2 triggers play a major role at this stage: a further investigation of the adverb-initial clauses shows that more than two thirds (69%) of the initial adverbs in XVS sentences belong to the so-called then-group (then, now, thus; Warner Reference Warner2007), known to invariably trigger inversion. As another factor, to be as a main verb seems to have some association with inversion: 23% of the XVS sentences and 30% of the XVxS sentences have to be, against only 14% of the XSV examples.
To conclude, in OE, inversion is clearly the majority pattern, even though uninverted orders are also quite frequent. In many contexts inversion is the only option (as part of the V2 system) or the most frequent option. Syntactically, XVVS and XVxS are infrequent but distinct categories of inversion; at the same time, these patterns stand out less than PDE full-verb inversion because inversion in general is much more common. Finally, while V2-inversion is determined by initial elements, full-verb inversion does not seem to be associated with a particular type of initial element.
ME: Increasing Variation and Development towards Preverbal Subjects
In the ME corpus, the overall percentage of subject-initial clauses is a little higher than in OE (57% compared to 44%). The proportions among the X-initial clauses have changed: AdvP-initial clauses have decreased (from 48% to 40%), as have NP-initial clauses (from 15% to 12%), while PP-initial clauses have increased (from 17% to 31%). Table 4.3 presents the distribution of the subject positions, again organised by clause-initial element.
Table 4.3Long description
Table showing the distribution of subject positions and clause-initial elements in the Penn Parsed Corpus of Middle English, version two, abbreviated as P P C M E subscript 2. The table is structured into six main columns representing syntactic structures: X S V V, X S V S, X V S, X V x S, X V V S, and total. Each column contains two sub-columns labeled frequency and percentage.
The rows are labeled with five categories of initial elements: A d v P, P P, N P object, N P other, and X P other. A final total row sums all values across initial elements and syntactic types.
The detailed data in the rows from left to right are as follows:
For A d v P, the corresponding values are 3089, 58, 404, 8, 1544, 29, 158, 3, 93, 2, 5288, and 100.
For P P, the corresponding values are 2026, 49, 374, 9, 1363, 33, 140, 3, 191, 5, 4094, and 100.
For N P object, the corresponding values are 109, 20, 63, 12, 319, 58, 39, 7, 16, 3, 546, and 100.
For N P other, the corresponding values are 754, 71, 58, 5, 199, 19, 32, 2, 25, 2, 1061, and 100.
For X P other, the corresponding values are 1026, 79, 116, 9, 126, 9, 19, 1, 10, 1, 1297, and 100.
For total, the corresponding values are 7004, 57, 1015, 8, 3551, 29, 381, 3, 335, 3, 12286, and 100.
Overall, preverbal subjects now constitute the largest category, accounting for just over half of all X-initial clauses (57%). Yet at the same time preverbal categories can also not yet be described as the default or norm, especially because there is considerable variation between the clause-initial elements: the ‘NP-Other’ and ‘XP-Other’ categories have a much higher percentage of preverbal subjects than the average (71% and 79%, respectively), while ‘NP-Object’ has a much lower percentage (20%). The decrease among inversion patterns is most striking for XVS, the ‘mixed bag’ category, now at 29% (down from 47%), although it is still the largest inversion category. The straightforward V2 pattern, XVSV, is a little smaller than in OE, with 8%, and the likely full-verb inversion pattern, XVxS, has also decreased, to 3%, the same percentage as unambiguous full-verb inversion, XVVS.
In the ‘mixed bag’ category, XVS, the influence of the then-group, as one trigger of V2 inversion, seems to have become smaller, accounting only for roughly 20% of adverb-initial XVS clauses. XVS and XVxS still have a higher percentage of to be, but the difference with uninverted subjects is smaller (29%, 26%, and 21%, respectively). Finally, the results show no obvious association between initial elements and late subjects: XVVS is a little more frequent after PPs (5% against 3% overall) and XVxS after ‘NP-Other’ (7% against 3% overall), but these differences are extremely small.
To summarise, inversion orders have decreased overall in ME but are still common, at a little under 50%. As in OE, the full-verb inversion patterns, XVVS and XVxS, represent infrequent but distinct patterns. What stands out most in this period is the extensive variation in the percentages of preverbal subjects and inversion patterns among different clause-initial elements. In other words, neither inversion nor non-inversion can be described as the default or canonical option, neither from a frequency-based nor from a theory-based perspective.
EModE: Preverbal Subjects as the New Norm, Inversion Increasingly Infrequent
In the EModE corpus, the overall proportion of subject-initial clauses has risen to 64%, meaning that only one third of clauses now have the potential to show inversion. In the X-initial clauses, the proportions of AdvP-initial clauses (38%), PP-initial clauses (27%), and NP-initial clauses (13%) have largely remained similar. Table 4.4 presents the distribution of subject patterns for the entire EModE corpus.
Table 4.4Long description
A table showing the distribution of subject positions and clause-initial elements in a corpus. The table is organized into six columns: X S V V, X V S V, X V S, X V x S, X V V S, and Total. Each column has two sub-columns labeled frequency and percentage.
The rows categorize five types of initial elements: A d v P, P P, N P Object, N P Other, and X P Other. A total row at the bottom summarizes all values across the initial elements. The data across the rows from left to right is as follows:
For A d v P:
The corresponding values are 2594, 73, 191, 5, 603, 17, 68, 2, 74, 2, 3530, and 100.
For P P:
The corresponding values are 2008, 78, 122, 5, 282, 11, 42, 2, 104, 4, 2558, and 100.
For N P object:
The corresponding values are 93, 56, 19, 11, 48, 29, 6, 4, 1, 1, 167, and 100.
For N P other:
The corresponding values are 612, 84, 37, 5, 62, 9, 2, 0, 16, 2, 729, and 100.
For X P other:
The corresponding values are 1822, 86, 77, 4, 207, 10, 9, 0, 8, 0, 2123, and 100.
For totals:
The corresponding values are 7129, 78, 446, 5, 1202, 13, 127, 1, 203, 2, 9107, and 100.
Preverbal subjects are now clearly the dominant category, accounting for over three quarters of the X-initial clauses overall (78%). ‘NP-Object’ has a much lower percentage (55%) and ‘XP-Other’ a much higher percentage (86%), but the range of variation is smaller, and in all clauses, preverbal subjects are the most frequent pattern. Among the inversion patterns, XVS is still the largest category, but it has become much less significant, with only 13% overall (with ‘NP-Object’ as the exception, at 29%). The continued decline of V2-type inversion is reflected in the low percentage for XVSV clauses: 5%. Finally, XVxS and XVVS lose a little more ground, at 1% and 2%, respectively.
By this time the V2 system has largely collapsed. Indeed, the straightforward V2 examples of the XVSV pattern in EModE are now restricted or archaic options, with many Bible translations and formulae such as in (16).Footnote 12
[To the worshipfull Robart Plompton, knight,] be this byll delivered in hast.
XVS, the ‘mixed bag’ category, is a little more frequent after adverbs (18%) and objects (29%), but is likely to contain far fewer cases of typical V2 inversion. Indeed, only 4% of the adverb-initial XVS clauses now contain an adverb from the then-group, as one of the typical V2 triggers. There are, however, other adverbs that occur quite frequently in the dataset, such as here in (17).
And [here] are the names of them that hath payd me; Robart Wood, Peter Cott, John Gloster, Robart Taler, William Bentham.
In other words, while the XVS category still contains leftover V2 patterns, other clause-initial triggers found in this period will continue to exist into PDE.
From the perspective of the defining characteristics of PDE full-verb inversion, two further factors are worth investigating. The first factor is the type of verb. The verbs in the PPCEME corpus were previously annotated for type of verb (for details, see van Kemenade Reference Kemenade, Nevalainen and Traugott2012: 828), which allows for a comparison of the inversion patterns in terms of the type of verb. Table 4.5 presents the results for the clauses with only a finite verb, distinguishing between to be, an unaccusative verb, or another verb.
Table 4.5Long description
The table presents the frequency and percentage of different types of verbs in three syntactic structures: X S V, X V S, and X V x S. The table is divided into three vertical sections, one for each structure. Each section contains two columns labeled frequency and percentage. The rows in the table are organized by type of verb: to be, unaccusative, and other. A fourth row labeled total summarizes the data within each structure. The data from left to right are filled as follows:
For to be, the corresponding values are 1527, 32.4, 950, 20.1, 2241, 47.5, 4718, 100, 560, 43.3, 31, and 19.5.
For unaccusative, the corresponding values are 950, 20.1, 356, 27.6, 83, and 52.2.
For other, the corresponding values are 2241, 47.5, 376, 29.1, 45, and 28.3.
For total, the corresponding values are 4718, 100.0, 1292, 100.0, 159, and 100.0.
Table 4.5 shows that almost three quarters of the inverted subjects occur with either be (43.3% for XVS; 19.5% for XVxS) or an unaccusative verb (27.6% for XVS; 52.2% for XVxS). For the preverbal subjects, these percentages are much lower, with just over half of the clauses occurring with be (32.4%) or an unaccusative verb (20.1%). In other words, unaccusative verbs and be are dominant in inversion patterns.
The second factor relevant to the diachrony of full-verb inversion are proxies for discourse function, in particular weight and definiteness of the subject. While Dreschler (Reference Dreschler, Los and de Haan2017) and Warner (Reference Warner2007) show that OE/ME late subjects are preferred by indefinite and long subjects, no comparable data are available for later periods. Table 4.6 presents the distribution of subjects of different length across the five subject positions.Footnote 13
Table 4.6Long description
The table presents the relationship between subject length and syntactic structure across five different structural types: X S V V, X V S Y, X V S, X V x S, and X V V S. Each structure is represented by two columns: frequency, labeled as freq. and percentage, labeled as %. The rows categorize the length of the subject in number of words, grouped into four categories: 1, 2, 3, and greater than 3 words. A row at the bottom labeled total summarizes the total frequency and percentage for each structure.
The data filled in the rows from left to right are as follows:
For 1 word, the corresponding words are 2086, 29, 120, 25, 225, 17, 14, 9, 5, and 2.
For 2 words, the corresponding words are 2688, 37, 188, 39, 372, 29, 29, 18, 47, and 14.
For 3 words, the corresponding words are 992, 14, 78, 16, 186, 14, 15, 9, 49, and 15.
For greater than 3 words, the corresponding words are 1539, 21, 102, 21, 509, 39, 96, 60, 232, and 70.
For the total, the corresponding values are 7305, 100, 488, 100, 1292, 100, 154, 100, 333, and 100.
The effect of weight is clearest for the XVVS category, with only 2% of subjects consisting of just one word and 70% consisting of more than three words. The same trend – few short subjects and many long subjects – is visible for XVxS and XVS, but the differences are smaller and two-word subjects are in fact quite frequent. Interestingly, the percentages for preverbal subjects and V2-inverted subjects are very similar, both showing a preference for one- or two-word subjects, although longer subjects are not uncommon. To further examine syntactic indicators of discourse status, I also compared definiteness of the subjects, with a definite article or demonstrative as an indicator of definiteness, and indefinite article or absence of definiteness indicators as corresponding to indefiniteness. However, these results did not show any observable differences between the subject positions. This is perhaps surprising considering the results for OE and ME, but not when we take into account Birner and Ward’s (Reference Birner and Ward1998) findings for PDE inverted subjects, which were equally as often definite as indefinite.
In summary, inversion is a minority pattern in EModE. The majority of the examples, also in the ‘mixed bag’ category of XVS, are no longer triggered by V2 factors; rather, while some can be specified by a specific triggering initial element, there is a large category that can be grouped together based on shared characteristics: inversion with either specific verbs and/or specific subjects. These factors are not new per se, but rather, because other factors are lost, they are now more evident as the defining factors of remaining inversion orders. All inversion patterns stand out more as a non-standard pattern now that preverbal subjects are the norm.
MBE: Preverbal Subjects as the Default, Inversion as a Rare Option
By the Modern English period, 74% of the main clauses are subject-initial. In terms of frequency, then, preverbal subjects are clearly the default, regardless of (non-)inversion in X-initial clauses. In the X-initial clauses, PP-initial clauses have further increased (from 27% to 38%), while AdvP-initial clauses have further decreased (from 38% to 27%), just as NP-initial clauses (from 13% to 7%). Table 4.7 presents the distribution of the five subject positions in X-initial clauses.
Table 4.7Long description
The table is divided into 6 columns and is labeled X S V V, X V S V, X V S, X V x S, X V V S, and total. Each column is divided into two sub-columns labeled frequency, freq., and percentage, %.
The rows are categorized by initial element: A d v P, P P, N P object, N P other, X P other, and a final total row that summarizes each column. The data of the rows filled from left to right are as follows:
For AdvP, the corresponding values are 1169, 88, 44, 3, 107, 8, 7, 1, 4, 0, 1331, and 100.
For P P, the corresponding values are 1712, 92, 36, 2, 75, 4, 3, 0, 27, 1, 1853, and 100.
For N P object, the corresponding values are 41, 75, 5, 9, 8, 15, 1, 2, 0, 0, 55, and 100.
For N P other, the corresponding values are 283, 93, 8, 3, 8, 3, 2, 1, 2, 1, 303, and 100.
For X P other, the corresponding values are 1136, 88, 15, 1, 130, 10, 2, 0, 4, 0, 1287, and 100.
For total, the corresponding values are 4341, 90, 108, 2, 328, 7, 15, 0, 37, 1, 4289, and 100.
Table 4.7 clearly demonstrates the dominance of preverbal subjects, with an overall percentage of 90% for XSV(V); only ‘NP-Object’ has a slightly lower percentage, at 75%. Among the inversion patterns, XVS is still the largest category, but it represents only 7% of the clauses. XVxS is by now almost extinct and XVVS is extremely rare. XVSV has a slightly higher percentage, which, as in EModE, is mostly due to the corpus containing some archaic examples and several Bible translations, where these orders are still relatively common.
Table 4.7 also shows that most of the XVVS examples (27 out of 37) occur in PP-initial clauses, which suggests an association between full-verb inversion and clause-initial PPs at this stage, although for XVS clauses, AdvPs and NPs seem more relevant. The PPCMBE corpus is not annotated for type of verb, but we can still examine the number of examples with be: in the clauses with a single verb and an inverted subject, be accounts for 54% of the examples, against only 36% of the uninverted subjects; in other words, be seems to have a larger influence than in EModE. Table 4.8 presents the distribution of subject length across subject positions.
Table 4.8Long description
The table is divided into 7 columns with the headings labeled as subject length in number of words, X S V V, X V S V, X V S, X V x S, and X V V S. The columns of subject position are sub-divided into two labeled frequency and percentage. The data in the rows filled from left to right are as follows:
For 1 word, the corresponding values are 1108, 26, 17, 16, 44, 13, 3, 20, 2, and 5.
For 2 words, the corresponding values are 1582, 36, 46, 43, 107, 33, 2, 13, 1, and 3.
For 3 words, the corresponding values are 585, 13, 17, 16, 41, 13, 4, 27, 4, and 11.
For greater than 3 words, the corresponding values are 1065, 25, 28, 26, 136, 41, 6, 40, 30, and 81.
For total, the corresponding values are 4340, 100, 108, 100, 328, 100, 15, 100, 37, and 100.
Table 4.8 shows the same tendencies as in the PPCEME corpus. The percentage of subjects longer than three words in XVVS has increased to 81% (against 70%), while two-word subjects especially have become less frequent (3% against 14%); most categories are largely unchanged, except for XVxS, which has few examples. With respect to definiteness, as in EModE, the percentages of definite and indefinite subjects are again largely the same for all subject patterns.
A final question is to what extent the data in the PPCMBE corpus already resemble the observations from studies on PDE full inversion. A further analysis of all inversion examples in the final subperiod (B3, 1840-1914; 176 examples) reveals that there are still a considerable number of sentences from Bible translations or otherwise archaic examples, as in (18).
Moreover, [herein] is revealed that of which thou didst erstwhile profess thyself ignorant.
The selection also contains some examples of quotative inversion and subject-auxiliary inversion, as well as initial adverbs such as then, so, here. However, apart from these specific types, many of the examples seem unremarkable from a PDE perspective. The majority of examples have be as the main verb, a passive verb or an unaccusative verb, and/or have a strong discourse-link, as in (19).
Prince Rupert threw away the best of his horsemen in attempts to break the solid masses of the London train-bands, who showed a steady power of resistance very admirable in such young soldiers. [In one of these desperate charges] fell Lord Falkland, the wisest and most moderate of the king’s councillors, who is said to have deliberately thrown away his life because of his sorrow at the long continuance of the war.
This example meets all descriptions for PDE: it has an unaccusative verb, a clause-initial element that has a strong link to the preceding discourse (charges refers to attempts), and a long (and discourse-new) subject.
In summary, in the MBE corpus, inversion has clearly become a non-standard option due to the dominance of preverbal subjects; it is non-canonical both in terms of frequency and in terms of the default word order options in the language. There are still some differences from PDE full-verb inversion, which is not surprising considering the fact that there is still a time gap of roughly a hundred years between the end of the final MBE subperiod and Birner and Ward’s (Reference Birner and Ward1998) study. Nevertheless, the majority of inversion patterns (XVS, XVxS, and XVVS) can be categorised together in terms of the characteristics of PDE full-verb inversion.
4.5 Conclusion
This chapter has examined the history of full-verb inversion in English, aiming to connect earlier research on full-verb inversion in PDE (often referred to as ‘locative inversion’) with research on full-verb inversion in OE and ME (known as ‘late subjects’). The central question was whether PDE full-verb inversion is syntactically a continuation of the OE-ME full-verb inversion pattern or whether it emerged as a new pattern, while I also wanted to address the question to what extent full-verb inversion already constituted a non-canonical word order option in earlier stages of English, as it does in PDE.
Syntactically, PDE full-verb inversion is a mix of remnants of V2 inversion and those clauses described as featuring late subjects in the literature, which represent a word order option that was already in use as a distinct pattern in OE. Full-verb inversion has developed a more specific profile over time through its increasing association with certain verbs, certain clause-initial elements, and certain subject characteristics. In OE, the two word order options in the corpus indicative of full-verb inversion – XVVS and XVxS – are part of a larger set of inversion orders, with which they overlap, while in the MBE corpus, the three remaining inversion patterns – including the ‘mixed bag’ category of XVS – are conflated through a shared set of characteristics, certainly syntactic and presumably information-structural. Based on the current study, it seems that part of the specialisation that characterises the full-verb inversion pattern may have happened quite late in the history of English, as the data from the MBE corpus, which runs until the beginning of the twentieth century, do not yet show these syntactic characteristics to the same extent as they have been reported for PDE by Birner and Ward (Reference Birner and Ward1998). Crucially, it is clear that PDE full-verb inversion is both a continuation of an older pattern and also a new phenomenon, in that the characteristics of both the verbs, the initial elements, and subjects involved in the pattern have developed over time, into a word order pattern with a more specific profile.
The results in this study also indicate that full-verb inversion has become more non-canonical over time, both from a theoretical and a frequency-based perspective: in OE, while XVVS was already a distinct and infrequent pattern, it stood out less as a non-canonical option because there were several more or less similar patterns, and inversion in general was more frequent than non-inversion was. As inversion became a minority pattern, any inversion pattern, including full-verb inversion, became more non-canonical. This is further strengthened by the increase in subject-initial clauses overall, which further limited the potential for inversion. From a pragmatic perspective, too, the unique opportunity of reordering arguments that full-verb inversion achieves is obvious in PDE, while in OE many other options were available to achieve a similar reordering effect. An interesting question for future research, then, is to what extent the increasing non-canonical status of full-verb inversion relates to its information-structural function. This, in turn, would shed further light on the motivation behind non-canonical word order patterns more generally.
Open access