1 Introduction

Old English (henceforth OE) was characterised by word order variation; in fact, as scholars such as van Kemenade (1987), Pintzuk (1999), Roberts (1997), among others, have noted, there is variation in the respective order of Verb (V) and Object (O), as well as in the respective order of V and Auxiliary (Aux), as examples (1)–(4) show. The Auxiliary and the non-finite verbs are indicated through underlining, whereas the objects are given in italics¹:

By the start of the Middle English (henceforth ME) period, VAux orders have become rare (Struik 2022; Struik & van Kemenade 2022), but there is still OV/VO order variation, even though non-pronominal objects are found in the majority of cases in postverbal position (with textual and regional differences, see De Bastiani 2020, 2022):

Variation also affects the position of pronominal objects, which are found before the auxiliary, between the auxiliary and the non-finite verb, and after the finite verb, as examples (7a–c) show:

The frequency of AuxVO orders increases in the course of the ME period, but there are remnant OV orders up until the late ME period, which present distinct features. Firstly, these orders are mostly found in poetry, with their frequency being six times higher than in prose around 1320, to eventually become, by 1470, twenty times more frequent than in prose. Preverbal objects tend to be short, adjacent to the verb and represent given or inferable information (Foster & van der Wurff 1995; Moerenhout & van der Wurff 2000, 2005). Some of the orders Foster & van der Wurff (1995) and Moerenhout & van der Wurff (2000, 2005) analyse as OV might be instances of scrambling, however, as example (8) shows, where the negated object precedes the finite verb. This object is clearly moved above TP, probably to express contrastive focus:

By 1500–50, OV order is limited to three constructions: clauses with an auxiliary and a quantified or negated object, and clauses lacking an overt subject, as work by Moerenhout & van der Wurff (2000, 2005) and Ingham (2000, 2002) shows:

Negated and quantified objects persist in preverbal position the longest, up until the Early Modern English period (Foster & van der Wurff 1995; Moerenhout & van der Wurff 2000, 2005; Ingham 2000, 2002, 2007; Pintzuk & Taylor 2006). Ingham (2002) shows, however, that OV order in 15th-century non-literary correspondence is essentially restricted to what he calls embraciated objects: objects that are found between the tensed verb and a non-finite verbal form. Moreover, 90% of these objects are negated, whereas preverbal quantified objects constitute a residual word order. Prior to the 15th century, as Moerenhout & van der Wurff (2000, 2005) show, quantified objects are also placed in preverbal position, but it must be kept in mind that the instances in poetic texts far outnumber those in prose.

This evidence led scholars to conclude that the syntax of quantified and negated objects is different from that of non-pronominal positive objects, since the latter are virtually all placed in post-verbal position by the end of the ME period (cf. Taylor & Pinztuk 2006). Evidence from OE, however, might point at different conclusions. In fact, Ingham (2000, 2002, 2007), Struik (2022) and Struik & van Kemenade (2022) conclude that higher OV rates of quantified and negated objects in OE are probably an effect of translation, and that negated and quantified objects in non-translated texts should not be treated any differently from other kinds of objects. For OE, Struik (2022) finds out, in fact, that quantified and negated objects surface preverbally in comparable rates to other objects in non-translated texts, whereas they are more frequent in translations. These conclusions open up new questions for research, since the quantitative work by Pintzuk & Taylor (2006) highlights a higher frequency of preverbal quantified and negated objects compared to positive objects. It must be noticed, however, that the number of negated and quantified objects in preverbal position is not high in absolute terms and that, already in the Early ME period, they are found in postverbal position in around 60% or 70% of the cases (Pintzuk & Taylor 2006, 275–276).² Pintzuk & Taylor (2006) do not consider possible influence from translations a factor in their research, thereby leaving some space for further discussion of quantified and negated objects. I will focus on this factor in the present contribution.

The empirical data briefly presented in examples (1)–(11) have received different syntactic treatments in the literature, and the accounts given differ with respect to both the underlying base order of OE and the triggers for language change. In a nutshell, there are accounts which consider OV the base order, with different rightward extraposition movements or Verb (Projection) Raising to account for VO orders. Scholars such as Pintzuk (1999), Kroch & Taylor (2000), Pintzuk & Taylor (2006), Taylor & Pintzuk (2011, 2012a, b, 2015) have argued for the Double Base hypothesis, according to which grammars which vary in the headedness of the IP and VP projection are in competition throughout the OE period, until the head-initial grammar eventually gains in frequency and becomes the base one. Finally, scholars such as Roberts (1997), Fischer et al. (2000), Biberauer & Roberts (2005), Hinterhölzl (2014, 2015, 2017), Elenbaas & van Kemenade (2014), Struik & van Kemenade (2018, 2022), De Bastiani (2019, 2020, 2022) Struik (2022) have argued for a Universal Base Hypothesis accompanied by different leftward movement operations to derive OV orders.

Against this empirical and theoretical background, the goal of this paper is to determine which constraints govern preverbal placement of negated and quantified objects in Late ME and Early Modern English, and to embed them within a head-initial account of word order variation in the history of English. In Section 2, the constraints governing OV and VO order in OE and Early ME are presented; these constitute the backdrop against which the data in Section 3 will be compared, where negated and quantified objects in ME and Early Modern English are analysed. In this section, I provide a comparison of data from two corpora, the PPCME2 and the PCEEC corpus, which highlights an asymmetry between translated and non-translated texts in the preposing of quantified and negated objects in late ME. Moreover, this Section provides quantitative evidence for the Early Modern English period that confirms Ingham's (2000, 2002, 2007) findings and sheds further light on the relevance of the Negative Cycle for the preposing of negated objects. Section 4 evaluates two main theoretical derivations of residual OV orders in ME and Early Modern English; I will argue that in OE and Early ME, quantified and negated objects do not behave differently from positive objects, thereby corroborating Struik & van Kemenade's (2022) analysis. I will furthermore argue that different rates of preposed quantified objects in the two corpora examined might be linked to the influence of translations. Section 5 concludes the paper.

2 Constraints governing preverbal placement of positive objects in OE and ME

Word order variation in OE has been the subject of a decade-long debate in generative historical linguistics, and, as mentioned briefly in Section 1, the syntactic accounts either resort to an OV or VO base word order, or rely on the grammars in competition scenario (Kroch 1989).³ Pintzuk (1999) observed that an account postulating an OV base with rightward extraposition movements would not account for cases in which light elements, such as pronouns and demonstratives (cf. example (7c) in Section 1), are found to the right of the verb, since these are not usually extraposed in OV languages such as Dutch, for instance. Given the variation presented in Section 1, she argues for a scenario of grammars in competition, which vary according to the headedness of the IP and VP projections. By combining a head-initial or a head-final IP and VP, four logical possibilities are derivable, as shown in (12a–d):

As can be seen from example (12d), the fourth logical possibility is not only not attested in OE, but is also rare across the world languages (it is in fact ruled out because of the Final Over Final Constraint, put forth by Biberauer et al. 2014). Pintzuk's (1999) proposal accounts for the variation attested, but the factors ultimately driving the head-initial grammar to “win” the competition are not clear. Pintzuk (1999) shows that this grammar gains in frequency through time, and other scholars adopting this framework such as Kroch & Taylor (2000) and Trips (2002) have argued that the SVO grammar is introduced and eventually spreads due to language contact with the Scandinavian settlers in the Danelaw. It is not the aim of the present paper to enter this discussion, it suffices to say that both internal and external causes to this language change have been put forth.

Finally, the accounts building on the Universal Base Hypothesis take the syntax of OE to be uniformly VO, with different leftward licensing movements to derive OV order.⁴ The accounts put forth by Roberts (1997) and Fischer et al. (2000) have in common leftward licensing movements to check strong case features. In the proposal by Fischer et al. (2000), it is spell out that is crucial: if it takes place before movement, then we obtain VO word order; we obtain OV order if spell out takes place after movement. Fischer et al. (2000) also show that there are sentences in OE that are clearly instances of leftward movement operations, since objects are raised above adverbials. If such objects can be preposed, nothing prevents us from postulating that movement operations have also taken place in sentences with OV order lacking such a clear signpost of leftward movement.

Recent investigations have shown that there is indeed evidence to sustain the claim that OV order is derived, the relevant trigger being IS and weight interface conditions rather than case-checking, thereby giving support to the Universal Base Hypothesis. Preposing of positive pronominal and non-pronominal objects is in fact triggered by information structure and weight of the constituents (Elenbaas & van Kemenade 2014; Struik & van Kemenade 2018, 2022; Struik 2022; De Bastiani 2020, 2022). These works follow the methodology developed in Pintzuk & Taylor (2006), Taylor & Pintzuk (2011, 2012a, b, 2015): sentences with an auxiliary and a non-finite verb are analysed in order to track both the position of the object with respect to the non-finite verb, and the position of the auxiliary/modal verb with respect to the non-finite verb. The approaches by Elenbaas & van Kemenade (2014), Struik & van Kemenade (2018, 2022), Struik (2022), De Bastiani (2020, 2022), however, differ from the one in the works by Taylor & Pintzuk, in that no double-base hypothesis is postulated, but OE is analysed as a head-initial language with leftward licensing movements to derive OV word order.⁵

Objects in AuxOV sentences present uniform features: they have a given information structural status, either because they have been mentioned in the previous context, or are part of encyclopaedic knowledge, or are inferable. Preverbal objects are also usually light.⁶ Objects in AuxVO sentences are instead heterogeneous both in terms of their prosodic weight and their IS value. This is taken in Struik & van Kemenade (2018) as a strong indication for an underlying AuxVO syntax in OE. Example (13) illustrates an AuxOV sentence with a given preverbal object:

As can be seen from the context preceding this sentence, given in (14), the referent writings/books has already been introduced in discourse:

The evidence briefly presented here is therefore analysed through a Universal Base account combined with pragmatic and prosodic interface conditions postulated by Hinterhölzl (2015, 2017), given in (15)–(16):

These conditions interact with spell out at LF and PF, determining the spell out of either the higher or the lower copy after the movement operation has taken place; according to Struik (2022), objects are attracted to Spec, vP and this movement is triggered by an Edge Feature. If we combine the two approaches, we get the derivations depicted in (17)–(18):

As for the VAux clauses in OE, there are indications that not only the objects, but also the whole sentence conveys backgrounded information (Milicev 2016; Struik & De Bastiani 2018); more research is however needed to corroborate this claim.

The previous context of example (19) makes clear that the act of fasting constitutes information already given, and the relative clause constitutes the background for the sentence headed by the discourse marker þa, by anchoring it to a specific point in time. The main clause headed by þa progresses the narration (cf. also van Kemenade & Los 2006; van Kemenade 2009 about the discourse markers þa and þonne).⁷ VAux clauses decrease significantly in the transition from OE to Early ME English. By 1150–1250, they have decreased significantly to the point of being virtually non-existent (Struik & van Kemenade 2022). Struik (2022) analyses these clauses, building on Biberauer & Roberts (2005), by assuming that the object is raised to Spec,vP of the embedded non-finite verb and that the whole vP is further raised to the Spec,vP of the matrix restructuring verb.⁸

It has been shown that OV order with positive objects is constrained by either IS or prosodic weight of the object already in OE. There is, moreover, reason to claim that the order VAux was already declining in frequency in OE, and that this order can also be explained by derivation from an AuxVO base with different information-structurally driven operations. In Early ME, there is still (Aux)OV/VO order variation and objects in AuxOV sentences are given in terms of their information structural value (Elenbaas & van Kemenade 2014; Struik & van Kemenade 2022). A progressive loss of OV orders can however be observed; Struik & van Kemenade (2022) select sentences with an auxiliary and a non-finite verb and show that first VAux orders are lost, and that, by 1500, AuxVO is the main word order pattern, while AuxOV orders constitute residual strategies. While the syntax of AuxOV clauses in Early ME is driven by the same IS requirements, there is also evidence that shows that non-pronominal objects are predominantly found in post-verbal position in Early ME texts (De Bastiani 2020, 2022). This finding is also confirmed by Pinztuk & Taylor (2006, 259), who report that non-pronominal positive objects in the Helsinki period M1 (1150–1250) surface pre-verbally in only 28% of the cases. The texts examined in De Bastiani (2020, 2022) differ from one another in the positioning of pronouns. In the texts which are more advanced in the post-verbal spell out of all objects, Peterborough Chronicle and Katherine Group, these are also predominantly postverbal, as shown in De Bastiani & Hinterhölzl (2020). As a matter of comparison, in the texts of the South-East Midlands area, which are more conservative (Kroch & Taylor 2000), pronouns surface in post-verbal position in 22.9% of the cases in main clauses, and in 13% of the cases in subordinate clauses (De Bastiani & Hinterhölzl 2020). The distribution in the texts of the Katherine Group is very different: pronouns are found in post-verbal position in 70.4% of the cases in main clauses, and in 43.8% of the cases in subordinate clauses (De Bastiani & Hinterhölzl 2020). The data are taken to indicate that the G-Transparency condition is subject to blurring, since most non-pronominal objects are already postverbal, and that in those varieties that retain preverbal pronouns it is the Weight Transparency condition that is responsible for postverbal spell out of heavy elements (cf. De Bastiani 2020, 2022). As the texts of the Katherine Group and the Peterborough Chronicle show, the Weight Transparency condition is subject to blurring in these varieties, since pronouns are also progressively being spelled out in post-verbal position with a higher frequency.

The data summarised in this section were analysed employing the same methodology De Bastiani (2020, 2022) used for OE: clauses with a finite and a non-finite verb were selected in order to track the linear position of the object. The non-finite verb and the auxiliary, and referential non-negated and non-quantified objects were annotated for their IS value, whereas all objects were annotated for their syntactic weight (cf. De Bastiani 2020, 2022).

The evidence presented in this section shows that in Early ME, leftward licensing movement is governed by different constraints than in OE; in fact, VAux clauses have become rare and non-pronominal objects are predominantly postverbal. Moreover, in the more advanced varieties, pronouns are also found in postverbal position in the majority of cases. This has been taken to indicate in De Bastiani (2020, 2022) that the G-Transparency condition is overruled by the Weight Transparency Condition in the M1 period, the latter being also subject to blurring in the more advanced varieties.

In this section, the discussion revolved around positive referential objects, which can be analysed for their IS value. In the later periods, as already shown by Foster & van der Wurff (1995), Moerenhout & van der Wurff (2000, 2005) and Ingham (2000, 2002), evidence for OV is mostly represented by quantified and negated objects. The same IS analysis is not applicable to these items since they do not have a referring potential on their own (cf. Petrova & Speyer 2011; Struik & van Kemenade 2018). These have in fact been treated separately in investigations on OV/VO alternation in the history of English (Pintzuk & Taylor 2006). A confirmation of this is given in Ingham (2002), who examines the IS value of preposed negated and quantified nouns and concludes, albeit acknowledging the limitations of such an analysis of NPs that are negated or quantified, that their IS status is not relevant for the leftward licensing, since they do not appear to correlate with givenness. We now turn to the analysis of these items in the next section.

3 A qualitative analysis of negated and quantified objects

Of all object types, quantified and negated objects have been shown to persist in preverbal position the longest, up until the Early Modern English period. This has been taken to indicate that quantified and negated objects are subject to different derivations than positive objects (van der Wurff 1999; Pintzuk & Taylor 2006). If negated and quantified objects persist in preverbal position longer, it is reasonable to postulate that they are subject to different constraints, but as Struik (2022) has shown, it is also crucial to understand whether their placement can be ascribed to translation effects, a distinction which Pintzuk & Taylor (2006) do not make. Therefore, I present a quantitative and qualitative investigation of negated and quantified objects with the goal of determining whether influence from translations can be detected, and whether these types of objects behave differently from positive objects.

To this end, diagnostic structures, i.e. AuxOV main and subordinate clauses from the PPCME2 (Kroch, Taylor & Santorini 2000) and PCEEC (Taylor et al., 2006) corpora have been extracted.⁹ The sentences have been selected in order to obtain structures comparable to the ones analysed in relation to positive objects in OE and Early ME, i.e. main and subordinate clauses with a finite and non-finite verb. For the investigation in this section, I am concentrating on quantified and negated preverbal direct and indirect objects, in order to obtain data comparable to the analysis of Pintzuk & Taylor (2006). Only AuxV structures are elicited in this search, since the VAux order had become extremely rare by the start of the ME period (Struik & van Kemenade 2022).

The structures analysed are exemplified in (20)–(21), where a subordinate clause with a preverbal quantified object and a main clause with a preverbal negated object are illustrated respectively:¹⁰

Single word objects such as none, all and combinations such as no thing, any thing are also included in the search.¹¹

The analysis starts with the PPCME2 corpus, which comprises the whole ME period and consists of both translated and non-translated texts. This corpus comprises 56 text samples (as illustrated at the philological description section, https://www.ling.upenn.edu/hist-corpora/PPCME2-RELEASE-4/index.html), of which 22 (i.e. 39.2% of the texts in the corpus) are indicated as translations in the description of the samples provided by the corpus compilers. The period which has the highest rate of translated texts is the M2 (1250–1350), and this period is known for being problematic in the literature on ME.

As a first step, I provide a quantitative distribution of quantified and negated objects in main clauses, divided per Helsinki periods (Figure 1).

View Full Size

Percentage distribution of all preverbal quantified and negated objects in main clauses divided per time periods, PPCME2 corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

As can be observed from Figure 1, the most hits (= 47.6%) of quantified and negated objects in AuxOV main clauses come from the M1 (1150–1250) period, which still presents OV/VO variation with positive objects, as illustrated in Section 2. The texts in this time-frame are all native compositions except for one text, and the only text whose syntax is subject to different constraints is the text of the Ormulum, a poetic text with a very precise metrical structure.¹² The distribution of quantified and negated objects in subordinate clauses mirrors that in main clauses, as Figure 2 illustrates.

View Full Size

Percentage distribution of all preverbal quantified and negated objects in subordinate clauses divided per time periods, PPCME2 corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

Table 1 compares the rate of AuxOV sentences with the total amount of clauses per each time period in the PPCME2 corpus.¹³

The data in Table 1 confirm the observation that AuxOV order decreases steadily in the course of the ME period (cf. Pintzuk & Taylor 2006; Struik & van Kemenade 2022). The M1 period presents the highest percentage of AuxOV structures, compared against the common denominator of the number of all main and subordinate clauses in each time-period provided in the PPCME2 corpus. Moreover, the evidence presented in Section 1 shows that OV/VO variation with positive objects is characterized by IS and prosodic constraints in this period. By the M4 period, AuxOV structures have become relics (cf. also Struik & van Kemenade 2022).

Table 2 presents the percentage of sentences with a preposed negated or quantified object compared to the number of AuxOV sentences in the relevant ME period.

Table 2 shows that in the M4 period, if an object was preposed, it was either negated or quantified in half of the cases, whereas in the M1 period, AuxOV sentences present a quantified or negated object in 10.2% and 6.7% of the cases in main and subordinate clauses, respectively. In other words, whereas different kinds of objects were preposed in the M1 period, in the M4 period preposed objects were most likely quantified or negated. It must be noticed that in the M3 and M4 periods, when an object is preposed and not quantified or negated, it is in most cases a pronoun. This is consistent with the framework illustrated in Section 2 and, furthermore, confirms the observation that OV order persists the longest with quantified or negated elements (cf. the literature quoted in Section 1 & 2).

Given that the earliest ME period, M1, is still characterized by OV/VO variation with both pronominal and non-pronominal objects, albeit with dialectal differences (cf. for instance Kroch & Taylor 2000; De Bastiani 2020), the next question to be asked is which constraints govern the preposing of objects in the subsequent periods. The distribution of quantified and negated objects in the periods M2–M4 according to the division native/non-native text is presented in Figures 3 & 4. This distinction is relevant in light of the evidence presented by Struik (2022), and Struik & van Kemenade (2022).

View Full Size

Percentage distribution of all preverbal quantified and negated objects per text type in the M2-M4 periods, main clauses

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Percentage distribution of all preverbal quantified and negated objects per text type in the M2-M4 periods, subordinate clauses

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

As can be seen from Figures 3 & 4, the hits from the Helsinki periods M2–M4 (1250–1500) are mostly concentrated in translated texts, or in texts which are not coded as translated in the information provided by the PPCME2 corpus compilers, but are based on foreign sources. The M2 period is not particularly revealing, since it only comprises one native text, whereas the M3 and M4 periods are more balanced, and more insights can be gathered by examining the hits from these two periods. The M3 and M4 periods do not consist of exclusively translated texts, unlike the M2 period, and there are some texts that do not show any OV orders with negated or quantified objects at all, such as Middle English Sermons, for instance. As can be observed from Figures 3 & 4, most hits are found in translated texts, but the M3 and M4 periods (1350–1500) also have some hits that come from native texts. Most of these share one common feature: the authors of these texts are linked to the Norfolk area. It is the case of Julian of Norwich's Revelations of Divine Love and the Book of Margery Kempe. A further five examples do not share these features, however, and they come from two texts, dating to the end of the ME period: Richard Fitzjames’Sermo die Lune and Gregory's Chronicle.

Apart from the two texts mentioned above, the data examined so far reveal that most of the negated and quantified preverbal objects are found in texts translated from French or Latin, or which constitute rewriting of foreign sources. Gianollo (2016a, b) shows that negated elements are strongly preferred in preverbal position in Late Latin, and this could have had an effect on texts translated from this language. For the texts translated from French, a systematic comparison is needed in order to determine to what extent this language influences preposing of negated and quantified objects. Evidence from Old Italian, however, shows that bare quantifiers strongly favour preverbal position in this Romance language (Garzonio & Poletto 2018). This can be taken to indicate that in Romance languages, quantifiers prefer pre-verbal position.

So far, negated and quantified objects have not been treated separately. The distribution of quantified vs. negated objects in main and subordinate clauses is given in Figure 5.

View Full Size

Raw counts for preverbal quantified and negated objects, main and subordinate clauses, PPCME2 corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

Figure 5 shows that the distribution of quantified and negated objects is different across the ME periods. In fact, the majority of quantified objects are found in the M1 period in both main and subordinate clauses; in the M3 and M4 periods, the number of negated objects is higher than the number of quantified ones. This shows that the preposing of quantified objects was diminishing at a faster pace when compared to the preposing of negated objects and confirms Ingham's (2002) findings.

How can we account for the data analysed so far? For the first period, which is still characterized by OV/VO variation, we can postulate that the preverbal placement of negated and quantified objects is subject to similar requirements as positive objects. One can suppose that these objects are attracted to Spec,vP in order for them to check a uNeg or uQuant feature located in T (Struik 2022), as shown in (22):

Tables 1 & 2 show, in fact, that object preposing in the M1 period is not limited to negated and quantified objects, which constitute around 10% and 6% of the preposed objects in AuxOV main and subclauses respectively. The proportion of preposed quantified and negated objects increases through time, paralleling the decrease of AuxOV structures in general. This finding can be taken to indicate that these elements are indeed governed by different syntactic constraints if compared to positive objects, but Figures 3 & 4 show that most of the hits from the M2–M4 periods are found in translated texts, whereas native texts presenting quantified and negative elements in AuxOV sentences come from the Norfolk area. For the latter ME period, it can be hypothesised that at least in the variety spoken in the Norfolk area, attraction to Spec,vP and preverbal spell out are still active for these objects, but according to Ingham (2000, 2002, 2007), the development of negation is also intertwined with the preverbal placement of negated elements. Ingham (2000, 2002, 2007) links the persisting of negated objects in AuxOV structures to the development of the Negative Cycle in the course of the late ME and Early Modern English period. The fact that in later periods the number of negated objects is higher than that of quantified ones can be taken as a confirmation of Ingham's (2002) analysis. This point will be further discussed in Section 4.

Whereas the PPCME2 corpus comprises both native and translated texts across the whole ME periods, the PCEEC corpus (Taylor et al., 2006) contains only native compositions; this corpus spans across the M3 and M4 periods and, moreover, comprises the whole Early Modern English period, ranging from 1350 to 1710. This corpus consists of letters; letters provide a valuable source for syntactic analysis because their syntax can be expected to be much more representative of the spoken vernacular. Although they can, of course, differ in degree of formality (cf. Davis 1971; Hanham 1975), they are unlikely to show any influence from the kind of stylistic awareness and monitoring expected in literary texts, or to show interference from translations. Ingham (2000, 2002) shows that in the collection of the Paston Letters OV is only productive with negated objects. Ingham (2002) samples other letter collections, but his investigation is limited to a subset of letters, since a parsed corpus of Early English correspondence was not available at that time. It is therefore worth applying the extraction of data to the whole PCEEC. In fact, this corpus lets us zoom into the periods for which we found negated and quantified objects in non-translated texts in the PPCME2 corpus and follow their distribution in later periods.

In the same way as for the investigation of the PPCME2 corpus, this corpus was also searched for AuxOV structures with a direct or indirect quantified and negated object. In Figures 6 & 7 the distribution across time frames is given.¹⁴ As Figures 6 & 7 show, most of the hits from this corpus come from the late M3 period; the majority of examples come from the Paston correspondence (1425–ca.1519), which, moreover, comprises the most extensive collection of texts. It should be noted that this family was also based in Norfolk. Ingham (2002) also samples the Stonor, Cely and Plumpton collection, but finds only a handful of examples of negated and quantified objects. The corpus query lets us retrieve a couple more hits for these correspondence collections, 9 for the Stonor collection and 7 for the Cely collection, respectively. Our calculations shows that the frequency with which negated (and quantified) elements surface in the sentences of the Paston and the Stonor letter collections amounts to circa 0.20% in both.¹⁵ These two letter collections therefore have a similar frequency of preposing of negated elements; this piece of evidence shows, furthermore, that the Norfolk area is not the only area for which we find preposed negated and quantified objects in native texts, as the data in the PPCME2 corpus seem to suggest. In fact, the Stonor letter collection, which comes from a family based in Oxfordshire, shows a similar rate of preposing of such objects as the more extensive Paston collection. It is, moreover, important to underline that Ingham (2002) argues that evidence for OV in these texts is scant overall, since it is limited to formulaic expressions such as so God me helpe (Ingham 2002, 297).

View Full Size

Percentage distribution of all preverbal quantified and negated objects in main clauses divided per time periods, PCEEC corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Percentage distribution of all preverbal quantified and negated objects in subordinate clauses divided per time periods, PCEEC corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

Table 3 shows that the overall percentage of AuxOV structures declines steadily in the Early Modern English period.

The percentage of AuxOV sentences with quantified and negated objects compared to the overall number of AuxOV sentences is given in Table 4.

Table 4 shows a striking asymmetry if data are compared to the same data from the PPCME2 corpus; in fact, in the PPCME2 corpus, the percentage of negated and quantified objects in AuxOV sentences corresponds to 17.7% of AuxOV main clauses and to 18.1% of subordinate clauses, respectively, for the M3 period. For the M4 period, AuxOV sentences with quantified and negated objects make up almost half of the hits for main and subordinate AuxOV clauses. Despite what the data from the PPCME2 corpus seem to suggest, then, preposing of negated and quantified objects is found also in native texts. Moreover, the evidence from the PCEEC corpus shows that in the M3 and M4 periods, when an object is preposed, it is most likely to be either quantified or negated. The separate figures for quantified and negated objects are given in Figure 8.

View Full Size

Raw counts for quantified vs. negated objects, main and subordinate clauses, PCEEC corpus

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

As can be observed from Figure 8, the negated elements outnumber quantified elements in the PCEEC sample. Moreover, the percentage of quantified preverbal objects in the PCEEC texts from the M3 and M4 periods is lower, if compared to the percentage of quantified preverbal objects in the PPCME2 texts from the same periods, cf. Table 5.

The evidence in Table 5 indicates an asymmetry between the two directly comparable time-periods in the queried corpora. The PPCME2 corpus, which is mixed in terms of the nature of texts, shows a higher amount of preposed quantified elements if compared to the native texts of the same time frame. This can be taken to indicate that native usage of the language preferred the preposing of negated elements by the end of the ME period (cf. also Ingham 2002). Preposing of quantified elements is still a grammatical strategy in late ME, but native texts show that negated elements are favoured in preverbal position by the late ME period. The higher amount of quantified elements in the PPCME2 corpus can be therefore linked to the influence of translations.

If we combine the data from the PPCME2 and PCEEC corpora, we can observe the distribution of quantified and negated objects across time. Figure 9 shows that quantified objects are found at lower rates than negated objects and their frequency decreases from the M1 to the M2 period. Negated objects raise in frequency through the M3 period, then drop in the course of the Early Modern English period. I will discuss in Section 4 that this development ties in well with Ingham's (2002, 2007) analysis. Before moving to the discussion of the data, I discuss the evidence in the Cely collection, which yields the majority of hits of preposed quantified and negated objects in the M4 PCEEC period.

View Full Size

Overall distribution of preverbal quantified and negated objects, corpora combined

Citation: Acta Linguistica Academica 70, 2; 10.1556/2062.2023.00650

• Download Figure
• Download figure as PowerPoint slide

All of the hits from this collection of letters present a similar structure, cf. (23):

Sentences with a similar structure are also found in the Paston collection, cf. (24):

In these sentences, the negated object none seems to act as a clausal negation: a topic is introduced at the beginning of the clause, and the sentence containing the negated element states that as far as the topic is concerned, there is nothing that can be done about it. Even though the element none is syntactically a negated object, its function is that to negate the clausal predicate, by means of an implicature. In fact, the authors of the sentences in (23)–(24) state that they cannot write about even one piece of the news that the recipient is waiting for (23), or that they cannot find even one piece of cloth better than the one sent (24). As is known in the history of negative cycle, Negative Polarity Adverbs act in Jespersen's stage II as triggers for scalar implicatures, which are eventually conventionalised and then used as neutral negation (Breitbarth, Lucas & Willis 2020; Magistro, Crocco & Breitbarth 2022). Notice that Ingham (2002) dismisses these expressions as formulaic, but I will argue that they are relevant in the scenario he proposes. They are indeed fixed expressions, and might have been a stylistic feature of correspondence. However, their structure is clearly discourse configurational: the sentence opens with a topic, which restricts the reference the proposition is about, and the negated object, similarly to the Negative Polarity Adverbs investigated by Breitbarth, Lucas & Willis (2020), gives rise to a scalar implicature. Incidentally, notice that also the negator not in English has a similar etymology: nawith (not a thing/being). I therefore claim that these structures, though formulaic, are conventionalized implicatures that negate a whole proposition centred on the topic introduced at the beginning of the clause.

Finally, in the E1–E3 periods, negated (and quantified) objects surface preverbally in very small numbers: I retrieved 13 examples between main and subordinate clauses, in a subcorpus which comprises almost 2m words, showing that by the beginning of the Early Modern English period, preverbal spell out of such elements has become a residual strategy.

Basing ourselves on the evidence given in Section 2, and the literature discussed in Sections 1 & 2, it is evident that neither the IS nor the Weight interface conditions can account for their licensing. If we abstract away from translated texts, negated and quantified objects are licensed in preverbal position in a period in which OV/VO variation was still at play, i.e. in the M1 period. In later periods, preposing is limited to negated objects. In Section 4, we turn to the question of derivation of such orders within the wider scenario of the loss of linear OV order in the history of English.

4 Discussion of the data

The results presented in Section 3 corroborate Ingham's (2002) findings, and shed new light on the distribution of negated and quantified objects in ME. In fact, for the M2–M4 periods, these mostly come from translated texts in the PPCME2, whereas they are found in native texts in the earlier period, re-surfacing in the M3 and M4 periods in native texts composed in or linked to the Norfolk area, at least in the dataset from the PPCME2 corpus. The distinction between translated and non-translated texts has not been drawn in the study of Pintzuk & Taylor (2006), but this contribution shows that it is relevant for the ME period, as shown for OE by Struik (2022). Furthermore, by the Late ME period, most of the hits comprise negated elements, thereby corroborating Ingham's (2002) observations. Finally, the items in the Cely collection indicate that preverbal negated objects are used, at least in the variety spoken by the Cely family, in specific negative expressions, and I argue that this finding is coherent with the development of the Negative Cycle proposed by Ingham (2007). The direct comparison of the data from the M3 and M4 time frames in the two corpora, moreover, allows us to detect an asymmetry in the preposing of quantified and negated objects in native and non-native texts.

Regarding OV/VO variation in the history of English, Struik (2022) and van Kemenade & Struik (2022) propose that objects are attracted to Spec,vP via insertion of an Edge Feature. When the relevant feature is not present, no movement takes place, as they show that the underlying syntax of OE is already VO. In De Bastiani (2020, 2022), I instead argued that objects are covertly moved outside the VP and that IS and prosodic interface conditions regulate the spell out of either the higher or the lower copy after movement. In Section 2, it was shown that both the G- and the Weight transparency conditions were operative in the OE period, whereas in early ME, it is the Weight transparency condition that accounts for the persistent preverbal placement of pronouns, at least in those varieties that are less advanced in the progressive post-verbal spell out of all object types. This lets us postulate a uniform leftward movement operation to Spec,vP, as proposed in Struik (2022); the spell out of the objects is, however, licensed under different conditions in the case of quantified and negated objects. These are taken to be raised in order to check either a uNeg or a uQuant feature in Struik's (2022) and Struik & van Kemenade's (2022) accounts.

The data presented in Section 3 show that in the M1 period, the number of quantified elements in preverbal position is higher than the number of negated elements in the same position; moreover, quantified and negated elements are not the only elements that are preposed in this time frame, which is mostly represented by native texts in the PPCME2 corpus. This data is analysed as a confirmation of what Struik (2022) finds out for non-translated OE texts: in a subsample composed of native texts, where OV/VO variation with non-pronominal and pronominal positive objects was still at play, preposing of negated and quantified objects does not differ significantly from positive ones.

From the examination of the later periods, and of the PCEEC corpus, it emerges that during the course of the ME period, preposing of elements is virtually limited to quantified and negated elements. Moreover, evidence from the native PCEEC texts shows that negated elements outnumber quantified ones in preverbal position, but the amount of preposed quantified and negated elements eventually drops at the same time during the Early Modern English period. This evidence confirms the scenario proposed in Ingham (2000, 2002, 2007), which sees the lingering preverbal spell out of negated elements as a consequence of the Negative Cycle in the history of English. This piece of evidence had gone unnoticed in Pintzuk & Taylor (2006), since they only look at the PPCME2 corpus and therefore conclude that negated and quantified preverbal objects present similar rates of preposing. The direct comparison between the M3 and M4 periods in both corpora therefore provides novel evidence for the scenario proposed by Ingham (2007).

For OE, Struik (2022) and Ingham (2007) argue that negated and quantified objects do not behave differently from referential objects, therefore Struik (2022, 80–81) proposes that they are attracted to Spec,vP and spelled out there to check either a uNeg or a uQuant feature in TP. However, Ingham (2007), building on Zeijlstra (2004), links the preverbal licensing of negated objects to the syntax of negation in the history of English and claims that the developments we observe are part of a larger negative cycle, and possibly independent of the OV/VO variation. This view is corroborated by the observation that by Late ME, OV is productive with negated elements only, whereas quantified preposed objects represent a residual order, and our data show that the latter are more frequent in the PPCME2 corpus, which consists of both native and translated texts. According to Ingham's (2007) proposal, Spec, NegP was occupied by the negative operator ne in OE and Early ME. In this period, negated objects need not be raised to Spec, NegP and Ingham (2007) writes that they are moved to the same position where positive objects are moved: in the account adopted here, Spec,vP (Struik 2022). Once the ne operator is lost, uNEG n-words emerge, and these are moved to Spec,NegP to express negation. Ingham moreover argues that the option of marking negation by overtly moving a uNeg n-word to Spec,NegP emerges in Late Middle English. Ingham (2007) in fact observes that in Early ME, negated elements always correlate with the negator ne, as example (25) shows, whereas they are in complementary distribution with the new negator not in Late ME, as examples (26a–b) show.

In light of these data, spell out of negated elements in Spec,NegP is needed in late ME to check uNeg features expressed by the negator ne, but once that is lost and the new post-finite negator not emerges, Spec, NegP can be either occupied by not, or by a raised uNeg n-word.

My data from the Cely collection, dating to the M4 period, corroborate Ingham's (2007) proposal, since the structures extracted from the corpus have a specific function: the negated object negates the whole clausal predicate. This function is intimately connected to Negative Cycles. Evidence for raising of uNeg n-words to Spec, NegP is not only given by negated objects, but also by negated subjects, as Ingham (2000, 2002) shows. Therefore, there is strong indication that raising of negated elements is not only different from positive objects, but also licensed under different conditions than quantified elements, as Figure 9 in Section 3 shows. The higher rates of negated elements in preverbal position in the M3 and M4 periods and the sudden drop from the E1 period onwards provide further quantitative evidence for the account proposed by Ingham (2007). The increase of preposed negated elements in late ME follows from the loss of the ne operator. Finally, Ingham (2007) argues that Neg Movement is lost when NegP is altogether lost in Early Modern English, hence the drop in frequency of preposed negated objects observed in Figure 9. The data extracted from the PCEEC corpus and illustrated in Section 3 provide quantitative evidence for this. Interestingly, a similar phenomenon is found in the history of Hungarian (É. Kiss 2015), where the development of the negative particle sem correlate with word order change, showing that the relation of word order and the Negative Cycle is not exclusively limited to the history of English.

Quantified elements show a higher rate of preposing in the M1 period, and they progressively drop in frequency by the late ME period. Moreover, in native texts of the M3 and M4 periods, they surface less frequently in preverbal position when compared to non-native texts in the PPCME2 corpus. The question as to whether quantified elements are raised to Spec,vP and spelled out there remains open. As long as there is evidence in the language for preverbal spell out of different types of objects, as in the M1 period, it can be hypothesized that quantified objects are likewise spelled out in their preverbal position. With the evidence for preverbal spell out progressively being lost, it can be hypothesized that scope is realized covertly and that the lower copy after checking is spelled out, (see Hinterhölzl 2015 for an implementation of Chomsky's 1993 proposal about spell out of either the higher or lower copy after movement applied Old Germanic languages). More research is needed to reveal whether the behaviour of quantitative elements is linked to specific developments in the language, the data examined in this paper shows that in native texts, quantified preverbal elements are a residual strategy, so it is reasonable to postulate that they follow a similar decline in their preverbal spell out as positive nominal and pronominal objects.

5 Conclusions

In this paper, I presented new evidence on negated and quantified objects by examining their distribution in the PPCME2 and PCEEC corpus, paying attention especially to the nature of the texts in which these elements are found. This factor has not received attention in previous investigations, but is relevant in the discussion, given the evidence presented in Struik (2022) for OE. It emerges that when these elements are not found in translated texts, they are tied to a particular language variety, that of Norfolk, in the PPCME2 corpus. The search on the PCEEC corpus allowed to gather more insights on these elements and shows that Norfolk is not the only area where quantified and negated elements are preposed in late ME. Moreover, the data extracted from the PCEEC corpus confirm and corroborate Ingham's (2002) findings on Late ME and Early Modern English, since it emerges that preverbal placement of objects is productive mostly with negated elements in non-literary correspondence texts.

The data are interpreted within the wider language change scenario from linear OV to VO order in the history of English. In light of recent findings, I adopted a Universal Base for the syntax of OE, and it was shown that preverbal licensing of positive objects is governed by IS and Weight interface conditions in OE, whereas in Early ME it is predominantly pronouns which are still found in preverbal position. In accordance with De Bastiani (2020, 2022), this is taken to indicate that by Early ME, it is the Weight Transparency condition which drives the postverbal spell out of non-pronominal objects, a condition which is eventually blurred out already in some varieties of Early ME.

The data extracted confirm observations presented in the literature on preposed negated and quantified objects. On the one hand, the higher rate of negated objects in native texts and their increase during the M3 and M4 periods corroborate Ingham's (2007) theoretical account. The distribution of negated and quantified objects in the M1 period shows on the other hand that these are not the only objects preposed in this time frame, for which we have evidence from native texts, and confirms Struik (2022) and van Kemenade (2022)'s observations that in OE and Early ME quantified and negated objects are no different to positive ones. Therefore, it is assumed that they are moved to Spec,vP to check quantification and negation features in OE and in Early ME. As far as quantified objects are concerned, the fact that they are less frequent in late ME native texts might indicate that their higher preverbal placement in the M3 and M4 periods in the PPCME2 corpus may be influenced by translations, but more work is needed in order to fully understand their behaviour.