Syntactic effects of contact in translations: evidence from object pronoun placement in Middle English

Abstract

Whereas object pronouns regularly occurred before the main verb in Old and early Middle English, such word orders were to a large extent lost in Middle English prose by the end of the thirteenth century. Nevertheless, some isolated later texts still show regular preverbal occurrences of object pronouns. Such word orders are most frequent with three texts that are translations of French sources. This article closely examines one of these texts, the Middle English prose Brut, and its source, and argues that contact influence is the most plausible explanation for its distinct behaviour with respect to object pronoun placement. It is also shown that the translator does not slavishly follow his source and that the contact effects are mainly of the statistical type in that word orders occurring very marginally in other texts appear with high frequencies in the Brut while such a contrast is not found for a word order that is unattested elsewhere. These observations are compatible with the equally exceptional but slightly different distribution of object pronouns in another translation from French, the Ayenbite of Inwyt. The findings of this article show that translation-induced contact and, possibly, contact in bilingual language use more generally can have important quantitative effects and that these have to be seriously considered in any syntactic analysis of historical texts based on a foreign source text.

1 Introduction

As often observed in the literature on historical syntax, translations must be handled with care when one tries to determine the syntactic properties of an earlier stage of a language, since distinctive features of a translation could be the result of influence of the source text rather than the sign of a fundamental property of the language examined. Although the problem of potential influence is regularly acknowledged in the literature, there is relatively little work that has examined the exact nature of this influence in detail. The most notable exceptions in the field of historical English are recent studies focusing on the interaction between Old English (OE) translations and their Latin source texts (Cichosz, Gaszewski & Pęzik 2016; Taylor 2008; Timofeeva 2010). These have all demonstrated a certain degree of influence in various areas of the syntax.

The following main types of syntactic influence can be distinguished. First, given a certain construction that occurs in the source language, for which there is no exact equivalent in the target language, a translator might be led to render the source text in a way that is not truly native in his or her own language. Secondly, when the target language shows variation with respect to a given phenomenon and the source language does not or does in a different way, the source language may have an impact on usage frequencies. This is what Taylor (2008) calls statistical translation effects. These can be, in Taylor's terms, either direct or indirect. Direct translation effects occur when the translator tries to match the target sentence to the source sentence. Thus, for example if the target language has two variants, A and B, and the source only has A, the direct translation effect would lead the writer to use A. The result of direct translation effects is that overall the author makes more use of A than would be the case for a non-translated text. As for indirect translation effects, they occur when the translator recasts the source text or adds new elements but does so by making use of a syntactic option that is prominent in the source language. Assuming again the two variants A and B in the target language and only A in the source language, the writer may be led to use A more frequently even in non-translated clauses because of some kind of syntactic priming effect, i.e. the phenomenon known from the psycholinguistic literature whereby the likelihood of using a certain option is higher when it has recently been used (see e.g. Mahowald et al. 2016). Priming effects have also been found to occur across languages in bilingual language use (Travis et al. 2017).

Whether one or several of these types of influence characterize a translated historical text or not can only be determined on the basis of a close comparative analysis of the translated text and its source as well as, if possible, non-translated texts from around the same period. However, the interest of comparing translations and their sources goes beyond the simple question of whether the syntax of a given text can be considered as representative of its period or whether it is altered in any way. When translating, translators activate their linguistic competence in both the source and the target language. The translation process therefore represents a specific case of language contact, the effects of which are comparable to other contexts in which languages come into contact (see e.g. Kranich 2014).

As pointed out above, the source language may influence the use of the target language, a phenomenon that has been labelled interference or ‘shining-through’ in translation studies. This kind of influence has the potential of contributing to linguistic change. As Kranich (2014: 96) puts it: ‘If the same type of shining-through phenomenon occurs repeatedly in translations, it might spread to monolingual text production, that is to non-translated texts produced by [target language] authors.’ Given these observations, a close analysis of translations and their source texts may provide insights into the nature of language contact and its potential diachronic effects.

The aim of this article is to make a further contribution to the study of the nature of influence phenomena in contexts of language contact through translation. In contrast to the studies cited earlier, the source language of the translations to be considered will not be Latin but two medieval varieties of French, continental Old French (OF) and Anglo-Norman (AN). My empirical focus will be the placement of object pronouns in Middle English and I will argue that OF and AN as source languages can have important effects on the word order options used by the translator. However, these effects are mainly of what Taylor (2008) calls the statistical type.

The article is structured as follows. In section 2, I provide an overview of the syntax of object pronouns in Old English (OE) and Middle English (ME), and I will show that, although preverbal object pronoun placement declines rapidly in the ME period, there are some texts that still have a high frequency of OV order with pronominal objects, and that the most important exceptions to the general declining pattern are translations of OF and AN texts. Section 3 focuses on one text that has exceptional properties, the Middle English prose Brut, and closely examines its relation to a corresponding AN text to determine how the source language may have influenced the ME text. Some further texts with unusually high frequencies of preverbal object pronouns are then discussed in section 4. Section 5 summarizes the main findings and explores some general consequences of these findings with respect to the effects of language contact through translation.

2 The syntax of object pronouns in early English

OE has properties that are reminiscent of an OV language (e.g. van Kemenade 1987). These head-final properties are lost in the ME period. The frequency of head-final structure at the clausal level as measured on the basis of the word order variation auxiliary–verb/verb–auxiliary is already very low in the thirteenth century (Kroch & Taylor 2000a). As for head-final structure at the VP level, Pintzuk & Taylor (2006) conclude on the basis of the distribution of non-pronominal objects (henceforth full DP objects²) with respect to non-finite main verbs that it is lost to a large extent in the first half of the fourteenth century. After that, OV order with full DP objects is restricted to quantified and negative objects. Such cases are most plausibly analysed as involving leftward movement of the object rather than underlying head-final structure. These observations, when taken together, suggest that head-final structure does not play any role in ME syntax from the middle of the fourteenth century onwards.

The placement of object pronouns has not featured prominently in these discussions of headedness as it is well known that pronominal objects productively undergo leftward movement in early English and their distribution does therefore not allow one to draw any firm conclusions with respect to underlying structure. However, given that preverbal placement of object pronouns is ungrammatical in Present-day English (PDE), it is clear that this word order option must have declined during the ME period. In this section, I will briefly trace the diachronic development of object pronoun placement in OE and ME, first in clauses with finite main verbs and then in clauses with finite auxiliaries.³

2.1. Object pronouns and finite main verbs

In OE, object pronouns can either precede or follow finite main verbs, but preverbal placement is the majority option. ‘Subject – Object pronoun – Finite main verb’ (SOproV) order is illustrated in (1) (object pronoun in italics, finite main verb in bold print).

1. (1)

   

In the ME period, SOproV order declines. This is shown in table 1, which is based on all clauses containing an overt subject, a finite main verb and an (unmodified) object pronoun in the York–Toronto–Helsinki Parsed Corpus of Old English Prose (YCOE; Taylor, Warner, Pintzuk & Beths 2003) and the Penn–Helsinki Parsed Corpus of Middle English (PPCME2; Kroch & Taylor 2000b). As is common practice, I have divided the ME data into the four major periods, m1 to m4, as distinguished in the PPCME2. Only texts that can be clearly assigned to one of these periods are included.

Table 1. The distribution of object pronouns and finite main verbs in Old and Middle English

Table 1 suggests that preverbal placement of pronominal objects is very common until the middle of the fourteenth century and then drops abruptly at the end of that century. Nevertheless, the frequency of SOproV order remains non-negligible in period m3, and we see a further significant decline from period m3 to period m4 in the fifteenth century (chi-square = 28.98; p < 0.001).

However, once we consider the contribution of individual texts to this overall picture, our account of the decline of SOproV has to be modified. Let us start by taking a closer look at period m2. Prose material for this period is very scarce, and the PCCME includes only three texts in this period: the Kentish Sermons (c.1275), the Ayenbite of Inwyt (1340) and the Earliest Complete English Prose Psalter (c.1350). With respect to SOproV, table 1 hides a substantial imbalance in the way these texts use the different orders. Whereas the author of the Prose Psalter almost never uses SOproV (1.6 per cent; n = 309), the other two heavily favour this order. In the Ayenbite, the frequency of SOproV is 89.3 per cent (n = 300), and in the Kentish Sermons 67.2 per cent (n = 21).

A similar observation can be made for period m3. Among the sixteen texts included in this period, only half contain examples with SOproV orders. But among those eight, only one has frequent occurrences of SOproV, and that is The Brut or the Chronicles of England (c.1400). No fewer than 87 of the 111 cases listed in table 1 come from this text. With 178 examples with SVOpro order, the frequency of SOproV reaches 32.8 per cent in the prose Brut.

Finally, in period m4, the SOproV frequencies are very low. Nevertheless, a considerable imbalance remains among texts. Two of them have frequencies of SOproV order that are well above those found in other texts from the same period: The Siege of Jerusalem (c.1500; SOproV 34.8 per cent, n = 66) and The Commonplace Book of Robert Reynes (1470–1500; SOproV 45.5 per cent, n = 11). Once we remove these texts, 13 instances of SOproV order remain.

Table 2 provides revised figures for SOproV order, excluding outliers that have particularly high frequencies of this word order compared to other texts from the same period.⁴

Table 2. The distribution of object pronouns and finite main verbs in Old and Middle English – outliers separated

The overall picture emerging from table 2 is that SOproV is common until the middle of the thirteenth century, but it is then largely lost. No major developments occur after the thirteenth century. The loss of SOproV therefore seems to occur considerably earlier and more rapidly than suggested by the data in table 1.

2.2 Object pronouns and finite auxiliaries

Similar observations can be made for all clauses containing an overt subject followed by a finite auxiliary (Aux) followed by a non-finite main verb (V). In OE, the object pronoun generally precedes either the auxiliary or the main verb in such clauses. This is illustrated in (2a) (SOproAuxV) and (2b) (SAuxOproV).

1. (2)

   

During the ME period both of these word orders are lost. Table 3 presents this decline in the same way as table 2 in that clear outliers are listed separately.

Table 3. The distribution of object pronouns, finite auxiliaries and non-finite main verbs in Old and Middle English – outliers separated

If we first consider the different periods without the outliers, we can observe that pre-auxiliary placement of object pronouns (SOproAuxV) is very common in OE and early ME, but it is then entirely lost after 1250. But we note once again that the next data point (m2) corresponds to a single text from the mid fourteenth century and that the development for a century after the period m1 therefore remains uncertain. As for the order SAuxOproV, there is a substantial decline in its frequency in the transition from OE to ME already. After period m1, this word order becomes marginal, but residues can be found until the end of the ME period.

Once again, individual texts behave differently, and they are largely the same ones as those identified in table 2. In period m2, the Kentish Sermons and the Ayenbite show a pattern that is highly unusual even compared to OE and the ME period m1 in that SOproAuxV is by far the most frequent word order. But as in OE, there are only marginal findings of the innovative word order SAuxVOpro in these texts. In period m3, the most distinctive text is again the prose Brut, with the order SAuxOproV occurring as the majority option. Finally, in period m4, The Siege of Jerusalem and The Commonplace Book of Robert Reynes remain distinctive. Whereas the former text has a frequency of SAuxOproV order of 43.5 per cent (n = 23), the latter has a rate of 20 per cent (n = 20). This is at a time when other texts generally no longer have any orders of this type.

2.3. Variation in object pronoun placement

The data in tables 2 and 3 raise the question as to why certain texts have substantially higher frequencies of preverbal object pronoun placement than other texts from the same periods. For the three texts with the highest rates of preverbal objects a closer look at the historical background may provide a plausible answer. The Kentish Sermons, the Ayenbite as well as the prose Brut have the striking common property of being relatively close translations of source texts written in French, more precisely OF for the former two and AN for the Brut. This is of importance because object pronouns (clitics) generally precede the main verb in both OF and AN.⁵ Many instances of OV found in these three texts have a corresponding word order in the French versions of these texts. Influence of the source text is therefore a very likely explanation for the distinctive behaviour of these ME texts. As a consequence, residual preverbal object pronoun placement in ME is a phenomenon that may allow us to gain some insight into the nature of translation-induced effects in medieval texts.

3 The syntax of object pronouns in the ME prose Brut and its AN source

Assuming that the unusual properties of the ME prose Brut identified in section 2 can be related to influence of the AN source text, I will examine the nature of this influence in more detail in this section and provide further evidence in support of the influence hypothesis. But before turning to these issues, some observations concerning the textual basis for a comparison of the English and the AN version of the Brut are necessary.

3.1 The ME prose Brut and its source

The prose Brut is a chronicle that gives a comprehensive account of the history of England and survives in over 240 manuscripts in the three major literary languages of medieval England (Matheson 1998: 1). It is generally agreed that the earliest versions of the prose Brut were written in AN. On the basis of the historical events reported, different stages in the development of the AN prose Brut can be distinguished (Matheson 1998: 4): (a) Stage I, events up to 1272 (the original form of the text, sometimes referred to as the Common Text); (b) Stage II, events up to 1307 (first continuation); (c) Stage III, events up to 1333 (second continuation, either in a short version or in a long version). The original texts representing the different stages are likely to have been written shortly after the events they recorded. In the late fourteenth century, the long version of Stage III was translated into English. This initial translation then gave rise to most texts of the family of the ME Brut, with later versions regularly bringing the historical records up to date independently of an AN source.

In order to examine what influence AN may have had on the syntax of a translated ME text, I will compare an ME version of the Brut to an AN version. For ME, I will focus on the sections of the Brut contained in the PPCME2 (Kroch & Taylor 2000b). These are taken from the first part of Brie's (1906) edition, which represents the earliest English copy of the Brut from around 1400. As Brie (1906: ix) observes, this ‘first part of our Chronicle is a mere translation of the French Brut d'Engleterre’ (see also Matheson 1998: 79). The content suggests that the original AN source for this translation was the long version of Stage III (see Matheson 1998: 6; Pagan 2011: 4). Ideally, one would therefore use that version for a comparative study of ME and AN. Unfortunately, to my knowledge, edited versions of the AN texts are available only for Stage I of the AN version (Marvin 2006) and the short version of Stage III (Pagan 2011). I will therefore base my discussion on the latter AN text. This should not be too problematic, however. The main differences between the short and the long versions of Stage III concern the events from 1307 onwards. Brie (1905: 26) therefore claims that there is no doubt that the writer of the long version based it on a common text of the Brut up to the death of Edward I (i.e. 1307). There are only very minor factual differences between the long and the short versions of Stage III for the period before 1307. No indications can be found in the literature as to any significant linguistic differences between the different versions.

Given these observations, I will assume that Pagan's (2011) edition of the AN Brut is representative for the AN text that provided the basis for the ME translation, at least for the events up to 1307. I will therefore use this edition up to 1307 to explore the possible influence of AN on the ME used in the translation. It is important to stress, however, that this is unlikely to be the perfect match of translation and source text. One reason for this is that ‘[n]one of the English manuscripts preserves the original translation’ because ‘they represent copies written some time after the presumed date of composition of any portion of text’ (Matheson 1998: 84). Furthermore, with respect to AN, it is impossible to establish exactly what AN source the ME translator had access to. But since large parts of Brie's ME text read like a very close translation of Pagan's AN text, it seems to me that these two texts provide a sufficiently solid basis for a close comparative analysis.

3.2 Object pronoun placement in the ME prose Brut and the AN version

In this section, I will compare the placement of object pronouns in the ME and the AN Brut. As pointed out earlier, the hypothesis that increased preverbal object pronoun placement may be due to AN influence is based on the fact that AN object clitics precede the main verb. In clauses with finite main verbs, the presence of SOproV word order in the AN source text may therefore have led the translator to use the same order in English. This is illustrated in (3).⁶

1. (3)

   

In Taylor's (2008) terminology, the matching word orders in (3) could be considered as being due to a direct translation effect of the statistical type. The ME word order corresponds to the AN one, but it is not an innovation in ME. Instead, the contact situation leads to a substantial quantitative increase in the use of a declining word order that is still occasionally used in other texts.

However, it is not the case that the author of the ME Brut slavishly follows the AN text. As shown in table 2, postverbal placement of the object pronoun is the majority option with finite verbs in the ME Brut. This means that the author often uses postverbal object pronouns when the AN text has a preverbal clitic. This is shown in (4).

1. (4)

   

The second type of evidence suggesting that the ME author does not strictly imitate the AN syntax is presented in (5). In these examples, the ME translations (5a, b) are relatively free, and they introduce an object pronoun where there is none in the AN text (5a′,b′). Despite the absence of an AN model, the object pronoun precedes the finite main verb in the ME version.

1. (5)

   

In Taylor's (2008) terminology, (5) illustrates indirect translation effects. The systematic occurrence of preverbal object pronouns elsewhere in the AN source and the frequent use in the translation may have a priming effect that leads the translator to use this word order option even when it does not occur in the source. For the reasons discussed in section 3.1, this type of evidence has to be treated with some caution, however. Although no object pronoun is found in the corresponding passages in Pagan's (2011) edition shown in (5), we cannot be absolutely certain that this was also the case for the AN text on which the original translation was based. Thus, the data in (5) are suggestive (albeit not entirely conclusive) of the translator's independence with the use of SOproV order.⁷

In order to shed further light on the nature of the interaction between AN and English, I will now consider the placement of object pronouns in clauses containing a finite auxiliary and a non-finite main verb. In clauses with an auxiliary, a pronominal object could occur in three positions: (i) in a position before the finite auxiliary (SOproAuxV); (ii) in a position between the auxiliary and the main verb (SAuxOproV); (iii) in a position following the main verb (SAuxVOpro). AN generally uses option (i), even in contexts with modal auxiliaries where present-day French would now have option (ii) (clitic climbing; see example (6) below).⁸ Hence, if AN influence played a role in the choice of the placement of object pronouns in ME, we would expect a high frequency of option (i) in the prose Brut. As table 3 shows, this expectation is not borne out. The majority pattern is SAuxOproV, which is nevertheless highly unusual compared to other texts from the same period (50.7 per cent in the Brut vs 3.4 per cent elsewhere). However, here, it is generally not the case that this predominant order reflects a word order found in the AN source text. Instead, we can distinguish three different contexts in which ME has SAuxOproV order. Each of these options corresponds to about a third of all the cases of SAuxOproV in the Brut. First, this order may occur when AN has the order SOproAuxV.

1. (6)

   

Secondly, the ME text sometimes uses SAuxOproV order when, as in (7), the AN text lacks an auxiliary and has SOproV instead.

1. (7)

   

2. (7′)

   

Finally, SAuxOproV can also be found when there is no match in the AN text, either because the ME text contains elements that are entirely absent in the AN text or because the translation is not literal. Illustrations of the latter type are given in (8).

1. (8)

   

Although there is no exact match in word order in the sentences shown above, AN influence cannot be entirely excluded, at least for cases such as (6) and (7). What may have been the salient feature for the translator in these examples is simply the fact that the object pronoun occurs to the left of the main verb in AN. This distributional property is then reproduced in the ME version.

The main question these observations raise is why the most frequent word order in the Brut is one that is generally not found in the AN source text (i.e. SAuxOproV) and why the AN word order is virtually absent in the Brut. A plausible explanation for this state of affairs is that by about 1400, when the Brut was written, the order SOproAuxV had become an ungrammatical option in English. According to table 3, the last example of SO(pro)AuxV in a non-French-based text indeed goes back to the period 1150–1250, i.e. to a text written around 150 years before the Brut. The very low frequency of SOproAuxV order in the Brut may therefore be related to the reluctance of the translator to use an entirely ungrammatical option. Instead, he prefers the order SAuxOproV, which can still be found in some other texts from the same period, albeit at a very low rate (3.4 per cent).⁹ Furthermore, the position between a finite auxiliary and the main verb can also be occupied by quantified and negative objects in late ME. Pintzuk & Taylor (2006: 259) show that in the fifteenth century around 6 per cent of all quantified objects and around 20 per cent of all negative objects occur preverbally. Thus, the position between the auxiliary and the main verb is a position that is still available, at least marginally, for objects in late ME. This word order option then reflects that of the AN at least with respect to the placement before the main verb.

The frequent use of SOproV in the Brut is compatible with the scenario sketched above. As shown in table 2, this word order has a marginal status in the fifteenth century, but it is not entirely absent from other texts. Thus, as with SAuxOproV, the translation effect would be a quantitative reinforcement of a word order option that is nearly extinct but not fully ruled out by other authors.

I will finish the discussion of clauses with auxiliaries by briefly considering the status of the word order SAuxVOpro in the ME Brut. A comparison with the AN source leads to very similar results to those of the order SAuxOproV. SAuxVOpro can occur when AN has SOproV with a finite main verb, when it has SOproAuxV, or when no corresponding object pronoun can be found for AN. Once again, the ME SAuxVOpro examples are spread almost evenly across the three scenarios. An illustration of each option is given in (9).

1. (9)

   

The examples in (9a/a′) and (9b/b′) confirm the observation made in section 3.1 that the author of the ME Brut does not mechanically transfer the AN syntax into ME.¹⁰

To conclude this section, I add two observations that lend support to the influence hypothesis of object placement in the Brut. First, as pointed out in section 3.1, the AN text that was used as the basis of the ME translations (long version of Stage III) covers events up to 1333. Thus, if French influence were essential for the high frequency of preverbal pronominal objects, we would expect such word orders not to occur in the descriptions of events after 1333 as these are not based on a French source text. This expectation is indeed borne out. In samples that I have examined for different periods after 1333 in Brie's (1906) edition, preverbal pronominal objects are almost entirely absent. For the period 1333–77, I took the first 50 clauses containing a pronominal object, and found one single instance with preverbal placement. For the periods 1377–1419, 1418–30 (excluding John Page's poem on the siege of Rouen) and 1430–46, I examined the first 30 clauses containing a pronominal object, and preverbal placement did not occur at all. The contrast between the pre-1333 part of the ME prose Brut and the post-1333 part is thus very striking. However, whether there is an AN source or not is not the only variable that could be related to this contrast. Original authorship also varies with the continuations, and that variation may be of importance for the use of preverbal object pronouns. Thus, the pre-/post-1333 contrast is in line with the influence hypothesis for the pre-1333 text but is not entirely conclusive in this respect.

A second observation might provide stronger support for the influence hypothesis. If influence did not play a role, what alternatives would there be to account for the distribution of object pronouns in the ME prose Brut? One possibility would be that it is due to a stylistic choice by the author to use archaic word orders.¹¹ But if that were the case, it would be surprising for that stylistic choice to be restricted to object pronouns, and we would expect other archaic features to be present in the Brut as well, for example regular preverbal occurrences of full DP objects. As is well known, full DP objects frequently precede the main verb in OE. As for the development in the ME period, a distinction has to be made between clauses with a finite main verb and clauses with an auxiliary and a non-finite main verb. With finite main verbs, SO_DPV order quickly declines after the OE period and in the ME period m3 most texts no longer contain this word order. In the Brut, SO_DPV is a very marginal option (0.9 per cent; n = 701), so there is no indication here that the translator makes increased use of an archaic word order option. The same conclusion can be reached with clauses that contain a finite auxiliary. From over 50 per cent in OE (see Pintzuk & Taylor 2006: 259), the rate of SAuxO_DPV order has fallen to 2.5 per cent in the ME period m3 (n = 2,230). Separating the data for the Brut and those for the other texts from period m3, frequencies of 5.6 per cent for the Brut (n = 305) and 2.0 per cent for the other texts (n = 1,925) are obtained. The increase in the Brut is thus minimal compared to what was observed in table 3 for object pronouns. That there is a fifteen-fold increase in the frequency of SAuxOproV order in the Brut compared to other m3 texts but less than a threefold increase with full DP objects would be unexpected if the translator of the Brut simply had a stylistic penchant for archaic word order options.

In terms of the French contact influence hypothesis, the contrast between pronouns and full DPs with respect to preverbal placement can be accounted for straightforwardly. Whereas object clitics occur preverbally in OF and AN, OV order is in decline with full DP objects (see e.g. Zaring 2010 for OF). By the beginning of the thirteenth century, VO is ‘fairly well fixed’ (Zaring 2010: 5) with finite main verbs in OF, and the AN Brut is in line with this observation. In clauses with a finite auxiliary and a non-finite main verb, however, the decline of OV is slower. Zaring (2010: 8) shows that SAuxO_DPV still occurs with a certain frequency in thirteenth-century OF. The same is true for thirteenth- and fourteenth-century AN (Ingham 2017). In a sample of 1,500 lines of the AN Brut,¹² OV order is found in 29.6 per cent (n = 54) of all clauses containing an auxiliary (modal or auxiliary have) and a non-finite main verb. The regular occurrence of OV in the source text may have led the ME translator to use the corresponding word order more frequently than other authors from period m3. But with a frequency of a bit over a fourth in AN, SAuxO_DPV is far from being as salient as the systematic preverbal placement of object clitics. This could be argued to be the reason why the degree of influence is considerably smaller with full DPs.

In summary, although it is difficult to conclusively prove influence in a medieval situation of language contact through translation, the various observations made in this section make the hypothesis very plausible that the ME prose Brut has syntactic properties that are the result of influence of the AN source text.

4 High rates of preverbal object placement in other texts

In this section, I will briefly consider further texts that are exceptional with respect to object pronoun placement. Section 4.1 deals with PPCME2 period m2 and section 4.2 with period m4.

4.1 Period m2: The Ayenbite and the Kentish Sermons

As shown in tables 2 and 3, period m2 contains two texts that have unusually high frequencies of preverbal object pronoun placement, the Ayenbite and the Kentish Sermons. Just like the Brut, these two texts are translations from French, in this case OF. In the brief discussion in this section, I will focus on the Ayenbite because the amount of data from the Kentish Sermons is very small (34 clauses with a pronominal object as opposed to 418 for the Ayenbite). Given the context of the discussion here, I simply cite the following observation made by the editor of the Kentish Sermons: ‘The translator gives a very literal rendering of his original; it influences his idiom, order and choice of words, even to the borrowing of an occasional inflection’ (Hall 1963: 669). Hence, OF influence is likely to play a role with the distribution of object pronouns in the Kentish Sermons.

According to the information the author himself provides, the Ayenbite was written by the monk Dan Michel in 1340. It is a close translation of the OF Somme le roi, a religious treatise written between 1279 and 1280 (see Brayer & Leurquin-Labie 2008). As shown in section 2, the Ayenbite has very high frequencies of word orders in which the pronoun precedes the main verb. Compared to the Brut, there are two main differences. First, preverbal placement of object pronouns is much more frequent in the Ayenbite, to the point that postverbal placement is a very marginal option in this text. Whereas postverbal placement in the Brut is found in 67.2 per cent of the clauses with a finite main verb and in 45.7 per cent of the clauses with a finite auxiliary, the corresponding frequencies for the Ayenbite are at 10.7 and 2.5 per cent. Secondly, in clauses with a finite auxiliary, the predominant preverbal position is before the auxiliary in the Ayenbite whereas it is between the auxiliary and the main verb in the Brut.

At first sight, these differences suggest that influence of the OF source text has even bigger effects in the Ayenbite than in the Brut. Given the closeness of the translation, it is certainly very likely that influence crucially contributes to the Ayenbite’s unexpected object pronoun syntax. There are indeed numerous clauses with a preverbal object pronoun in the Ayenbite that have a direct match in Brayer & Leurgquin-Labie's (2008) edition of Somme le roi. In addition, there also seem to be indirect translation effects as preverbal object pronouns can be found in cases where there is no corresponding OF source or where the OF source does not have a pronominal object (see Hess 2016 for a comparison of the two texts).

However, there is an additional factor that should be taken into account when trying to explain the distributional properties of object pronouns in the Ayenbite. While the ME Brut was written in the second half of the fourteenth century, the Ayenbite is a somewhat earlier text (1340). Furthermore, as the author seems to have been around 70 years old when he wrote the Ayenbite (Gradon 1979: 12), his language can be argued to reflect the linguistic situation of the late thirteenth century. Thus, we get closer to a period (i.e. m1) when the orders SOproV and SOproAuxV still regularly occur across texts (43.2 and 36.2 per cent respectively for period m1). Compared to the author of the Brut, Dan Michel is therefore likely to have been exposed to such word orders more frequently, even independently of the presence of a source text that features preverbal occurrences of object pronouns.

However, even if we take this factor into account, the patterns observed in the Ayenbite remain surprising as the frequencies of SOproV and SOproAuxV go well beyond those found in texts from period m1. As a matter of fact, these frequencies are even higher than those found in OE. Thus, it is not simply the case that the Ayenbite maintains an archaic pattern, it even goes beyond what is attested for a period of almost 500 years before the composition of this text. This is in striking contrast with the fact that in other aspects of word order the Ayenbite is by no means similar to OE. For example, SO_DPV order with any type of non-pronominal object DP is less frequent in the Ayenbite than in most texts from period m1 (3.6 per cent (n = 578) vs an average of 10.0 per cent (n = 1,617) for m1). The same is true for preverbal placement of object DPs in clauses with finite auxiliaries and non-finite main verbs. The two orders SO_DPAuxV and SAuxO_DPV occur with a combined frequency of 17.9 per cent (n = 185) in the Ayenbite while the corresponding average frequency for period m1 is 31.8 per cent (n = 676). Finally, the Ayenbite is perfectly modern when it comes to the distribution of auxiliaries and main verbs. A single example of the common OE word order SVAux is found against 811 clauses with SAuxV order. In period m1, SVAux still occurs somewhat more frequently at an average rate of 3.9 per cent (n = 2,687).

In conclusion, the fact that the Ayenbite represents an earlier ME period may very well contribute to an increased rate of preverbal placement of object pronouns. However, the full extent of the difference between the Ayenbite and the Brut is unlikely to be due solely to these factors. Instead, influence of the OF source text remains a highly plausible factor as well. Without direct and indirect statistical translation effects, it would be very surprising to find rates of SOproV and SOproAuxV that go significantly beyond those found in OE. Furthermore, assuming that the influence hypothesis is correct and that the Ayenbite is diachronically less advanced than the Brut may allow us to account for a difference between these two texts. In clauses with auxiliaries in the Brut, the main effect of AN influence is an increase in SAuxOproV order as this is the only option with a preverbal object pronoun attested in ME when the Brut was written. For the author of the Ayenbite the situation is different and orders of the type SOproAuxV may still have been (at least marginally) in use as well. Since this word order corresponds exactly to what is found in OF, the statistical translation effect is strongest in this context, and the rate of SOproAuxV is substantially higher than in the Brut.

4.2 Period m4: The Siege of Jerusalem and The Commonplace Book of Robert Reynes

Whereas high frequencies of preverbal object pronoun placement in periods m2 and m3 can be related to language contact through translation, the same account cannot hold for the outliers in period m4. Neither The Siege of Jerusalem nor The Commonplace Book of Robert Reynes are translations.¹³

As for The Commonplace Book, I have no plausible explanation for its high rate of preverbal object pronoun placement at this point. A better understanding of this document would require a close analysis of the sources that the author uses. Given the properties of commonplace books, it may very well be that parts of the text are taken from sources that do not reflect contemporary usage and that it is this that makes the text look archaic in its use of preverbal object pronouns.

With respect to The Siege, I propose that its properties might be related to an effect that is very similar to a statistical translation effect. Kurvinen (1969: 27) observes that The Siege must be based on the late fourteenth-century ME poem Titus and Vespasian. What might be possible then is that The Siege does not only follow its model with respect to content but to some extent also with respect to form.

I examined the distribution of pronominal objects in a sample of 1,500 lines of the poem Titus and Vespasian (Herbert 1905).¹⁴ For the sake of comparison, data were also collected from the last five poems in The Parsed Corpus of Middle English Poetry (PCMEP; Zimmermann 2015), which correspond roughly to the period 1350–1400. As shown in tables 2 and 3, preverbal object placement is virtually absent in prose texts from this period. The results for poetry are shown in table 4.

Table 4. The distribution of object pronouns in Titus and Vespasian and five PCMEP poems from the second half of the fourteenth century

The data in table 4 show that preverbal object pronoun placement is strongly represented in late ME poetry, with frequencies reaching over 30 per cent with finite main verbs and over 50 per cent with finite auxiliaries in both Titus and Vespasian and the PCMEP data.¹⁵

Given these observations, a plausible hypothesis for the distinctive distributional properties of object pronouns in The Siege would be that its author was influenced by word orders found in the verse text in a way that is similar to how a translator is influenced by a source text written in another language. Thus, what we seem to find in The Siege are not effects of language contact but effects of register contact.

5 Conclusion

The empirical focus of this article has been the development of object pronoun placement in Middle English. I have shown that preverbal and pre-auxiliary placement of object pronouns is largely lost after the thirteenth century. However, there are five texts from the fourteenth and fifteenth centuries that have unexpectedly high frequencies of preverbal object pronoun placement. The three earliest texts (the Kentish Sermons, the Ayenbite and the Brut) are all translations from OF or AN source texts. Given the absence of any convincing alternative explanations for this discrepancy and the fact that OF and AN have regular preverbal placement of object pronouns, I have proposed that the unusual word order properties found in these three texts are translation-induced effects. If this hypothesis is correct, a certain number of conclusions with respect to the effects of language contact through translation can be drawn.

First, let us start by returning to Kranich's (2014: 96) observation, mentioned in section 1, that certain translation-induced phenomena could spread to monolingual text production if they occur regularly in translations and that therefore language contact through translation could have the potential of leading to language change. One aspect of the findings reported here indeed confirms this potential, namely the quantitative impact contact seems to have on the word order patterns examined. For example, compared to other texts from the same period, the increase in preverbal placement in the Brut reaches over 30 per cent with finite main verbs and nearly 50 per cent with SAuxOproV word orders. Given such strong effects, regular occurrence of texts of this type could indeed have the potential of affecting the language (or a specific register) more generally. This did not happen with the phenomena examined here, presumably because translations of French texts were not common enough to counteract the decline of preverbal object pronoun placement.

Secondly, we have seen that the influence effects are of a purely statistical type. Word orders with preverbal object pronouns that are sporadically found in ME texts up to 1500 show considerable increases in their frequencies of use in the translations. Particularly interesting in this respect is the status of SOproAuxV. This order occurs with an increased frequency in the earlier text examined (the Ayenbite) but hardly at all in the later text (the Brut). I argued that this contrast may be due to a difference in the status of pre-auxiliary placement of pronouns at the times when the two translators wrote, with the translator of the Ayenbite possibly still being exposed to SOproAuxV and the translator of the Brut writing when SOproAuxV was no longer in use. The reluctance of the author of the Brut to use what is likely to have been an ungrammatical option and what would therefore have been a syntactic innovation contrasts with his readiness to make increased use of an existing but marginal option. A possible explanation for this contrast can be based on a proposal made by Ingham (2016) in the context of an entirely different language contact scenario (effects of ME L1 on AN L2). Ingham (2016) makes a distinction between properties of ‘narrow syntax’ on the one hand, i.e. aspects of the syntax that concern formal grammatical features such as those involved in subject–verb agreement and the linearization of constituents, and syntactic properties that interface with semantics and possibly phonology on the other, and he proposes that the former are generally not affected by language contact whereas the latter can be. The status of the word order SOproAuxV concerns narrow syntax (linearization). The reluctance of the translator of the Brut to use this word order would therefore be in line with the hypothesis that properties of narrow syntax are generally not subject to contact-induced change. Instead, all that is possible with narrow syntax is for contact to have an influence on the frequency of use of already existing options such as SAuxOproV (direct or indirect statistical translation effects in Taylor's 2008 terms). If these observations are on the right track, they suggest that findings based on contact in translation are of relevance for issues related to bilingual language use in general.

A further main finding is that even when influence of the source language can be detected, the ME authors are far from slavishly following AN or OF syntax. Instead, they regularly alternate between adopting the source word order and not adopting it. Furthermore, they also use the French-style word orders in contexts where no matching construction seems to be available in the source.

Finally, I have shown that for a very late text with unusually frequent preverbal object pronoun placement a plausible explanation is not language contact but register contact. The Siege of Jerusalem is based on an earlier ME poem, and poetry features much higher rates of preverbal object pronoun placement throughout ME. This property then seems to have been transferred to prose by the author of The Siege.

The findings reported here also have methodological implications for the study of historical syntax. Our data show that, at least as far as the properties of narrow syntax are concerned, the use of translations in diachronic syntactic analyses is not necessarily very problematic with respect to determining what we can consider as grammatical or ungrammatical at a given moment in time. At least the translator of the ME Brut is clearly reluctant to use a word order that reflects the syntax of the source language but seems to be ungrammatical in the target language. Where the translation context may be of great importance, however, is in the quantitative domain. As the data in tables 1 to 3 have shown, when the diachronic trajectory of a given syntactic phenomenon is examined, the inclusion of one or several translated texts can substantially alter the picture and lead to potentially important differences with respect to the dating of a change. Since not all translated texts show the same translation-induced effects (see fn. 13), this finding should not indicate that translations must be entirely excluded from quantitative studies in historical syntax. However, it shows that translation as a potential source of influence must be seriously considered when interpreting quantitative evidence and that close attention must be paid to the contribution translated texts make to aggregate data for a particular historical period.