Topic and focus in Old French V1 and V2 structures

Abstract

The hypothesis that Old French was not a verb-second (V2) language, but rather a Topic-initial language, is evaluated in a corpus of verb initial (V1) and V2 matrix clauses extracted from a corpus of 12^th- and 13^th-century texts. It is shown that the initial constituent of V2 clauses is not always a Topic; it may be part of the informational Focus, or it could be an element that is neither Topic nor Focus. In addition, in V1 and V2 sentences with subject inversion, the postverbal subject may be an informational Topic, contrary to the hypothesis that subjects must move to the preverbal position to avoid being interpreted as part of the informational Focus. Therefore, from an Information-Structure point of view, Old French is similar to a standard V2 language like German. However, certain differences between 12^th- and 13^th-century texts could suggest that the use of the left periphery evolved during the period considered.

Résumé

L'hypothèse que l'ancien français n'est pas une langue de type verbe-second (V2), mais plutôt une langue à topique initial est confrontée aux données d'un corpus de textes des 12^e et 13^e siècles, dont on a extrait les propositions matrices ayant le verbe en première (V1) ou en deuxième position (V2). Il est montré que le constituant initial des phrases V2 n'est pas nécessairement un topique ; ce peut être un élément faisant partie du focus de la phrase ou un élément qui n'est ni topique ni focus. De plus, dans les propositions avec inversion du sujet, le sujet peut être le topique discursif, contrairement à l'hypothèse selon laquelle un sujet postverbal est nécessairement interprété comme faisant partie du focus. Du point de vue de la structure informationnelle, l'ancien français n'est donc pas une langue à topique initial, mais est semblable à une langue V2 comme l'allemand. Certaines différences entre les données du 12^e et du 13^e siècles pourraient toutefois suggérer une évolution dans l'utilisation de la périphérie gauche de la phrase au cours de la période considérée.

1. Introduction

A growing body of work explores the role of information structure on word order variation and change.¹ Information structure refers to the way languages structure discourse information like topic and focus. The cartographic approach to syntactic structures (e.g., Rizzi 1997, 2004; Cinque 2002, 2006; Cinque and Rizzi 2008), despite its shortcomings (van Craenenbroeck 2009), has shown that specific positions in the clause appear to be dedicated to topic and focus elements (Benincà 1999, 2006; Belletti 2004; Benincà and Poletto 2004; Benincà and Munaro 2010), and a number of authors argue that changes in word order reflect changes in the way a language expresses these notions (e.g., Benincà 2006; Laenzlinger 2006; Hinterhölzl 2009; Hinterhölzl and Petrova 2009, 2010; Kroch and Santorini 2009; and the various papers in Batllori and Hernanz 2011).

The present article was spurred by Rinke and Meisel's (2009) claim that Old French is not a V2 language, and more specifically, that it differs from German in the structuring of information structure. These authors develop an analysis in which Old French, contrary to German, is a topic-initial language. We undertook the task of verifying their claims by studying the informational role of preverbal elements and postverbal subjects in Old French V1 and V2 structures. Our research refutes Rinke and Meisel's analysis. We show that the distribution of Old French preverbal elements is similar to that of typical V2 languages like German. However, as discussed in section 7, we observed an evolution in the statistical distribution of preverbal constituents which suggests that the use of the left periphery changed during the period considered.

2. The Corpus

The data are extracted from 19 parsed Old French texts of the MCVF corpus (Martineau et al. 2010) and the Penn supplement to this corpus (Kroch and Santorini 2012) (see Appendix). In order to be able to create tables and figures, a date of composition was assigned to each text, ensuring that no two texts had the same date. These dates should, of course, be taken only as an approximation of the year the text was written.

We used Corpus Search (Randall 2005) to automatically extract all positive declarative matrix clauses with a full subject (i.e., excluding null and clitic subjects) and to code the data. We report here our analysis of V1 and V2 clauses only. V ≥ 3 clauses were kept for later investigation.

One characteristic of the corpus used is that the 12^th-century texts are all in verse except for those of Li Quatre Livre des Reis, 1170 (henceforth QLR). This is due to the lack of 12^th-century prose texts, all surviving texts of some length being in verse. Conversely, all of the 13^th-century texts in the corpus are in prose, with the unfortunate consequence of a near conflation of text genre and time variables.² In the figures to be presented, both the date and the genre are indicated.

3. Operational Definitions Relative to Information Structure

The following operational definitions of topic and focus are adopted:

Information topic: The information topic, or I-topic, also called aboutness topic (Reinhart 1981), is the subject of discourse, the entity on which the comment is predicated. It corresponds to “the entity or set of entities under which the information expressed in the comment constituent should be stored in the [common ground] content” (Krifka 2008), the common ground being the information shared by the discourse participants. The I-topic is a discourse-old entity, already present in the discourse or accessible in the common ground, and it is therefore typically a definite constituent. This notion corresponds to Vallduví’s (1993) link.

Information focus: The notion of information focus, or I-focus, is linked to the pragmatic principle of Progression (e.g., Charolles 1978) stating that if a sentence is to be informative, it must contain new material. The I-focus of the sentence is brand-new information that enriches the common ground. It is both discourse-new and hearer-new. Thetic sentences are all-focus sentences, that is, sentences containing only new information. In the dialogue in (1), the answer is all-focus.

1. (1) What happened?

   The telephone rang.

An I-focus is not necessarily a contrastive focus, which corresponds to “material which the speaker calls to the addressee's attention, thereby often evoking a contrast with other entities that might fill the same position” (Gundel and Fretheim 2004; Krifka 2008). A contrastive focus may be marked by position, by prosody, or by expressions like even, only, and also.

According to Gundel and Fretheim (2004), the notions of I-topic and I-focus correspond to relationally given/new information respectively, and are equivalent to the notions of theme/rheme, or topic/comment. This is largely correct, with some caveats: the comment may contain background, discourse-old, information (shared information that is already in the common ground) in addition to new information (the I-focus per se). For example, in (2), taken from Krifka (2008: 42, ex. 41), married her in (2b) is background information within the comment (Büring 2007: 5; Vallduví’s 1993 tail).

1. (2) a.  When did [Aristotle Onassis]_I-topic marry Jacqueline Kennedy?

   b.  [He]_I-topic [married her [in 1968]_I-focus]]_Comment

4. Old French Information Structure and the Verb Second (V2) Problem

Since the end of the 19^th century, it has been observed that the Old French finite verb tends to be found in the second position in main clauses: when an element precedes the verb, the subject is postverbal (Thurneysen 1892, Foulet 1928, Skårup 1975). Accordingly, many authors analyze the Old French clause with the V2 structure typical of Germanic languages, as shown in (4) (e.g., Adams 1987, 1988; Roberts 1993; Platzack 1995; Vance 1995). In the Government and Binding framework, the verb was assumed to occupy C and the preverbal constituent [Spec,CP]. With the development of an articulated left periphery (Rizzi 1997), the verb tends to be analyzed as filling Fin, with the preverbal constituent occupying [Spec,FinP] or a higher position (see Holmberg 2015 for a review; see also Ledgeway 2008, Labelle and Hirschbühler 2017, Zaring this issue). [Spec,CP] (or [Spec,FinP]) corresponds to Skårup's (1975) place du fondement, or the Vorfeld in German.

1. (3) messe  e  matines  ad  li  reis  escultet

   mass  and  matins  has  the  king  heard

   ‘The king has attended mass and matins.’ (1100, Roland 11;139)³

2. (4) [_CP/FinP messe e matines ad [_TP li reis ad escultet messe e matines]

This V2 analysis has been challenged by a number of authors, in part because V1 and V3 constructions are more frequent in Old French than in either modern German (Kaiser 2000, 2002; Ferraresi and Goldbach 2002; Rinke and Meisel 2009; Kaiser and Zimmermann 2011) or Middle High German (1050–1350) (Elsig 2009). Rinke and Meisel (2009), for instance, argue, on the basis of a study of two early 13^th century texts (Villehardouin and Les 7 sages de Rome), that Old French is not a V2 language but a topic-initial language. They derive Old French matrix clauses within TP, with [Spec,TP] a topic position. The main claim of Rinke and Meisel is that the preverbal position hosts the aboutness topic – I-topic in our terms – and the verb phrase contains the I-focus portion of the clause. Old French is said to differ from contemporary German, where the preverbal position is not restricted to topics but may also host an I-focus or an adverb that is neither a topic nor (part of the) I-focus. If their analysis proves correct, Old French would be similar to Old High German, given Hinterhölzl and Petrova's (2010, 2011) claim that the Old High German verb served to separate the I-topic from the comment.

For Rinke and Meisel (2009: 109), subjects move to the preverbal position to avoid being interpreted as I-focus: “An incompatibility of the post-verbal subject with an interpretation as information focus or as part of a thetic sentence would cause the subject to move to the pre-verbal position.” This claim would be disproved if the I-topic of the clause may appear postverbally; this is the subject of sections 5 and 6, where postverbal I-topics are shown to be frequent in the corpus.⁴

Scene-setting elements, namely, elements specifying the time and location of the situation under which the proposition is evaluated, are often considered to be types of topics (Speyer 2004: 534–535). Their discourse role, however, differs from that of I-topics, and the two types of elements may co-occur in a clause (Nikolaeva 2001; Andréasson 2007), as (5) illustrates.

1. (5) John had a busy day. At twelve[_{Scene-setting}], he[_I-topic] was eating pizza with his brother…

In the following quote, Rinke and Meisel suggest that, in Old French, scene-setting adverbials fill the preverbal position only when no I-topic is overtly realized:

In addition, sentence-initial (scene setting) adverbial phrases may serve a special discourse function. According to Reinhart (1981), they can establish discourse cohesion by linking the sentence to the previous discourse. Preverbal adverbial phrases thus fulfill the same discourse function as preverbal topics. These adverbial expressions are therefore frequently found in sentences in which no topic is present that would normally assume the discourse-linking function.

(Rinke and Meisel 2009: 115)

If this is correct, Old French differs significantly from German. In German, a sentence may contain both a scene-setting element and an I-topic, and Speyer (2004, 2008, 2010) established that, in a sentence where both are present, the scene-setting element preferentially occupies the Vorfeld, while the I-topic remains postverbal. Consequently, in section 6.2, we study the informational role of postverbal subjects when a scene-setting element is preverbal. We will see that there are many V2 sentences with a preverbal scene-setting element and a postverbal I-topic subject.

Both in Old French and in German, I-topics are often preverbal. This is expected because they are cohesive devices linking with the previous discourse. According to Speyer's (2004, 2008, 2010) research on German, about 82% of the referential elements filling the Vorfeld in written German are elements establishing a link with the preceding discourse (scene-setting, I-topics, and elements anaphoric with a discourse-old set). The remaining cases include brand-new elements (i.e., I-focus in our terms). Hence, placing the I-focus in clause-initial position is not the dominant option in German, although it is grammatical. We will see that I-focus elements may fill the preverbal position in Old French as well. This evidence goes against the hypothesis that Old French is a topic-initial language.

Finally, if the Vorfeld in Germanic V2 languages is the host of various types of informational elements, including those pertaining to the I-Focus of the clause, but the preverbal field in Old French is a Topic position (holding I-Topics and scene-setting elements), we would expect the statistical distribution of preverbal elements to differ when the two are compared. The comparisons discussed in Section 7 do not bear out this prediction.

Our conclusion will be that, from the informational point of view, Old French does not appear to differ markedly from standard Germanic V2 languages.

5. Verb-initial (V1) Clauses

We begin with V1 clauses. If the Old French postverbal field is interpreted as the I-focus of the clause, V1 clauses should be thetic (all-focus). In other words, if subjects must move to the preverbal position in order to avoid being interpreted as part of the I-focus, the postverbal subjects of V1 clauses should be construed as belonging to the I-focus portion of the clause. To verify this prediction, we first considered sentences other than those where the verb introduces direct discourse (section 5.1), and then turned our attention to verbs of saying (section 5.2).

5.1 Verbs other than those introducing direct discourse

We extracted the strict V1 clauses of the corpus (i.e., those not introduced by a coordinator) having a full subject and a verb distinct from those introducing direct discourse. Taking the context into account, we hand-coded the subjects as being I-topic, I-focus, or Unclear⁵. Table 1 shows that, while there is a dominance of I-focus subjects, over a third of the postverbal subjects were coded as I-topic. Examples (6) and (7) illustrate postverbal subjects coded as I-topics.

1. (6) (The Philistines ask a question of the Jews and the answer doesn't please them.)

   Curecerent  s’  en  les  princes  des  Philistiens

   got-angry  refl  gen  the  princes  of.the  Philistines

   ‘The princes of the Philistians got angry at this.’ (1170; QLR, 1–2; 1332)

2. (7) (Brendan is the main protagonist. He and his followers land, and hope to relax after a long journey. However, a tempest approaches.)

   Cunuit  Brandans  a  l’  air  pluius/Que  li  tens ert mult annüus.

   knew  Brendan  from  the  air  rainy/that  the  weather was very worrisome

   ‘Brendan knew from the wet wind that the weather was worrisome’ (1120; Brendan, 56; 675)

Table 1: Informational role of postverbal subjects in V1 declaratives.

The results in Table 1 counter Rinke and Meisel's claim that, in Old French, postverbal subjects are internal to the verb phrase and interpreted as part of the I-focus. Rather, they confirm Rouveret's (2004: 196) comment that “It does not seem that the postverbal position in [Old French] V1 sentences is pragmatically specialized.”

Figure 1 illustrates the evolution of V1 declaratives over time. In Roland (1100), the vast majority of postverbal subjects were coded as being I-focus, whereas in Brendan (1120) and the QLR (1170), they were more often I-topics. The V1 construction dropped out of the language at the beginning of the 13th century, when the language became more strictly V2. This is why there are few examples of V1 declaratives after 1170; three texts dated between 1205 and 1267 contain no V1 declaratives. After 1267, V1 declaratives reappear. They contain no clear I-topic subject, but there are only seven examples. Thus, at least for the period before the 13^th century, when V1 declaratives were productively used, I-topic subjects did not need to move to the left of the verb; they could remain postverbal in V1 declaratives.

Figure 1: V1 declaratives–Informational nature of subjects: Number of postverbal subjects that are I-topic, I-focus or Unclear. The numbers at the bottom indicate the total number of relevant examples.

The postverbal position of I-topic subjects in V1 sentences is not compatible with the hypothesis that Old French sentences are dominated by TP, with the verb under T and the postverbal field expressing the comment and containing the I-focus portion of the clause. The facts are explained if we assume, following Labelle and Hirschbühler (2005), that, in V1 constructions, the emphasis is on the verb, which moves to a discourse-related position within the left periphery. Labelle and Hirschbühler propose that the verb in V1 clauses first moves to Fin, then, in a second step, to a higher discourse-related head, which they label ‘Z’, and which could correspond to focus in the articulated structure of the left periphery proposed by Benincà and Poletto (2004), shown in (8).

1. (8) [ HT [ Scene setting [ Force [ topic [ focus [ WH [ Fin ]]]]]]]

Under Labelle and Hirschbühler's analysis, postverbal subjects could occupy either [Spec,vP] or [Spec,TP], and one could explore the hypothesis that I-focus subjects occupy [Spec,vP] while I-topic subjects occupy [Spec,TP], as proposed by Rinke and Meisel.

1. (9) a.  [_FocP Cunuit [_FinP cunuit [_TP cunuit [_vP Brandans cunuit …] (=(6))

   b.  [_FocP Cunuit [_FinP cunuit [_TP Brandans cunuit [_vP Brandans cunuit …]

But to account for the fact that postverbal subjects could be I-topics, we need to assume a derivation of V1 clauses that involves the left periphery in a manner similar to what happens in V2 sentences. Such a structure is compatible with the V2 nature of Old French if we assume, with Ledgeway (2008: 439), that V2 is not a surface linear constraint, but a syntactic requirement that the verb raise to the C-domain in general, though not necessarily, with fronting of a pragmatically salient constituent to its left.

5.2 Verbs introducing direct discourse

Let us now turn to verbs introducing direct discourse. As schematized in (10)–(12), with these verbs, V1 declaratives (10) alternate with V2 declaratives, some of the SV(X) type (11), and some of the XVS type (12), where the preverbal element is either a cataphoric object or an adverb.

1. (10) Dit Roland: “ … ”     = VS

   ‘Says Roland: ….’

2. (11) Roland dit: “ … ”      = SV

   ‘Roland says: ….’

3. (12) a. Ço dit Roland: “ … ”  = Cataphoric object VS

   ‘This says Roland: ….’

4. b. Donc dit Roland: “ … ”  = Adverb VS

   ‘Thus says Roland: ….’

We did not consider verbs of saying in parentheticals, nor cases where the verb of saying follows the direct discourse, because there is no word order variation in these cases: the subject is always postverbal.

As far as we could see, the informational role of the subject is the same in examples (10)–(12). If this is correct, the preverbal or postverbal position of the subject is not dictated exclusively by its informational role. The following analysis attempts to verify this by focusing on the following question: with the disappearance of VS declaratives at the beginning of the 13^th century, which construction replaces VS declaratives? If the postverbal position of the subject in VS declaratives was dictated by Information Structure, VS sentences are expected to be replaced by XVS sentences, to keep the subject postverbal. However, Figure 2 shows that, when the rate of VS sentences dropped dramatically (between 1177 and 1225), the proportion of SV(X) sentences increased, but not that of XVS sentences.

Figure 2: Verbs introducing direct discourse: Rates of SVX sentences as compared to VS and XVS sentences. The numbers at the bottom correspond to the number of relevant examples.

The fact that SV sentences are replacing VS sentences casts serious doubts on the idea that the subject's informational role dictates its position. Moreover, there is no support for the hypothesis that I-topic subjects could not remain postverbal. The vast majority of the subjects are definite and discourse-old both in SV and in VS sentences. In 99% (573/580) of the examples containing a verb of saying, the subject is definite.⁶ As definite subjects correspond to known information, they tend to represent the I-topic, particularly in the discourse context of verbs of saying.

Figure 2 suggests that VS order in declaratives resulted from a desire to place the verb first, and not from the need to leave the subject in a postverbal I-focus position. When speakers started to avoid placing emphasis on the verb in declaratives, a V2 construction was used instead, with the verb remaining under Fin and some constituent filling [Spec,FinP]. The fact that I-topic subjects were largely preferred over either objects or adverbs to fill the preverbal position is coherent with the distribution of preverbal elements in V2 sentences (see section 7, table 4).

To summarize this section, the hypothesis that subjects need to move to the preverbal position to avoid being interpreted as I-focus is refuted. Postverbal subjects in V1 clauses are often the I-topic of the clause. It could very well be that subjects move to [Spec,TP] in order to be interpreted as I-topics. However, we must conclude that the Old French clause structure projects higher up than [Spec,TP] and that the Old French verb moves to the left periphery in V1 main clauses. This movement to the left periphery is in line with the standard V2 analysis of Old French.

6. V2 Clauses

Let us now turn to V2 clauses. Two questions are addressed in this section: First, what is the informational role of the postverbal subjects in V2 clauses? Second, what is the informational role of the preverbal constituents?

6.1 Subjects

Because definite subjects tend to be I-topics, whereas indefinite subjects make bad I-topics and are often found in thetic sentences, we asked ourselves whether, in V2 sentences, there is a marked tendency to find definite subjects preverbally and indefinite subjects postverbally. To answer this question, we automatically coded the nature of subjects in the 4397 non-coordinated V2 sentences. Subjects coded as “definite” include those introduced by a definite, possessive or demonstrative determiner, DPs headed by a demonstrative or other referential pronoun, and proper nouns; “indefinite” subjects include DPs headed by a partitive or indefinite determiner. Bare NPs and DPs headed by a quantifier (like tous ‘all’, plusieurs ‘many’), which could be definite or indefinite, were counted separately. Figures 3a and 3b show that subjects are overwhelmingly definite, both preverbally and postverbally. Given that definite subjects mostly introduce known or retrievable information, and rarely new information, the large number of postverbal definite subjects casts serious doubt on the idea that postverbal subjects could not be I-topics.

Figure 3a: Preverbal subject types: Rates of subject types in preverbal position.

Figure 3b: Postverbal subject types: Rates of subject types in postverbal position.

Preverbal definite subjects are expected to be I-topics. Because this would be compatible with Rinke and Meisel's claims, we did not study them further, but rather focused our attention on the 192 indefinite subjects. We found a small tendency to place indefinite subjects postverbally, but clearly no exclusion of indefinite subjects from the preverbal position: 44.7% of indefinite subjects are preverbal in verse texts (12^th century) and 43% in prose texts (1170 and 13^th century). Examples (13) and (14) show that these indefinite preverbal subjects may be the I-focus of the clause; the contexts should make clear that the indefinite subjects carry brand new information. We conclude from this that the preverbal position may host informationally new, I-focus, elements.^{7, 8}

1. (13) (Eliduc is looking for a burial place for his love. He remembers that near his domain there is a large forest. A holy hermit lived there, and there was a chapel there. He had talked to him many times. He will bring his love to him, for him to bury her in his chapel.)

   Un  seinz  hermites  i  maneit  et  une  chapele i  aveit

   A  holy  hermit  loc  lived  and  a  chapel  loc  had

   ‘A holy hermit lived there and there was a chapel there.’

   (1180; Marie de France, Eliduc 182;3715)

2. (14) (While Eliduc's wife is crying next to a dead girl, a weasel comes running out from under the altar and passes over the corpse. The valet kills it and throws its corpse in a corner.)

   Une  musteile  vint  curant,  desuz  l’  auter  esteit  eissue

   A  weasel  came  running,  under  the  altar  was  come.out

   ‘A weasel came running out from under the altar.’

   (1180; Marie de France, Eliduc 187;3815)

Considering only texts containing more than ten indefinite subjects, we observe a small increase in the tendency to place indefinite subjects in postverbal position (Figure 4), but the percentage of preverbal indefinite subjects at the end of the period is still 35%. Thus, there does not seem to be any constraint against preverbal I-focus subjects during the period under investigation.

Figure 4: Postverbal indefinite subjects over total indefinite subjects (in text with more than 10 indefinite subjects): Rates of postverbal indefinite subjects. The numbers at the bottom correspond to the total number of relevant examples.

To summarize this subsection, the distribution of definite vs indefinite subjects in V2 clauses does not support the hypothesis that subjects move to the preverbal position to avoid being interpreted as I-focus. This conclusion will be supported in Section 6.2.3, which presents a quantitative analysis of the informational role of postverbal subjects in sentences containing a preverbal adverbial.

6.2 Preverbal nonsubject constituents

Rinke and Meisel (2009) claim that, while in German the preverbal constituent may be a topic, an I-focus, or an element that is neither topic nor focus, the preverbal constituent in Old French is always a topic (an I-topic, or an adverb with topical properties). In this section, we consider XVS sentences with the aim of determining whether the preverbal position is a topic position, and whether postverbal subjects are interpreted as being part of the I-focus portion of the clause.

Since clause-initial contrastive elements could result from movement to a dedicated contrastive focus projection within the left periphery (Rizzi's 1997 FocP), we will set aside examples like (15) containing an overtly contrastive preverbal constituent.

1. (15) Meïsmes  a  l' empereour  sont  les  lermes  venues  aus  yex,

   even  to  the  emperor  are  the  tears  come-ptcp  to-the  eyes

   ‘Even to the emperor, the tears came to the eyes.’ (1267; Cassidorus, 664;4447)

6.2.1 Preverbal objects.

The OVS word order in 13^th century texts has been studied by Rickard (1962), who shows that, in addition to establishing a link with the previous discourse, the preverbal position may also be a prominence position serving to convey a brand-new and affectively important idea. Marchello-Nizia (1995, 1999), who studied the role of preverbal direct objects in two texts, Roland (1100) and La Queste del Saint Graal (1225), noticed a change in the informational role of preverbal objects. In Roland, there is no restriction on the informational role of preverbal objects, but they tend to be, in her terms, more often rhematic (I-focus) than thematic (I-topic) (Marchello-Nizia 1995: 99–100). By contrast, the OV(S) order is more restricted in the Queste, where, according to Marchello-Nizia, it serves mainly to thematize the object. When the object is rhematic, the OVS order is marked and is found mainly in idioms of type mander saluz ‘send greetings’, as in (16), or when O is either strongly focused or cataphorically linked to the discourse.⁹

1. (16) Saluz  vos  mande  li  bons  chevaliers

   Greetings  you  send  the  good  knight

   ‘The good knight sends you his greetings’   (Queste, p. 30, cited in Marchello-Nizia 1999: 44)

Our corpus study confirms the change in the informational role of preverbal objects. Taking context into account, we coded all preverbal objects in V2 sentences as I-topic, I-focus, or unclear. As shown in Figure 5, a preverbal object tended to be the I-focus of the clause before 1205, but the I-topic of the clause, after 1225.

Figure 5: Informational status of preverbal objects: Rates of preverbal objects that are I-focus, I-topic, or Unclear. The numbers at the bottom correspond to the total number of relevant examples.

The examples in (17)–(20) contain preverbal objects analyzed as part of the I-focus:

1. (17) (Marsile comes through a valley with the great army he had assembled)

   .XX.  escheles  ad  li  reis  anumbrees.

   twenty  columns  has  the  king  counted

   ‘The King has organized them in twenty divisions.’¹⁰ (1100; Roland, 112;1459)

2. (18) (After the storm, a large quantity of birds assembled on the pine tree, they all sang with perfect harmony.)

   et  divers  chanz  chantoit  chascuns;

   and  various  songs  sang  each.one’

   ‘and they all sang a different song’ (1177; Yvain, 15;453)

3. (19) (Lunete asks the lady to swear that she will help the knight)

   La  main  destre  leva  adonques  la  dame,

   The  hand  right  raised  then  the  lady

   ‘Then the lady raised the right hand’ (1177; Yvain, 202;7065)

4. (20) (While the king fortified Sayete, merchants arrived and told us that the king of the Tartars had taken the city of Baudas.)

   La  maniere  comment  il  pristrent  la  cité  de  Baudas et  le  calife  nous  conterent

   The  manner  how  they  took.pl  the  city  of  Baudas and  the  caliph  us  told

   les  marcheans;

   the  merchants

   ‘The merchants told us how they took the city of Baudas and the caliph’

   (1309; de Joinville 289; 3370.¹¹)

It can be seen that the postverbal definite subjects are not part of the comment or I-focus of the clause, but constitute the old information of the clause and its I-topic (the entity under which the comment must be stored in the common ground).

A change in the informational ordering of constituents seems to have happened at the beginning of the 13^th century. This change led to a tendency to leave I-focus objects in postverbal position. This trend could suggest that the language changed from having a preverbal position unrestricted with respect to information role, towards a situation where the preverbal position tends to be restricted to I-topics. However, since the OVS order with a preverbal I-focus object counts for twenty to forty percent of preverbal objects until the end of the period, one cannot say that the language becomes a topic-initial language at the beginning of the 13^th century. The detailed analysis of the OV order carried out by de Andrade (this issue) confirms this finding. De Andrade shows, among other things, that it is only in the 15^th century (that is, in Middle French) that informational focus objects become rare in preverbal position.¹² We will come back to the change that affected preverbal objects in section 7 below.

We conclude that, in Old French, the preverbal position may be occupied by an I-focus, in which case a postverbal subject may be the I-topic.

6.2.2 Preverbal predicates and quantifiers

In Old French, the preverbal position may be filled by constituents that are not good candidates for topic-hood. These include quantifiers, non-finite verbs, adjectives and other predicates.

In (21), the preverbal predicate is known information. It establishes a link with the previous chapter and introduces a contrast with the second clause. It is neither the I-topic nor the I-focus of the discourse.

1. (21) (Beginning of the section on rebellion, after a section on the sin of grumbling [murmuring against an authority].)

   Male  chose  est  murmure,  mes  trop  vaut  pis  rebellion.

   bad  thing  is  grumble  but  too.much  be.worth  worse  rebellion

   ‘Grumbling is a bad thing, but rebellion is much worse.’ (1279; Somme, 1,64; 1758)

However, in general, preverbal predicates and quantifiers are interpreted as part of the I-focus. When that is the case, subject may appear postverbally and be construed as I-topics. Examples (22)–(24) illustrate this situation.

1. (22) (Lanval leaves his hosts, rides towards the town. He often looks behind him.)

   Mut  est  Lanval  en  grant  esfrei!

   much  is  Lanval  in  great  fear

   ‘Lanval is greatly afraid.’ (1180; Marie de France, Lanval, 196; 78.1592)

2. (23) (Count Roland says to Marsile “…today you'll learn my sword's name”. He goes to strike him.)

   Trenchet  li  ad  li  quens  le  destre  poign

   cut-ptcp  him  has  the  count  the  right  hand

   ‘The count has cut his right hand.’ (1100; Roland, 142; 1926)

3. (24) (The girl, hiding in a shelter, disinherited and disconsolate, becomes full of joy upon hearing that the Lion Knight is coming (to defend her). She thinks that, now, her sister will give her part of her inheritance.)

   Malade  ot  geü  longuemant  la  pucele

   sick  has  laid  a.long.time  the  girl

   ‘The girl had been sick for a long time.’ (1177; Yvain, 177; 6235)

Preverbal predicates and quantifiers were always rare in preverbal position compared to adverbials and PPs. Table 2 shows that their percentage is higher in verse texts (12^th century) than in prose texts (mainly 13^th century). It could be that versification constraints favored the use of an atypical word order. It might also be that after the 12^th century, speakers were less likely to place I-focus elements in preverbal position, echoing what was found with objects.

Table 2: Distribution of adjectives, predicative NPs and non finite verbs in preverbal position.

6.2.3 Adverbials

We coded the informational status of postverbal subjects in V2 sentences introduced by an adverbial, or a scene-setting temporal DP like le lendemain ‘the next day’. There were about 9% of cases that we were unable to code (86 cases out of 975); these were removed from the analysis. The majority of the postverbal subjects (71%) were coded as I-topic. Only 29% of the postverbal subjects were coded as new information, that is, as I-focus. Figure 6, illustrating only the texts for which we had at least nine relevant examples, demonstrates that no observable change is apparent within the period considered.

Figure 6: Information function of postverbal subjets in Adv-V-S clauses: Percentage of postverbal subjects that are coded as I-focus or I-topic in texts containing more than nine relevant examples.

In the majority of cases, the preverbal adverb is a short temporal adverbial like lors, puis, après, donc, adont, or the particle si. In these cases, the preverbal element is neither a topic nor an I-focus, but a cohesive element, linking to the previous discourse. When such an element is preverbal, the I-topic appears postverbally. The postverbal placement of I-topics is expected in a V2 system, and it is observed in German.

1. (25) (Ydoines asks Helcanor if he prayed God to give him his hands back; Helcanor answers “Yes I did.”)

   Dont  sot  Ydoines  vraiement  que  Diex  l’  amoit,

   Then  knew  Ydoine  truly  that  God  her  loved

   ‘Then Ydoine truly knew that God loved her.’ (1267; Cassidorus, 351; 3031)

2. (26) (Perceval wants to leave immediately, but his aunt refuses to give him leave, saying that he should stay overnight and leave after hearing mass in the morning. Perceval stays. The servants install the table. Perceval and his aunt eat the meal prepared for them.)

   Si  demora  laienz  Perceval  avec  s’  antain.

   thus  stayed  therein  Perceval  with  his  aunt

   ‘Thus, Perceval stayed there with his aunt.’ (1225; Queste, 107; 2806)

A second category of preverbial adverbials includes locative or temporal scene-setting elements. Here again, the postverbal subject is usually interpreted as the I-topic of the clause. The examples in (27) and (28) illustrate this type of situation. These examples show that in Old French, scene-setting adverbs do not necessarily occupy the preverbal position when no I-topic is available to fill it. On the contrary, Old French appears to function like German, where, according to Speyer (2008), when a clause contains both a scene-setting element and an I-topic, the scene-setting element has priority for filling the Vorfeld.

1. (27) (Cassidorus takes his spear and runs towards Lapsus to fight.)

   La  fist  Cassidorus  plus   d’  armes

   There  made  Cassidorus  more  (feats) of  arms

   que  onques  Achilles  ne   fist  devant  Troies.

   than  ever  Achilles  neg  made  in.front.of  Troy

   ‘There, Cassidorus made more feats of arms than Achilles ever made in front of Troy.’

   (1267; Cassidorus, 26; 627)

2. (28) (During Lent, the king had his boats prepared to go back to France. On the eve of Saint Mark's day, after Easter, the king and the queen prayed on their boats; they set sail with fair winds.)

   Le  jour  de  la  saint  Marc  me  dit  le  roy que  a  celi  jour  il  avoit  esté  né;

   The  day  of  the  Saint  Marc  me  told  the  king  that  at  that  day  he  had  been  born

   ‘On Saint Marc's day, the king told me that he was born that day;’

   (1309; de Joinville, 305; 3574)

Our results can be summarized as follows. In the preverbal position of V2 sentences, we observed not only elements treated as topics by Rinke and Meisel (I-topics and scene-setting adverbials), but also elements that were the I-focus of the clause as well as adverbs that are neither topics nor focus. In postverbal position, we counted a large number of definite subjects interpreted as the I-topic of their clause. We conclude that Old French was not a topic-initial language and that subjects could remain postverbal without being interpreted as part of the I-focus. Therefore, from the informational point of view, Old French resembles German. The standard analysis whereby the verb moves to the left periphery in main clauses accounts adequately, not only for the V2 examples, but also for the V1 examples.

7. Comparison with Germanic Languages

In this section, we compare the distribution of preverbal constituents in Old French and in Germanic languages. If the information role of the preverbal constituent in Old French is similar to what is observed in Germanic V2 languages, we expect the distributions to be similar. If, however, Old French is not V2 but is a topic-initial language, we expect this fact to be reflected in the distribution. We base our comparison on the study by Bohnacker and Rosén (2007).

Bohnacker and Rosén (2007) compare the prefields in German and Swedish, and they arrive at the figures in Table 3 (which combines their Tables 1–3, pages 34 and 36). Table 3 can be compared with our results for Old French in Table 4. The percentages of the various types of preverbal constituents are of the same order of magnitude in Old French and in the two Germanic languages. This comparison reinforces our claim that during the 12^th and 13^th centuries, Old French was not a topic-initial language, but, rather, a V2 language similar in the relevant respects to German and Swedish.

Table 3: Constituents in the prefield in German and Swedish (from Bohnacker and Rosén 2007).

Table 4: Constituents in the prefield in Old French.

Commenting on the figures in Table 3, Bohnacker and Rosén explain the difference in the percentages of fronted objects in German (about 7%) as compared to Swedish (about 3%) as follows: both Swedish and German tend to start declaratives with a subject coinciding with the theme and topic, and to place the theme before the rheme. However, Swedish has a stronger tendency to place the rheme after the verb, to start with a light element or one of low informational value (e.g., an expletive, det, så), and to use few fronted objects. Although the constructions with a heavy rhematic (i.e., I-focus) object found in German are considered grammatical in Swedish, they are dispreferred. Swedish typically fronts objects that are themes (i.e., I-topics).

Turning to Old French, Table 4 shows that the percentage of preverbal objects in verse texts (12^th century) is similar to what is found in German, while the corresponding percentage in prose texts (mainly13^th century) is similar to what is found in Swedish. If we add to this observation the decrease of I-focus preverbal objects in prose texts (section 6.2.1), the lower percentage of preverbal predicates and non-finite verbs in prose texts (section 6.2.2), and the small increase in the proportion of postverbal indefinite subjects (section 6.1), we are tempted to conclude (if the differences are not exclusively due to text genre) that the use of the prefield during the period considered evolved from one similar to that of contemporary German to one similar to that of contemporary Swedish. These changes might indicate a change towards a preference for keeping I-focus constituents within the VP. It is clear, nonetheless, that I-topics are not restricted to the preverbal position during the period considered, contrary to what was argued by Rinke and Meisel (2009). Moreover, as German and Swedish are both V2 languages, the shift observed in Old French cannot be taken as an argument against the analysis of Old French as a V2 language.

8. Conclusion

We have argued that, from an information-structure viewpoint, Old French was essentially a V2 language of the Germanic type until the end of the 13^th century. Just as in German, the preverbal position of an Old French sentence could be filled by an I-focus, by an I-topic, by a scene-setting topic, or by an adverb that was neither a topic nor part of the I-focus. Moreover, postverbal subjects were frequently interpreted as I-topics; therefore, subjects did not need to move to the preverbal position to avoid being interpreted as part of the I-focus. This is what is expected of a language in which the verb moves to a position within the left periphery (FinP or higher) with some constituent merged to its left in V2 sentences. If the verb occupies a position within the left periphery, postverbal subjects in SpecTP may be construed as I-topics.

Our research allowed us to advance in the description of the evolution of Old French, in that we noticed a slight change in the use of the left periphery. The 12th-century grammar is statistically closer to that of German, while the 13th-century grammar is similar to that of Swedish. This shift appears to reflect a change in progress towards a grammar where new information preferentially occurs in postverbal position. Specifically, noncontrastive elements carrying the I-focus of the clause are less frequently preverbal at the end of the period, compared to what was found at the beginning. If the differences are not exclusively due to text genre, they reflect one of the first steps towards the loss of the V2 grammar and the emergence of an SVO grammar, as argued by Marchello-Nizia (1995). It would be interesting to study Middle French V2 sentences to better understand this evolution.