Contrast and word order: A case study on the history of Portuguese

Several studies have observed that older Romance languages had more frequent object-verb order than their modern counterparts. This article explores the idea that contrast is crucial to understand the shift to verb-object order, as part of a more encompassing notion of boundedness, which has been frequently associated with the V2 parameter. To do so, we first show that some fronting constructions involving demonstratives were available in Old and Classical Portuguese, but not in Modern Portuguese, as a consequence of the existence of a KP projection hosting contrastive items of different sorts acting as delimitators. Second, we present some changes between Old and Classical Portuguese: (i) a decrease in fronted objects with demonstratives; (ii) an increase in the frequency of null subjects; and (iii) a small decrease in the frequency of frame setters. We propose that these shifts are to be attributed to an information-structural change correlated with a syntactic change concerning the position of complementizers and verb movement in subordinate clauses. To wit, while Old Portuguese was a symmetrical relaxed C-V2 language, Classical Portuguese was an asymmetrical one, which is reflected in terms of a reduced left periphery of subordinate clauses in the latter, in which KP cannot be projected. The article presents some implications of this account for the derivation of basic word order patterns and of clitic interpolation in Old Portuguese and for the loss of V2 in the history of Portuguese. These results suggest that the connection between boundedness, contrast and V2 grammars can be profitably explored in Old Romance languages.


Introduction
Several studies have observed that older Romance languages had more frequent preverbal objects than their modern counterparts (Antonelli 2011;Sitaridou 2011;Zaring 2011;Poletto 2014, among others). In generative works applied to Romance varieties, this word order has been explained as the result of movement to left-peripheral projections, such as those related to topic or focus (Benincà 2004;Benincà & Poletto 2004;Wolfe 2015), whereas for Old Portuguese IP-scrambling was suggested as a possible solution as well (Martins 2002). It is less consensual whether all Romance languages went through a V2 stage, though in recent years there is a growing tendency to admit so, while at the same time the notion V2 grammar has become nuanced (Jouitteau 2010;Holmberg 2015;Wolfe 2015).
We put forward the idea that contrast is the relevant feature for the description of the V2 nature of Old Portuguese (OP) and Classical Portuguese (ClP). We then pinpoint some differences between OP and ClP which are somehow related to the marking of contrast and other features of boundedness, a cognitive notion that is closely related to V2 grammars. We argue that bounded languages present some element in the left periphery of the clause Glossa general linguistics a journal of de Andrade, Aroldo and Charlotte Galves. 2019. Contrast and word order: A case study on the history of Portuguese. Glossa: a journal of general linguistics 4(1): 107. 1-29. DOI: https://doi.org/10. 5334/gjgl.897 that encodes what has been termed local anchoring, so that the clause is associated with the immediately previous discourse. Demonstrative pronouns usually play this role, thus consisting in a system of marked anaphors vis-à-vis personal pronouns.
These observations suggest that the study of Old Romance syntax must take into account the intertwining between syntax and information structure (as well as with textual organization), and therefore also fine quantitative evidence related to the frequency of alternative constructions. This text reassesses previous work by Eide & Sitaridou (2014), which is devoted to the study of contrastive items in Ibero-Romance languages. They suggest that no important change has taken place regarding contrast, in a comparison between the old and modern stages of these languages. However, we defend that different assumptions and research methods may lead to the opposite conclusion. 1 The paper is organized as follows. In section 2 we detail why contrast should be understood as a central informational category in V2 grammars. In section 3 we organize the arguments for the V2 status of OP and ClP. Section 4 details the position to which fronted contrastive items move in these language stages, in comparison with similar data in Modern Portuguese (MP), and offers some pieces of evidence to this end. In section 5 we explore some information-structural differences between OP and ClP concerning the behavior of different phenomena expressing boundedness: object position (especially those with a demonstrative), subject expression, and frequency of frame setters. In section 6, we put forward an account for a syntactic change between OP and ClP involving verb movement in subordinate clauses. Section 7 concludes the text.

CP-level Verb-Second grammars and contrast
A series of papers defend the existence of contrast as an autonomous informational category. Among those, for Vallduví & Vilkuna (1998), contrast may be connected to either topics or foci. Applying Neeleman & Vermeulen's (2012) featural matrix, it appears that the four main information-structural categories may be derived from three basic features, as shown in Table 1, adapted from Neeleman & Vermeulen (2012: 25).
The table suggests that [contrast] should act as a freestanding feature as well. We believe the resulting category corresponds to a delimitator in a more general sense than that explored in Krifka (2007): it may be a frame setter, which refers to an element that sets the frame under which the following expression should be interpreted, as in (1a), or a delimitator proper, an element indicating that the current communicative needs of the common ground are not wholly satisfied by a proposition, as in (1b): (1) Modern English (Krifka 2007: 45, 48) 1 Eide & Sitaridou's Portuguese corpus is composed of two narrative texts, Crônica Geral de Espanha and Vidas de Santos, the first one of which was also used in our research. Since the two corpora are strikingly similar, one must consider that the more restrictive notion of contrast adopted in Eide & Sitaridou (2014) is at the base of the mentioned differences, along with other methodological considerations. Notice the similarity between frame setters and delimitators, in that the two have a focus inside them, and another one outside. In (1a), the context would be a debate about the state of an enterprise business, where alternative statements are {S 1 = In the United States they are losing money; S 2 = In Argentina they are stable, etc.}. Similarly, (1b) may be uttered in a context where people are debating about the quality of a mathematician, where alternative statements are {S 1 = He is a mediocre mathematician, S 2 = He is a really bad mathematician, etc.}. The difference is that a delimitator shows a clearer difference between it and the internal focus. (Despite this conjoined term, we may use the term frame setter when we intend to single out the temporal or locative usage of delimitators.) A necessary implication of the featural approach presented above is that contrastive items do not form a uniform category; instead, there are degrees of contrast, which stand in a continuum, possibly applied as in the following: (2) Delimitators < Contrastive topics < Contrastive foci This is a consequence of the fact that only with contrastive foci (CFs) is there an implicature of exhaustive denial of the alternatives (Lee 2003). Besides that, contrastive topics (CTs) and CFs share a clearer identification of the alternatives, whereas delimitators only imply a looser set thereof. Notwithstanding these differences, the following basic definition is valid for all the various contrastive items mentioned above: "if an expression a in a declarative sentence S is contrastively interpreted, a set M of expressions, |M| ≥ 2, is assumed to exist which contains a and the expressions which denote salient alternatives to the denotation of a" (Frey 2006: 246). We would like to explore an analysis of the expression of contrast in C-V2 grammars according to the following hypothesis. Although contrast has relevance in virtually any language, it is only syntactically encoded apart from focus and topic features in C-V2 languages, i.e. those in which there is V-to-C movement (Vikner 1995). The idea of the syntactic visibility of contrast-in terms of movement of the contrastive constituent to a relevant left-peripheral position-has been explored in different languages in Vallduví & Vilkuna (1998), Molnár (2002) and Frey (2006). Such an element is usually associated with some sort of pragmatic and/or prosodic highlighting as well.
For simplicity reasons, we will consider Modern German as the prototype of a C-V2 language. In Modern German there is a preference for encoding contrastive items at the left periphery of the clause, using a monoclausal construction. By contrast, in a language like Modern Portuguese, CFs may occur in postverbal position or in a cleft construction, and these are preferred strategies vis-à-vis the monoclausal structure with Focus Fronting: (3) Modern German (Frey 2006: 247, with  CTs and Delimitators show similar behavior, in that in C-V2 grammars there is a strong tendency for them to occur in preverbal position, a facet of what has been termed local anchoring in Los (2012). Therefore, topics and delimitators consist in local anchors, because they ensure a link to the immediately preceding discourse, in terms of either a topic change or a frame setting change: 2 (5) Cintra ( These examples show that contrastive items usually trigger subject-verb inversion in OP, unlike in MP. As discussed in the following section, this is an important piece of evidence for the C-V2 status of OP (and of ClP). In this scenario, contrastive constituents by hypothesis move to the left periphery, from a position inside the clause, in accordance with the fact that phrases that bear a [contrast] feature tend to occur in the prefield more than in the middlefield (Speyer 2008a;b). 3 In terms of prefield ranking, contrastive items (including frame setters) are the ones most frequently occurring in this position, a fact that may find a motivation in the framework described below.
What stands behind this special role of the prefield in C-V2 languages (and probably in V2 languages generally) is the idea of boundedness. Psycholinguistic studies have explored the fact that the marking of temporal frames is linked to how the preverbal position is expressed in different languages. A crucial experiment involving the retelling of a film shows that English speakers tend to link the subevents of the story to a main temporal frame, whereas German speakers tend to bound each subevent, i.e. to present them with endpoints: Modern English (Carroll et al. 2004: 190) A -What happens in the scene? B -a. A young man is surfing.
b. The wind is blowing him off the board.
(8) Modern German (Carroll et al. 2004: 190)  The strategy to mark the temporal shift in German involves the use a frame setter adverbial (dann, 'then'), which allows keeping the reference on the same subject; on the other hand, in English it consists in changing the subject for an inanimate participant (the wind). As a consequence, it seems that the German participant tells the story "from within", as if s/he were the protagonist. 4 This contrasts to the external viewpoint of the English participant. According to Los (2012) and Komen et al. (2014), the loss of this system had greater consequences to the history of English, such as the fixation of word order and the development of progressive tenses.
A consequence of local anchoring is that referential chains tend to be kept for longer stretches of text. For some languages, this implies a system with two pronouns, one unmarked and the other one marked. This has been observed for Dutch in Comrie (1997), and seems to be valid for C-V2 languages generally. 5 Markedness is embodied in the functional difference between personal and demonstrative pronouns, the marked status of the latter being noticeable in the syntax from its privative usage in the prefield, but in variation with the former in middlefield positions, as shown for German in the pair (9b-c): 4 The prefield ranking model foresees the same result from a distinct theoretical standpoint. It is interesting to observe the much higher probability of frame setters to appear in the prefield, if compared to aboutness topics (usually the clausal subject). Besides, according to Speyer (2008b: 482), in Early New High German, frame setters had a specialized position, corresponding to a higher TopP, and could co-occur with other prefield elements. 5 Studies have shown this pattern for Dutch and German. In Scandinavian languages, which are usually considered to be I-V2 languages, the distinction is not so strong: personal pronouns may be used instead of demonstrative pronouns, or the latter ones may have been reanalyzed as [-animate] markers, especially when used in Contrastive Left Dislocation structures (cf. van Gelderen 2013). The explanation may be that a canonical subject position and peripheral positions would not be distinguished in these languages. Notice though that the generalization according to which a C-V2 language implies a distinction between marked and unmarked pronominal forms is not valid in the reverse direction. This is expected because the distribution of marked and unmarked forms seems to be well widespread (Diessel 1999), more than the C-V2 condition is.

(9)
Modern German (adapted from Bosch et al. 2003: 61;Wiltschko 1998: 178) a. Paul i wollte mit Peter k laufen gehen. Paul wanted.3sg with Peter run.inf go.inf 'Paul wanted to go running with Peter.' b. Aber {er i/k / der *i/k } war leider erkältet. but he this.m.nom was.3sg sadly cold 'But he had a cold.' c. {Den *i/k / Ihn *i/*k } habe ich gesehen. this.m.acc 3sg.m.acc have I seen 'I have seen this one.' The following excerpt shows that a similar distribution was found in OP: Cintra ( In information-structural terms, the demonstrative is marked because it refers to the less salient referent, whereas the personal pronoun is unmarked because it is used to keep the same discourse topic (Hinterwimmer 2015). 6 In other words, the idea of a marked pronoun refers to the usual existence of two possible antecedents, "one of which is excluded, the other actualized" (Comrie 1997: 53). The examples above show that Modern German and OP are similar in this respect, and give further evidence for the idea that contrast is visible in C-V2 grammars, where a contrastive item stands in a set of alternative referents explicitly or implicitly retrievable in the context (Repp 2016).

Arguments for V-to-C movement in Old and Classical Portuguese
In this section we present a summary of findings of previous works exploring the main syntactic arguments presented in the literature for the characterization of OP and ClP as C-V2 grammars, i.e. showing V-to-C movement, at least in non-dependent clauses (cf. Before delving into the specifics of the arguments, we crucially observe that this classification takes V2 in its configurational aspect, instead of focusing on the word-order effect derived from it. In fact, OP and ClP show different amounts of actual V2 orders, with relevant amounts, but not corresponding to 100% of the cases. However, as it is well known, the V2 characterization is couched on other pieces of evidence that confirm a different clausal configuration in these languages, whereby there is no canonical subject position, as detailed below.

The relative position of adverbs and postverbal subjects
vP adverbs such as bem ('well') can appear either on the left (11a) or on the right (11b) of postverbal subjects. The latter position is impossible in Modern Romance languages but is typical of V2 languages like Icelandic (Belletti 2004). This suggests that the subject is in Spec,IP and that the verb has moved up to C (Antonelli 2011: 156) b. E nos gasalhados e abraços mostraram os cardeais legados and in.the affections and greetings showed.3pl the cardinals delegates bem este contentamento well this satisfaction 'And the delegate cardinals have very well shown this satisfaction (to me) in their affections and greetings'

The position of postverbal subjects in verbal compounds
Sentences like the one exemplified in (12) are also usually pointed out as evidence for Germanic Inversion, in which there is I-to-C movement of the auxiliary verb whereas the subject remains in Spec,IP, between the auxiliary and the main verb: Note that an additional piece of evidence is that the subject esta Religiosa ('this nun') is not interpreted as the focus of the sentence, as further discussed in the following section.

The informational role of postverbal subjects
In ClP and in OP, postverbal subjects may be interpreted as having the role of informational focus, as in (13a). This is compatible with they being kept in their base-generated position, Spec,vP. Nevertheless, they may also receive another kind of interpretation, as a familiar topic, i.e. one which refers to salient information in the whole text, as illustrated in (13b)

The frequency of XV and of VS
Both the occurrence of VS order and the usage of preverbal constituents other than the subject (XV) are taken as crucial triggers for the acquisition of V2 grammars. As discussed in Galves & Paixão de Sousa (2017: 166-167), the amount of XV in ClP, including sentences with null subjects, is higher than that of SV, at a proportion of 43% against 16% in authors born the 16 th century. This count is higher than the one found for Modern German (Lightfoot 1997). This is in sharp difference with respect to MP in the 19 th century, in which the proportion is inversed: 17% (XV) against 56% (SV).
Besides, the frequency of preverbal subjects was smaller in OP and ClP, if compared to the one found in MP texts. This suggests that in the older grammars they competed with fronted objects and adjuncts. To wit, the overall frequency of VS in OP is of about 20% (Medeiros 2018), whereas in ClP it is of 28% (Galves To appear) and in MP it occurs by 10% (Galves & Paixão de Sousa 2017). This difference is attributable to the loss of Germanic Inversion, which finds a natural explanation in terms of V-to-C movement. 9,10 Regarding the derivation of C-V2 in OP and ClP, we assume that C has an EPP-feature which triggers verb movement, due to the rich verbal agreement of null-subject languages (cf. Antonelli 2011, following ideas on Alexiadou & Anagnostopoulou 1998). In terms of Rizzi's (1997) articulated left periphery, there would be V-to-Fin movement, and no Spec,FinP would be projected. As a consequence, no bottleneck effect-according to which Spec,FinP would be the entrance target to other left-peripheral categories (Roberts 2004)-would be available, unlike in Germanic languages under a C-V2 approach. This entails that Old and Classical Portuguese may be classified as relaxed V2 grammars (cf. Medeiros 2018 and Galves To appear, following ideas on Wolfe 2015). Still according to another classification, that of Vikner (1995), OP and ClP are classified together as C-V2 languages. Nevertheless, under a third classification, i.e. the one separating symmetrical from asymmetrical V2 languages, OP and ClP are indeed different, as we shall discuss later on. 9 An anonymous reviewer suggested that these empirical points could be undermined by a theoretical account according to which any fronted constituent could occur in Spec,IP. In terms of Camacho-Taboada & Jiménez-Fernández's (2014) proposal for Focus Fronting in Modern Spanish, some evidence in this direction could be drawn from differences between A-and A′-movement. According to the authors, a crucial piece of evidence for supposing that a focus-fronted constituent las peras ('the pears') sits in Spec,IP is the possibility of having a floating quantifier in a raising (or passive) construction in parallel to in a focus-fronted construction: Los niños parecen haber terminado todos la tarea. the children seem.3pl have.inf finished all.m.pl the homework 'The children seem to have finished all the homework.'

b.
Las peras se ha comido todas Maria, no las manzanas. the pears refl has eaten all.f.pl Maria not the apples 'Maria has eaten all the pears, but not the apples.' Crucially, no similar construction to (ib) is reported, neither in the older grammars (OP and ClP) nor in Modern Portuguese, thus undermining the idea that Spec,IP could host one of the constituents that we consider as left-peripheral. 10 An anonymous reviewer suggests that the derivation of sentences showing SV order could be said to involve just a V-to-I movement, in line with Zwart (2001), among others. In terms of acquisitional effort, it seems equally complicated to assume that verb movement is variable. We direct the interested reader to Holmberg (2015) for further arguments against this asymmetrical account for V2.

The left periphery of Old and Classical Portuguese and the position of demonstrative pronouns used as anaphors
Following Galves (To appear), we propose that OP and ClP have a KP projection whose specifier may host a delimitator, as in the following scheme, representing the left periphery of these older stages of Portuguese: Except for KP, (14) follows Wolfe's (2015) proposal for relaxed V2 languages. 11 In particular it excludes either a topic (as in Rizzi 1997) or a second focus position (as in Benincà & Poletto 2004) under Foc. We refer the reader to Galves (To appear) for a detailed argumentation in support of (14). What is important for us here is the existence of the K head under Foc, which is largely similar to Frey's (2006) KontrP. The main evidence for it stems from object fronting, in particular the one creating an OSV sentence in which the object is not doubled by a pronoun. This is a rather infrequent order, but it appears in a strictly recurrent pattern, both in OP and in ClP: the object is quantified or focalized and the subject is interpreted as a delimitator: In the examples above, the object is in Spec,FocP and S in Spec,KP. In the generative literature on the left periphery, fronted quantified expressions are said to occur in Spec,FocP (Rizzi 1997). Besides, in the two examples the object the and subject are somehow focalized, because they express alternatives, as discussed in connection with (1). Considering co-occurrence constraints, there is no other way to derive this word order, if the verb is taken to be in Fin.
On the other hand, the structure in (14) allows us to analyze SOV constructions in two different ways: either with the subject as a topic and the object as a delimitator, as in (17), or with the subject as a topic and the object as a contrastive focus, as in (18) The structure makes available two positions for the object in SOV constructions, either in Spec,FocP or in Spec,KP, considering that these are specialized regarding their informational role.
In the change between ClP and MP, the KP projection is lost, a development which will be argued to finally imply the loss of V2. As a consequence, delimitators must now check their [contrast] feature with another projection able to do so, which we consider to be TopicP for reasons to be made clear below. In connection with this issue, consider how demonstrative pronouns occur in different syntactic contexts in ClP and MP, thus illustrating the role of KP in the former grammar. The ClP examples were obtained from de Almeida (1681) and Holanda (1984Holanda ( [1548), two texts available in the Tycho Brahe Corpus. These texts have been compared to the corresponding modern New Testament version in de Almeida (2001), and to some MP sentences either tested with native speakers or found in de Andrade (2018a).
First, consider the usage of demonstratives as pronominal anaphors. If MP is not a C-V2 grammar, we expect not to find demonstratives used as pronominal anaphors, or only very marginally so, since there is no need to distinguish between marked and unmarked pronominal forms. The replacement of the demonstrative pronoun by a personal pronoun, done by the editors of the revised New Testament version, supports this hypothesis. It is worth noting that in the remaining cases in which a demonstrative is kept in the revised New Testament version-due to the fact that its referent was newly introduced in the previous clause-it tends not to appear in a fronted position codifying contrast: Classical vs. Modern Portuguese (John 13: 24 apud de Almeida 1681; 2001) a. A este pois fez sinal Simão Pedro, que perguntasse, quem to this.m then made.3sg sign Simon Peter that asked.sbjv.3sg who era aquele de quem dizia. was that.m of whom said.3sg 'So Simon Peter motioned to him to ask (Jesus) of whom he was speaking.' b. Então Simão Pedro fez sinal a este, para que perguntasse then Simon Peter made.3sg sign to this.m for that asked.sbjv.3sg quem era aquele de quem ele falava. who was that.m of whom he spoke.3sg 'So Simon Peter motioned to him to ask (Jesus) of whom he was speaking.' In sum, these comparable examples from two versions of the New Testament show two related changes: the decrease in the usage of demonstratives as anaphors and the loss of their movement to a functional projection at the left periphery, the two trends being very likely connected to changes in local anchoring.
A possible implication is that dislocation constructions involving demonstratives should be more freely available in OP and ClP, in interaction with the complex left periphery. It is noticeable that Germanic languages, to the crucial exception of Modern English, display Contrastive Left Dislocation (CLD), a construction in which a topical constituent is resumed by a demonstrative pronoun. This type of left dislocation is generally unfound in modern Romance languages, but is present in Portuguese, along with Hanging Topic Contrastive Left Dislocation (HTCLD):

(21)
Holanda ( We consider that (21a) is a case of CLD because the topic is a PP, and both the topic and the resumed pronoun show a preposition (cf. Grohmann 2003, among others). On the other hand, (21b) would be a case of HTCLD because the topic is a DP (it could also be a DP preceded by a topic introducer, such as quanto a 'as for'), and topic and resumptive do not share the dative marker a. 12 Similar structures may be found in MP (and Galician): Modern Portuguese a. [Aos que se comportaram bem,] i a esses i dei os to.the.pl which refl behaved.3pl well to these.m gave.1sg the rebuçados. sweets 'I gave the sweets to the well-behaved (kids).'

b. [Quanto aos
alunos,] i esses i eu não culpo. regarding to.the students these.m I neg blame.1sg 'As for the students, I don't blame them.' 12 In view of the lack of morphological case in Portuguese, an anonymous reviewer pointed out that another possible difference between CLD and HTCLD is the adjacency requirement between the initial topic and the resumptive pronoun, holding only in the former (Grohmann 2003). However, in MP there is no clear example in which these elements may appear non-adjacent to the initial topic, according to the intuition tests shown in de Andrade (2018a).
This would suggest, at first glance, that there has been no change regarding these constructions between ClP and MP. However, a crucial distinction is found regarding the respective positions of demonstrative resumptive pronouns in ClP (and OP) with respect to MP, in the domain of clitic placement. In ClP, clauses with either CLD or HTCLD only show proclisis, which suggests that the demonstrative pronoun does not occupy a position which allows variation between proclisis and enclisis, as topic positions do (cf. Galves  These facts may be explained in the following way. In OP and ClP the demonstrative resumptive was hosted in KP, and the initial topic in TopP, as in (25a). By contrast, KP is not projected in MP; thus the resumptive pronoun occurs in a higher projection, Inner-TopP, leaving only the OuterTopP position available for the initial topic, as in (25b). 13 The demonstrative works as a proclisis trigger in the former due to its [contrast] feature, which is not crucially relevant for word order in MP, and consequently it does not play a role in clitic placement either.  Andrade's (2018a) suggestion that the reason why CLD is (apparently) not available in MP is that OuterTopP tends to be mapped to the outside of the clause, thus turning the initial topic a hanging topic. Irrespectively of the possibility of distinguishing between CLD and HTCLD, the fact that the resumptive does not function as a proclisis trigger is an important piece of evidence for the proposed left-peripheral change. Following de Andrade (2018a), we consider that the initial topic is base-generated in the clausal left-periphery, whereas the resumptive moves by topicalization or left dislocation (besides the demonstrative, a coreferent clitic pronoun may appear as well, optionally in MP and categorically in Galician). The interim conclusion is that contrastive elements are mapped to different positions of the clausal structure of OP and ClP. CTs are in TopP and CFs are in FocP, whereas KP-a projection that has by hypothesis disappeared in MP-is reserved to delimitators. The latter elements are usually expressed by demonstratives, whose contrastive features can be explained in terms of their pragmatic context (Light 2012). In the following section we further explore how demonstratives are connected to V2-like grammars and, consequently, to boundedness.

Contrastive items and boundedness in Old and Classical Portuguese
If OP and ClP are indeed C-V2 grammars, they should be similar regarding general characteristics, namely V-to-C movement in matrix clauses and the projection of the left-peripheral KP projection. Differences, however, are found between these two language stages, which affect demonstrative fronting, anaphoric subject expression, and the frequency of frame setters. These phenomena are directly related to the expression of boundedness because demonstratives in preverbal position and frame setters express local anchoring; and subject expression by a demonstrative pronoun, a personal pronoun or a null pronoun reflects an anaphorical system related to boundedness.
In order to study such differences in more detail, we have developed a database upon ten parsed texts, the first two from OP, available in the WOChWEL Corpus (Martins et al. 2012), and eight from ClP, available in the Tycho Brahe Corpus ). 14 These texts represent a similar corpus size (2,000~3,000 sentences in each clausal context, i.e. matrix or subordinate, for each language period) because the texts from the former period are much longer than the more recent ones. They have been organized by century according to the authors' birth dates, as shown in Table 2. 15 The data were obtained with the help of CorpusSearch 2 (Randall 2005). 16 14 Following previous work on periodization (cf. Galves et al. 2006, for a proposal and a discussion), we call Classical Portuguese the language observed in texts of authors born in the 16 th or in the 17 th century. 15 With OP texts whose authorship is unknown, we have calculated about 50 years before the publication date, following Medeiros' (2018) study on the average aging of authors. Notice that de Galhegos (1641) was included in the 17 th century because this author's birth date is too close to the end of the previous century. 16 All queries used to make the tables and figures of this paper are available in the Appendix. In face of the relevance of demonstratives for the marking of contrast and local anchoring, we have compared preverbal nominal objects with and without a demonstrative. 17 Object-verb order in OP is frequently related with the usage of demonstrative pronouns, in both matrix and subordinate contexts, as shown in (26) and (27) Let us first consider object position in matrix contexts. Tables 3 and 4 display the results of accusative and dative objects of the corpus, with and without a demonstrative. Objects with a demonstrative front in 40% of the cases in OP, but only 15% in ClP. On the other hand, objects without a demonstrative front by about 5% in both periods, showing that a change occurred specifically in sentences with demonstratives. In a nutshell, the weight of the presence of a demonstrative on object fronting is at least three times stronger in OP (10 to 1) than in ClP (3 to 1).
The results for subordinate clauses are shown in Tables 5 and 6, where it is possible to notice that the change concerning demonstratives is more abrupt. Objects with a demonstrative front by 67% in OP, but only 1% in ClP. The latter result may be considered irrelevant, once the percentage of fronted objects with and without a demonstrative in ClP is virtually the same. In this context, a decrease of about 5% is observed in preverbal objects.
It is misleading, however, to think that all types of subordinate clauses contribute equally to the high frequency of demonstratives in OP. As Table 7 reveals, almost all cases of fronted objects with a demonstrative consist of adverbial clauses.  Therefore, most occurrences of demonstratives in subordinate clauses serve to express local anchoring with impact on the information contained in matrix clauses as well.
Subject expression is the second aspect studied in our quantitative study. It relates to boundedness because the tendency to keep the perspective on the subject leads to a greater usage of non-expressed subjects (cf. Komen et al. 2014 on subject ellipsis in the history of English). However we posit that, in null-subject languages such as Portuguese, the frequency of null subjects tends to be smaller in a bounded system because of the mentioned interplay between personal pronouns and demonstrative pronouns as delimitators. For this quantitative study, only anaphoric subjects have been selected, i.e. 3 rd person personal pronouns, DPs with a demonstrative and null subjects, irrespectively of their clausal position: Lord pleases that he be.sbjv saved he 3sg.m saluara. will.save 3sg 'If Our Lord is pleased that he be saved, he will save him.'

Relative clause
Old Portuguese 90% (355) 4% (14) 6% (25) Classical Portuguese 67% (4) 33% (2) - (0) c. Bem assi como Aldrēt conselhou a el rey Mars o well likewise as Aldret counselled.3sg to the king Mars 3sg.m fez elle made.3sg he 'Aldret counselled king Mars, who followed his words likewise' The query included clauses with a nominal or a pronominal accusative object, in order to ensure comparability of the results. 18 Figure 1 shows that the frequency of clauses with null subjects is much higher than that of clauses with overt anaphoric subjects, a number that becomes even bigger in ClP, reaching 94% of the occurrences.
The higher percentage of null subjects in ClP is probably the result of a weaker interplay between demonstrative pronouns and personal pronouns in this language stage. In order to verify the overall frequency of pronominal subjects vis-à-vis the one in OP, we have complemented the previous query with one for nominal subjects, with the ensuing quantitative results presented in Table 8.
The data above shows that the change in null subjects is not to be related with a change in the frequency of demonstratives, but to a lower frequency of 3 rd person personal pronouns in subject position, which has decreased from 17% to 5%. This suggests that  pronominal subjects were more frequent in OP because in that system local anchoring was more active. Recall that both personal pronouns and demonstrative pronouns were part of an anaphoric system in which the former was used to refer to unmarked referents. A brief excursion at this point might point to a connection between these results and the distribution of null subjects in older Romance in general. According to Ingham (2018), who developed a case study on Old French, preverbal foci or null topics favor the occurrence of null subjects, whereas preverbal overt topics favor overt postverbal subjects. Among these overt topics, demonstrative pronouns tend to cooccur with postverbal subjects, an unambiguous V2 context, following his results. 19 The third and final aspect related to boundedness which will be commented on is the number of frame setters in the first position of clauses with an accusative object, by searching for clauses with PPs, AdvPs, adverbial DPs, adverbial clauses and participial clauses, compared to the total of matrix clauses with an accusative object.  Ingham's (2018) corpus study comprises only main clauses. If the data of preverbal objects in Tables 2 and 3 above-which correspond to main clauses in Old Portuguese-were reclassified into unambiguous V2 (with an overt postverbal subject) and ambiguous V2 contexts, we would have the following results (isolating cases with expletive subjects): -For Table 2, 70% (40) versus 30% (17) in Old Portuguese, and 46% (6) versus 54% (7) in Classical Portuguese; -For Table 3, 10% (5) versus 80% (41) in Old Portuguese, and 46% (14) versus 54% (16) in Classical Portuguese.
The special role of demonstrative pronouns in unambiguous V2 contexts is confirmed by a comparison between the revised numbers in the Old Portuguese period. Notice, however, that Ingham's account considers that Old French was a partial null subject language, in which subject non-expression follows the continuation of topic chains, similarly to what happens in Modern Chinese. Although the similarity is very much appealing, we do not claim that this system is applicable to Old Portuguese. At first glance, it seems important to differentiate subtypes of foci (and topics) in the latter, because overt subjects frequently cooccurred with preverbal intensifiers, usually taken to express contrastive focus. 20 The results from the automatic query were manually selected according to the semantic value of these elements, thus eliminating circumstances such as cause, goal, and instrument. e. Compostas assim as jurisdições, e os entendimentos, dirigiu composed so the jurisdictions and the understandings addressed.3sg então o Padre António Vieira a sua eloquência then the Father António Vieira the his eloquence 'Once the jurisdictions and the understandings were thus and so composed, Father António Vieira addressed his eloquence' Figure 2 shows that a decrease by 5 points in the frequency of frame setters took place between OP and ClP. This again suggests that local anchoring was less active in ClP than in OP. 21 The findings reported in this section indicate that local anchoring expressed by the use of frame setters and demonstratives is more clearly expressed in OP. The higher frequency of preverbal personal pronouns with subject function in OP suggests their role as unmarked anaphors in this language period.
The interim summary is that boundedness was more visible in OP than in ClP. The results for matrix clauses related to object position, subject expression and frame setter usage converge to this conclusion. Besides, in the literature there is some room for the idea that bounded systems may form a continuum. 22 We leave for the next section a specific analysis for the complete loss of object demonstratives in the preverbal position of subordinate clauses. A possible supplementary argument supporting the weakening in the boundedness parameter between OP and ClP is that there was a fullfledged system of demonstrative pronouns in the former, with simple and reinforced 21 χ 2 = 48.532; p = 0. The absolute values are 937 out of 6342 (Old Portuguese) and 400 out of 3982 ( Classical Portuguese). 22 Petré (2013) puts forward the existence of fully bounded and moderately bounded languages, a classification that seems to be closely related to the typology relaxed versus strict V2 languages. In this sense, both OP and ClP may be considered moderately bounded languages. forms, except for the distal paradigm. 23 This is summarized in Table 9. This system loses the proximal and medial reinforced forms by the 15 th century (Teyssier 1981). Thus in ClP demonstratives come to have only three paradigms (proximal, medial, and reinforced distal).
Probably the weakening of boundedness in ClP has later on paved the way for the loss of V2, in the transition into MP. We will come back to this issue later on.

A syntactic change in subordinate clauses from Old to Classical Portuguese
The results presented in the previous sections suggest that, although OP and ClP share features related to boundedness, these are more clearly expressed in the former than in the latter. Besides, results concerning demonstrative objects show that the change in subordinate clauses was abrupt in this context, suggesting that in this case we should consider a syntactic change instead of an information-structural change.
We would like to put forward that the change in the usage of demonstratives in subordinate clauses is directly related with the parameter regulating verb movement. While OP is a symmetrical V2 language, ClP is an asymmetrical V2 language. In other words, in OP inflected verbs moved to Fin in both matrix and subordinate clauses. This entails that in this grammar complementizers would be merged to Force instead of being merged to Fin, as it is the case in ClP (Antonelli 2011). This similarity between the left periphery in matrix and subordinate clauses is represented below for the examples (17) and (28), according to of the structure proposed in (14) 23 The corresponding simple forms to the distal paradigm became 3 rd person pronouns (ele, elo, ela). The uninflected form elo was finally lost, unlike in Spanish. While the configuration in (31b) was kept, the one in (32b) was virtually lost in ClP. 24 We propose that this is the consequence of there being a reduced left periphery in ClP subordinate clauses. Because it is an asymmetrical V2 grammar, the complementizer merges to Fin and the verb stops in I. Therefore, no other element may appear above Fin, making no room for KP to be projected in this context: One could argue against our proposal on the basis of the phenomenon dubbed as recomplementation. Once this construction involves the expression of two left-peripheral heads-Fin and Force-it is expected that such phenomenon does not exist in OP, against the facts. However, in OP recomplementation may involve more than two complementizers. This suggests that the second complementizer, que 2 , is inserted in all heads whose specifiers are expressed (González i Planas 2011):

(35)
Cintra ( In this system, the left-peripheral heads showing que stay above Fin, which is free to receive the inflected verb, as shown in (35). The analysis presented so far allows us to propose new accounts of other phenomena concerning the left periphery of OP, as well as their change on the way to ClP, which, for lack of space, we shall only mention in very broad terms. The first phenomenon concerns generalized interpolation, which is typical of subordinate clauses where the clitic pronoun is separated from the verb by one or more fronted phrases, clitic and complementizer being adjacent: (36) Cintra (1951Cintra ( [1344: 528) Bem sabedes, coyrmãão, como e porque me el-rei dom Afomso well know.2pl cousin how and why 1sg the king don Afonso deitou da terra. expelled from.the.f land 'You know well, my cousin, how and why king Afonso expelled me from the land.' Generalized interpolation virtually disappears in texts by the 16 th century (Martins 1994) and, considering the authors' year of birth, from the mid-15 th century (Namiuti 2008 The author argues that once já ('already') is a focus element, nothing should intervene between it and the verb in a V2 system, contrary to the facts. According to the proposal put forward in (14), there are at least two positions that could host constituents between a focalized adverb and the verb: Spec,KP and an adjunct We summarize the relations between grammar type, according to the V2 parameter, and the role of KP and of contrastive items in Table 10.
In sum, in this section we have shown that, although both OP and ClP display V-to-C movement in matrix clauses, no KP is projected in embedded contexts in ClP as a consequence of the complementizer merging to a lower head of the clausal left periphery (Fin), which entails shorter verb movement. This change is probably connected to the weakening of boundedness referred to in the previous section, although we leave a more detailed explanation on this connection for future work.

Conclusion
We have discussed a change in the history of Portuguese involving the notion of contrast. We put forth the idea that contrastive items include not only contrastive topics and foci, but also delimitators. Besides, we posit that KP, a projection which encodes a contrastive feature independently of topic and focus features, is an active category in C-V2 grammars, as we argue to be the case for OP and ClP. In order to make this point, we have presented evidence related to the position of objects with demonstratives, comparing ClP data with MP data.
We have also presented quantitative data showing a decrease in demonstrative object fronting, which occurs in parallel with other syntactic phenomena reflecting boundedness: an increase in the frequency of null subjects and a decrease in the usage of frame setters. These changes represent positive evidence for the weakening of the boundedness to Spec,FinP. Alternatively, if we take into account the possibility of stylistic fronting in OP-understood as vP movement to the left of FinP (de Andrade 2017)-the constituent [nos este ceruo outra vez] could be said to move to Spec,KP for independent reasons. So, we consider that Martins's argument is not yet a definitive one against V-to-C movement in OP. system found in OP. Regarding the abrupt loss of demonstrative object fronting in ClP, we have shown that it is associated with a change in verb movement, resulting in a reduced left periphery in ClP subordinate clauses. In sum, while OP is a symmetrical C-V2 language, ClP shows asymmetrical V-to-C movement, with immediate consequences for the expression of fronted demonstrative objects. If these tenets prove to be correct, they present at least two important consequences for the nature of the relation between information structure and syntactic change. The first one concerns the debate on the V2 status of Old Romance languages, and its loss in their modern counterparts: the idea that there is a correlation between V-to-C movement and frequent XV order becomes stronger. The alternative proposal, which denies the V2 characterization to OP (as well as to ClP and to Old Romance languages generally) must rely only on informational differences in order to account for the difference in word order shown by OP and ClP (and even in comparison with MP; cf. Martins 2002;Fiéis 2003;Rinke 2009, a.o.), a strategy which apparently does not explain the shifts presented in this paper.
The second consequence concerns the motivation for grammatical change. Although we cannot tell for sure whether information-structural changes lead to syntactic changes or the opposite, the results discussed in this paper suggest that the weakening of boundedness has contributed to the loss of the C-V2 system later on. However, taken that independent information-structural changes are hard to conceive, these are probably the consequence of independent functional or external pressures. 27 This text opens up a new research agenda regarding the history of Portuguese that cannot be fully explored here. Nevertheless, we believe this is a good starting point for future work combining the V2 hypothesis and the boundedness parameter in the study of Old Romance languages.