Language Structure, Variation and Change: The Case of Old Spanish Syntax 9783030105662, 3030105660

Table of contents :
Front Matter ....Pages i-xii
Preliminary Concepts: Old Spanish, How to Measure the Speed of Change and the Structure of the Corpus (Ian E. Mackenzie)....Pages 1-21
Constituent Fronting: Focus, Discourse and Fashion (Ian E. Mackenzie)....Pages 23-73
Clitic Linearization: A Tale of Successful and Failed Changes (Ian E. Mackenzie)....Pages 75-121
DP Structure: From Multiple Determiners to Just One (Ian E. Mackenzie)....Pages 123-161
The wh-System: Free Relatives, Double Articulation and Free Choice (Ian E. Mackenzie)....Pages 163-203
Negation: Dispensing with the Clutter (Ian E. Mackenzie)....Pages 205-245
Conclusion: Change and Continuity (Ian E. Mackenzie)....Pages 247-256
Back Matter ....Pages 257-291

Citation preview

Language Structure, Variation and Change The Case of Old Spanish Syntax

Ian E. Mackenzie

Language Structure, Variation and Change “In his latest book, Mackenzie marries grammatical theory and quantitative methods to make significant advances in the analysis of a range of syntactic phenomena in Old Spanish. Alongside meticulous corpus-based empirical work, he critically engages with recent concepts in the theory of quantitative historical linguistics to great effect. This book is highly recommended to anyone interested in the history of the Spanish language and indeed in language change more broadly.” —Richard Waltereit, Professor of Romance Linguistics, Humboldt University Berlin, Germany

Ian E. Mackenzie

Language Structure, Variation and Change The Case of Old Spanish Syntax

Ian E. Mackenzie School of Modern Languages Newcastle University Newcastle Upon Tyne, UK

ISBN 978-3-030-10566-2    ISBN 978-3-030-10567-9 (eBook) https://doi.org/10.1007/978-3-030-10567-9 Library of Congress Control Number: 2019931381 © The Editor(s) (if applicable) and The Author(s) 2019 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Cover illustration: © Granger Historical Picture Archive / Alamy Stock Photo This Palgrave Macmillan imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Acknowledgments

The work leading to the publication of this book was supported by a generous Research Fellowship (2017–2018) from the Leverhulme Trust.

v

Contents

1 Preliminary Concepts: Old Spanish, How to Measure the Speed of Change and the Structure of the Corpus  1 2 Constituent Fronting: Focus, Discourse and Fashion 23 3 Clitic Linearization: A Tale of Successful and Failed Changes 75 4 DP Structure: From Multiple Determiners to Just One123 5 The wh-System: Free Relatives, Double Articulation and Free Choice163 6 Negation: Dispensing with the Clutter205 7 Conclusion: Change and Continuity247

vii

viii Contents

A  ppendix257 References271 Index285

List of Figures

Fig. 1.1 Linear growth and exponential growth 11 Fig. 1.2 Advance of do-support in affirmative transitive yes/no and adverbial questions (1400–1700) 13 Fig. 1.3 Evolving odds of do-support in affirmative transitive yes/no and adverbial questions (1400–1700) 14 Fig. 1.4 Rise and fall of do-support in unemphatic affirmative declarative sentences (1390–1700) 16 Fig. 2.1 Evolving rates of constituent fronting (1250–1609) 57 Fig. 3.1 Evolving rates of proclisis (1250–1609) (excluding negative contexts)92 Fig. 3.2 Regression trendlines (parameter values: slope 0.1355, intercept −1.5416 (infinitivals); slope 0.1227, intercept −1.4589 (main clauses)) for the advance of proclisis in main clauses (1250–1609) and post-prepositional infinitival clauses (1250–1489)93 Fig. 3.3 Evolving rates of proclisis in main-clausal micro-contexts (1250–1609)95 Fig. 3.4 Observed failures (unbroken lines) compared to S × (1 − S) and other curves proportional to S’s derivative 98 Fig. 3.5 Imperative, post-pause and post-coordinator micro-contexts 101 Fig. 3.6 Differential evolution of ‘clitic–VINF’ according to context 105 Fig. 3.7 Evolving rates of finite and infinitival interpolation compared to proclisis 110 ix

x 

List of Figures

Fig. 4.1 Evolving rates of occurrence of three recursive DP structures 147 Fig. 4.2 Late medieval increase in ‘definite article + possessive + NP’ 152 Fig. 5.1 Occurrence of ‘el  +  pronominal qual/cual’ and ‘el qual/ cual + NP’ (1250–1609) 181 Fig. 5.2 Quantitative evolution of ‘D + NP + definite article + relative clause’ (1250–1609) 189 Fig. 6.1 Decline in the use of no(n) after five medieval triggers (1250–1609)221 Fig. 6.2 Rise and decline of fronted nada (1600–2000) 237

List of Tables

Table 2.1 Fronting rates per syntactic category in period 1250–1289 (with comparator data from post-1950s texts) 27 Table 2.2 Object and PP fronting in the 1250–1289 sample 42 Table 2.3 Quantified objects and PPs in the 1250–1289 sample 49 Table 2.4 Rates of increase and decline in constituent fronting 60 Table 3.1 Rate of advance of proclisis in main clauses and postprepositional infinitival clauses 93 Table 3.2 Rate of advance of proclisis in main-clausal micro-contexts (1250–1609)95 Table 4.1 Recursive possessive DPs and their competitors in Medieval Spanish148 Table 4.2 Rate of change in use of the recursive possessive DPs (1250–1419)149 Table 4.3 Rate of change in use of the recursive possessive DPs (1410–1609)150 Table 5.1 Rates of increase and decline for el/la/los/las + qual182 Table 5.2 Decline and stabilization of ‘D + NP + definite article + relative clause’ (1250–1609) 190 Table 6.1 Rate of decline in use of no(n) in five contexts (1250–1609) 230 Table 6.2 Rates of growth and decay in nada-fronting (1600–2000) 238 Table A.1 Evolving rates of fronting (1250–1609) 258 Table A.2 Evolving rates of proclisis (1250–1609) (excluding negative contexts)259 xi

xii 

List of Tables

Table A.3 Evolving rates of proclisis in main-clausal subcontexts (1250–1609)260 Table A.4 Differential evolution of ‘clitic–VINF’ according to context (1250–1609)261 Table A.5 Interpolation versus proclisis (1250–1609) 262 Table A.6 Evolving rates of occurrence of three recursive DP structures (1250–1609)263 Table A.7 Three wh-structures: tokens per million words (1250–1609) 264 Table A.8 Decline of no(n) in five contexts (1250–1609) 264 Table A.9 Evolving rates of nada-fronting (1600–2000) 265

1 Preliminary Concepts: Old Spanish, How to Measure the Speed of Change and the Structure of the Corpus

1.1 Old Spanish 1.1.1 Deconstructing the Name The term ‘Old Spanish’ is widely used in the syntactic literature and indeed the linguistic literature more generally, as a counterpoint both to ‘modern Spanish’ and also to names for other varieties of Old Romance, such as ‘Old Portuguese’, ‘Old French’ and ‘Old Catalan’. The label is a useful one and it occurs frequently in this book, together with the term ‘medieval Spanish’, with which it can be regarded as being extensionally equivalent. It should nevertheless be noted that for most of the period of its existence, what we now think of as Old Spanish would not actually have been referred to as español (or any ancestral form of this word). Indeed, as late as the thirteenth century, usage does not point to the existence of any agreed name for the language. For example, in the prose manuscripts of the Alfonsine corpus, designations range from plain romanz or romançe through to lenguage castellano, romanz de Castiella and even the modern-looking castellano.1 To the extent, then, that there was any generalized awareness on the part of speakers that they spoke a © The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_1

1

2 

I. E. Mackenzie

particular language, the relevant linguistic identity was conceptualized in terms of a Castilian way of speaking or writing. Referring to the language of the medieval kingdom of Castile as ‘Spanish’ thus involves the retrospective application of a modern nomenclature. This approach should not necessarily be regarded as being ahistorical, however. For modern diachronic syntactic theory (see e.g. Roberts 2007) envisages language change as being realized in specific areas of the grammar rather than in the language as a whole. Moreover, these localized changes, as externally manifested, are long-term processes, evolving as gradual curves (Kroch 1989) rather than in discrete stages. At any given time, therefore, the different components of a language’s grammar will be at different evolutionary moments, implying, logically, that no particular temporal slice in a language’s overall history has any preferential claim to a specific identity of its own. Thus the use of a single term for the entire, seamless continuum is actually well motivated, and for that particular role, it makes sense to employ the familiar, modern name. Dissociated in this way from any linkage to a specific historical period, the word ‘Spanish’ comes to refer not just to the language of the post-medieval Spanish state, but also to all of its previous incarnations. The secondary label ‘Old Spanish’ should accordingly be seen as a meronym, designating no more than a part or segment of the diachronic whole. This perspective immediately invites the question of what temporal boundaries should be assumed to define the relevant segment of the continuum. Approaching this issue in the first instance with a comparison, it can be noted that what linguists have in mind when they talk about Old Spanish begins rather later than the language variety that falls under the label ‘Old English’. For while the term ‘Old Spanish’ is not generally used to refer to linguistic data which significantly predate the onset of the High Middle Ages, Old English is identified as the language of the Anglo-­ Saxons, who settled in Britain from the mid-fifth century. Indeed, despite the formal similarity of their names, Old Spanish and Old English occupy different positions within their respective genealogies. In Spanish terms, the nearest counterpart to Old English would be something like late spoken Iberian Latin, i.e. what Wright (1994) refers to as early Ibero-­ Romance. In practice, Middle English is a better match for Old Spanish,

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

3

even if, as is implied by the discussion below, the notional beginning of the latter predates that of the former by about two centuries.2 As with other Romance varieties, the Spanish case is complicated by the relatively late appearance of a customized way of writing the language. In its recognizable written guise, Old Spanish effectively dates from the thirteenth century, texts prior to this period tending to be Latinate in appearance.3 At the spoken level, how far back in time what we might be prepared to think of as Old Spanish goes is anyone’s guess. Spanish philological tradition takes its cue from Menéndez Pidal’s seminal work Orígenes del español (1926), a detailed reconstruction of the early spoken language based on documents from the tenth and eleventh centuries.4 Linguistically speaking, there is no particular reason to identify those centuries as marking the emergence of a new linguistic entity. However, there is a fairly widespread assumption that they represent, in some way, the época de orígenes ‘origins period’ of the Spanish language. The more fundamental point is that there is a disjuncture between Old Spanish as manifested with full clarity by the bespoke writing system that came on stream in the thirteenth century and Old Spanish as an older but largely presumptive linguistic variety, revealed to us through the prism of a written code devised originally for the speech of many centuries earlier.

1.1.2 Syntactic Continuity While the term ‘Old Spanish’ in principle covers both of the linguistic manifestations just highlighted, the second one can be referred to more specifically as pre-literary Spanish, where ‘literary’ alludes to writing in general rather than to literature specifically. This latter form of Spanish is itself presented to us in a wide variety of guises, close approximations to Latin grammar and spelling characterizing one polar extreme and innovative experimental forms the other. In between we find texts which, to varying degrees, mix Latin words and case endings with syntactic structures, vocabulary and spelling patterns that clearly belong to Old Spanish. Menéndez Pidal’s Orígenes del español references both the experimental glosses associated with the San Millán and Santo Domingo de Silos

4 

I. E. Mackenzie

monasteries and a variety of documents embodying the hybrid text type. The latter, it has to be said, point to the existence of a spoken language whose syntax in most major areas is very similar to what we find in the linguistically more transparent manuscripts of the thirteenth century. This is hardly surprising, given that in the majority of cases Menéndez Pidal’s documents pre-date those of the early literary period by no more than two hundred years, a timeframe which realistically does not allow for any dramatic transformation of the grammar. The similarity in syntax between the presumptive spoken language of the época de orígenes and the well-attested Spanish of the thirteenth century is illustrated in the extract in (1) below, which is taken from a document originating in the Palencia area and written, according to Menéndez Pidal, in 1097.5 (1) Et si ego mici mortem ante uobis uenerit, si de mea ereditate comodo et demeo ganato, aueatis uos jlas duas partes, et jla tertja, siue de ereditate comodo et deganato, jntre promea anima asancti Zoili. Et siuobis uiro meo aut germanis meis jla tertja quesieritis recolere, aprecient jla quantum ualere, et date precio pro jlas duas partes, et jla tertia lexola por amor de Dios; (Archivo Histórico Nacional de Madrid, San Zoil de Carrión P–7; Menéndez Pidal 1926: 35) ‘And if death comes to me before you, of my land as with my cattle, you shall have two parts, and the third, be it land or cattle, goes for my soul to San Zoil. And if you my husband or my brothers want to recover the third, its value should be established, then pay a price for two thirds of it, but the remainder I give for free.’

If one looks past the Latin interference, such as the anachronistic dative forms mici (Classical Latin: mihi) and uobis or outmoded spellings like aut, pro and comodo (Classical Latin: quomodo), the language in the extract should seem very familiar to anyone versed in the grammar of the post-1200 period. Rather obviously, the consistent use of jla and jlas to introduce noun phrases points to a fully operational definite article and the linear sequences aueatis . . . jlas duas partes ‘shall have (the) two parts’ and date precio ‘give a price’ indicate a predominantly VO word order. In addition, aueatis uos ‘you shall have’ appears to be an instance of the common medieval VS(O) pattern that is discussed here in Chap. 2, while jla tertja quesieritis recolere ‘the third [you] want to recover’ can be analysed

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

5

as discourse-driven constituent fronting, a wide-ranging and emblematic feature of Old Spanish which forms the principle subject matter of Chap. 2. As regards weak pronoun placement, the enclitic structures aprecient jla ‘they/one should weigh it up’ and jla tertia lexola ‘the third, I leave it’ are fully consistent with the Tobler–Mussafia system, which is commonly discussed in relation to thirteenth and fourteenth century Spanish (see Sect. 3.2.2 in this book). More specifically, aprecient jla, which immediately follows an adverbial clause, illustrates the inhibiting effect of a clausal boundary on proclisis, while jla tertia lexola involves Clitic Left Dislocation, which is known to correlate with pronominal enclisis in Tobler–Mussafia languages (see Benincà 2006; also Sect. 2.3.1.2 [latter stages] in this book). Finally, if ualere is assumed to be an early form of finite valiere ‘it is worth’, the jla quantum ualere component of the structure aprecient jla quantum ualere is no more than a pronominal counterpart to thirteenth century interrogative formulations of the kind discussed in Sect. 5.3.1.2, i.e. formulations like la caça qual es as it occurs in the example below (= (41) from Chap. 5): (2) E despues daquesto quando la sacaren a caça conuiene que uea el caçador antes la caça qual es. ca mas uale que la ecchen ante a la mayor caça que a la menor; (Libro de las animalias que cazan, fol. 62v) ‘And after this, when it is taken hunting, the huntsman should first of all check what is being hunted, for it is best to set [the bird] after larger game before smaller animals.’

As these brief comments on the language of (1) indicate, when texts from the pre-literary period exhibit a syntactic pattern that is also attested in the well-codified Old Spanish of the literary period, one can be fairly confident that we are dealing with the same phenomenon in both cases. In the converse situation, i.e. when an “expected” Spanish feature is missing in a Latinate text, the analysis is not necessarily as straightforward. In this type of case, it may in practice be impossible to say with certainty whether the scribe is merely applying a Latin rule, learned as part of his training, or whether the structure in question was actually subject to a degree of variation. An instance of this problem arises in connection with negation patterns in the Latinate section (dispositions I to CIX) of the Fuero de Madrid. This is a particularly interesting case, as the relevant part

6 

I. E. Mackenzie

of the Fuero’s manuscript dates from 1202, making it roughly contemporaneous with the emergence of literary Spanish. Now Camus Bergareche (1986) posits that preverbal negative quantifiers (other than nunca) were uniformly accompanied in medieval Spanish by the negative morpheme non until about the middle of the fifteenth century. Accordingly, with respect to examples like (3) and (4) below, he suggests (ibid. p. 113) that ‘probablemente la oración no deba considerarse propiamente castellana sino latina’6 and hence that the examples should not be treated as exceptions to the proposed general rule. (3) Nullus respondeat sine rancuroso (Fuero de Madrid, XXXII) ‘Nobody should be charged if no complaint is made.’ (4) [. . .] nichil pectet (Fuero de Madrid, LXV) ‘. . . he pays nothing.’

On the face of it, this analysis seems to be the obvious one to advance. A prominent feature of Latin is that its negative quantifiers nullus ‘no/ nobody’, nihil ‘nothing’, nemo ‘nobody’ etc. behaved exactly like their equivalents in English; that is to say, they could not construe with the clausal negator non (except to express a genuine double negation, amounting to an affirmative assertion), regardless of whether they were preverbal or postverbal. Examples (3) and (4) thus appear to reflect no more than the scribe’s disciplined adherence to a salient principle of Latin syntax. However, looking at the Fuero de Madrid overall, we find that rather than being categorically excluded, the co-occurrence of non with Latin negative quantifiers is actually variable. Thus in addition to examples like (3) and (4), we find other examples, such as (5) and (6) below, where the negation-related syntax is consistent with the known patterns of Old Spanish (see Sect. 6.3 in this book): (5) Nullus non pignoret qui uenerit cum mercadura. Todo el omne qui ad Madrid uenerit in requa, & alguna cosa adduxerit ad Madrit, nullus homo non pendret ei, et qui lo pendraret, pectet II morabetinos a los iurados del rei et tornet la pendra sene fiadura. (Fuero de Madrid, LXIV)

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

7

‘Nobody from outside may bring goods to pawn. In respect of any man who comes to Madrid with a mule caravan, bringing goods to Madrid, no man may take collateral from him, and anyone who does must pay two maravedis to the king’s officers and the loan is invalidated.’ (6) Toto homine de Madrid qui messare aut firieret uel mataret pastor aut bacherizo [. . .] & pignos noluerit dare cum bonas testemunas, non pectet nullam calumpnia, nisi calumpnia regi (Fuero de Madrid, XXII) ‘Any man from Madrid who assaults, wounds or kills a shepherd or cowhand and fails to provide character witnesses pays no fine, except to the king’

Interestingly, the use of non is categorical in the specific case illustrated by (6), i.e. when the negative quantifier is postverbal. In other words, the locus of the variation is the preverbal field, which is exactly the context in which the use of no(n) with negative quantifiers has been diachronically unstable, both in Spanish and Romance more generally (see Sect. 6.4). In light of the overall pattern of non-usage, then, one cannot be certain that the non-less syntax of examples like (3) and (4) is genuinely the product of Latin interference or whether, in actual fact, it should not be taken as evidence of early variation in this particular area of the spoken grammar. Such variation would not be entirely unexpected, given that overtly Old Spanish examples like (7) below, from a mid-thirteenth century manuscript, suggest that the non-less structure which came into its own in the fifteenth century may well have been available, if not widely used, at a significantly earlier time (in this connection, see Fig. 6.1 in Chap. 6). (7) E ninguno deue tomar della otra cosa; fueras aquello quel es otorgado por el derecho de santa eglesia. (Libro de las leyes, fol. 90r) ‘And nobody should take from it anything other than that to which they are entitled by the law of the holy church.’

As the discussion indicates, we are not in this this type of case talking about the wholesale absence of an expected Old Spanish feature, but its partial absence. The problem, therefore, is largely a methodological one;

8 

I. E. Mackenzie

namely, are the examples without the expected feature simply cases in which the scribe has remembered the relevant Latin rule, or might they be thought of as early tokens of a pattern that is independently known to have emerged? A case with potentially wider-ranging implications relates to the use of the definite article with prenominal possessives, as in el so nieto ‘his grandson’, a structure robustly attested from the thirteenth century onwards and, it is argued here in Chaps. 4 and 5, just one manifestation of a more general syntactic paradigm. As (8) below shows, the structure is attested in the famous eleventh century San Millán glosses: (8) Facanos Deus omnipotes tal serbitjo fere que denante ela sua face gaudioso segamos. (Códice Emilianense 60, fol. 70r; cited Menéndez Pidal 1926: 8) ‘May almighty God make us perform such service that before his face we are joyful.’

However, Batllori (2010: 420) finds that the structure is ‘prácticamente ausente’ in texts from the pre-literary period and Menéndez Pidal himself does not appear to have regarded it as belonging to the linguistic landscape of the época de orígenes, given that he makes no mention of it in his survey of the early possessive (ibid. § 67). One could infer from this that examples like (8) are not representative in this regard and that the use of the definite article with prenominal possessives reflects a type of syntax that did not acquire a detectable quantitative footprint until the latter stages of the High Middle Ages. That is of course perfectly possible. However, the quantitative data adduced here in Chap. 4 point to a rate of occurrence in the mid-thirteenth century of close to 20%, a level of productivity which remained approximately stable for the next two hundred years. This is suggestive of a structure that was already quite well embedded at the outset of the literary period. Viewed in that light, its apparently negligible presence in the (predominantly Latinate) documentation from before that time is surprising, particularly given its attestation in the early Romance gloss shown in (8). It is not implausible, perhaps, to link this state of affairs to the relatively small number of ‘possessive + noun’ combinations that in practice

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

9

occur in the documents of the pre-literary period. In a text type embodying the discourse of property transfers and municipal charters to the exclusion of almost everything else, possessive noun phrases such as sua parte ‘his/her part’, suo termino ‘its limit’ suos molinos ‘its mills’, suas sernas ‘its arable land’ etc. may well have become frozen in their article-less form at an early date, thereby developing an unusually high degree of resistance to the pressure of the spoken language. And it is of course worth remembering that the ‘definite article + possessive + noun’ structure was not a majority option even in the thirteenth century (or at any time afterwards). Presumably, therefore, any impulse felt by the scribes of the pre-­literary period to prefix the possessive determiner with the definite article would not have been a strong one. This particular combination of circumstances might conceivably have prevented a structure that one could reasonably expect to have been available in speech, if only at “background” levels of frequency, from appearing in the textual record. Notwithstanding this latter puzzle, one can be fairly confident that the syntax of pre-literary Spanish, certainly from the época de orígenes onwards, was fundamentally similar to the syntax which is revealed in the prose manuscripts of the thirteenth century, which marks the beginning of the time window with which this book is primarily concerned. The right-hand boundary of this window, assuming a left-toright visualization, is located in the early seventeenth century, a time that in conventional terms falls outside the Old Spanish period, which is usually assumed to come to an end in the fifteenth century (cf. Wanner 1991: 349). However, the majority of the changes which notionally represent the transition from Old Spanish to Modern Spanish are not actually complete by the end of the Middle Ages. Indeed, even the relatively late upper time boundary chosen here cuts off some of the changes before their diachronic curve has fully arrived at its saturation point, although the direction of travel is by that time usually pretty clear. This only reinforces the point implicitly made in Sect. 1.1.1, and explicitly made by Wright (1999), that linguistic periodization is ultimately self-defeating.

10 

I. E. Mackenzie

1.2 Measuring the Speed of Change 1.2.1 Constant Change One of the most significant findings in diachronic quantitative linguistics in recent decades is that syntactic change, as externally manifested, tends to evolve at the same speed across all of the affected contexts. That this was the case had long been entertained as a possibility, but Kroch (1989) was the first linguist to demonstrate this empirically. His statement of what is now called the Constant Rate Effect is given in (9) below: Constant Rate Effect (CRE) (9) When one grammatical option replaces another with which it is in competition across a set of linguistic contexts, the rate of replacement, properly measured, is the same in all of them. (Kroch 1989: 200)

Quite apart from being an interesting phenomenon in its own right, the CRE is a very useful tool for analysing groups of changes in a language. Searching for a CRE in any given case can often help to determine whether two or more changes are entirely separate events or whether, in contrast, they are manifestations of a single, more general change. Resolving this latter type of question is in turn fundamental to attaining a more structural or systemic perspective on a language’s evolution. The ability to measure the rate or speed of syntactic change is, therefore, an important component of the diachronic quantitative linguist’s toolkit. Quantitative change can obviously involve either an increase or a decrease, the first case being known as growth and the second as decay. Because linguists usually think in terms of one variant replacing another, discussions of syntactic change are typically about growth (in the frequency of the “incoming” structure). Clearly, though, structures may decline in frequency and in that case we need to measure the rate of decay. Growth and decay are measured in exactly the same way, but the interpretation of decay rates requires an additional consideration, whose discussion I will postpone until Sect. 1.2.3. As regards growth, Fig. 1.1

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

11

20 Linear growth 15

Exponential growth

10

5

0

0

1

2

3

4

5 Year

6

7

8

9

10

Fig. 1.1  Linear growth and exponential growth

illustrates what are probably the two commonest kinds, viz. linear growth and exponential growth. Linear growth is additive, in the sense that for each unit increase in the independent variable, a specific number of units is added to the value of the dependent variable. Thus the grey line in the figure represents a situation in which every year the value of the quantity shown on the vertical axis increases by two units: in year 1, the value is 2; in year 2, it is 4; in year 3, it is 6 etc. This type of growth, when represented in a graph or chart, corresponds to a straight line. In contrast, exponential growth is proportional, in the sense that each new value of the independent variable is proportional to the previous one. In Fig. 1.1 the exponential curve corresponds to a situation in which, each year, the value of the dependent variable increases by 70% with respect to the previous year. Expressing this in terms of a ratio, we can say that the value of the dependent variable each year is 1.7 times higher than the previous year’s value. As can be seen from the figure, this growth model delivers an approximately J-shaped curve. Now in both of the cases just considered, the growth rate (the rate of increase) can be said to be constant, in the sense that it can be characterized

12 

I. E. Mackenzie

in terms of a single numerical value. In respect of the grey straight line we can say that the growth rate is two units per year and as regards the black J-curve we can say that the growth rate is 70% per year (equating to a yearly growth ratio of 1.7). If syntactic changes were either linear or exponential, we could simply use one of these two measures to characterize the evolution of any given structure. However, as is well-known, syntactic changes (and linguistic changes generally) tend to describe an S-curve when their evolution is plotted on a graph. By definition, a sigmoidal curve cannot express a constant rate of change. In order, then, to bring syntactic change within the scope of the model implied by Kroch’s Constant Rate Effect, we need to conceptualize it in a different way from that implied by the usual graphical visualizations.

1.2.2 Logistic Growth and the Odds Ratio The problem can be illustrated using data from Ellegård’s classic 1953 study of the quantitative evolution of do-support in English, the process whereby simple verb–subject inversion (e.g. Knowest thou the ordinances of heaven? [Job 38:33]) in questions and negations was replaced by the modern periphrastic structure involving do. Figure 1.2 shows the curve corresponding to the advance of do-support in the context defined by Ellegård as including affirmative transitive questions that are either of the yes/no variety or involve an adverbial wh-word such as when or where. As can be seen, both from the unbroken line corresponding to Ellegård’s data and the dashed trendline, the growth in the percentage rate or (absolute) probability of do-support is neither linear nor properly exponential. Rather, it has an S-shaped trajectory, which diachronic linguists now usually identify with the logistic curve originally used by Pierre Verhulst to model population growth (see Altmann et al. 1983; Kroch 1989; Yang 2000; Kauhanen and Walkden 2017). Logistic growth is characterized by the existence of a natural terminus or saturation point, this being, in Fig. 1.2, the maximum percentage value of 100%. From an exponential perspective, the growth rate declines gradually over time, causing the curve to evolve from a convex shape to a concave one, the inflection point or change in curvature being located at the curve’s midpoint (around

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

13

100%

80%

60%

Ellegård’s data Trendline

40%

20%

0% 1412

1437

1462

1487

1512

1537

1562

1587

1612

1637

1662

Fig. 1.2  Advance of do-support in affirmative transitive yes/no and adverbial questions (1400–1700)

1540 in the present case). In linear terms, the growth rate increases up to the inflection point and then starts to decline. Viewed from either perspective, the growth rate in the absolute probability is constantly changing. In other words, if we use the percentage scale to quantify the evolving frequency of the incoming structure – as is completely normal – there is no single numerical value that will capture the rate of change. To resolve this problem, the absolute probability values need to be converted into the equivalent relative probability values, also known as odds. The odds are a measure of probability, but unlike absolute probability, which is calculated with reference to all possible outcomes, odds are calculated in relation to the likelihood of the relevant event not happening. For example, abstracting away from such matters as quality of squad and managerial skill, the absolute probability of any given team winning the English Premier League is 1/20 or 0.05, whereas the odds of this event happening are 1/19 (or 0.053 to three decimal places), which in betting parlance would be expressed as ‘nineteen to one against’.7 Absolute probability can be converted into odds using the equation below:

14 

I. E. Mackenzie

Odds =

probability 1 − probability

A useful property of odds is that they have no upper limit: they approach infinity as absolute probability approaches 1. Better still, logistic growth in the absolute probability equates to exponential growth in the odds.8 Accordingly, if we re-express Fig. 1.2 using the odds scale, we end up with the exponential pattern of evolution shown in Fig. 1.3. Viewed in this way, the advance of do-support in the relevant context does have a constant rate. Specifically, it can be estimated by logistic regression analysis that the odds of do-support in this context increase by 28.9%, or a factor of 1.289, each decade. The factor increase of 1.289 is also known as an odds ratio (OR), a concept which is of fundamental importance in many areas of statistical analysis. It should be noted that, rather than the OR itself, Kroch (1989) and his followers use the natural logarithm of the odds ratio, known as the 20 18 16 14

Odds

12

Ellegård's data Trendline

10 8 6 4 2 0 1412

1437

1462

1487

1512

1537

1562

1587

1612

1637

1662

Fig. 1.3  Evolving odds of do-support in affirmative transitive yes/no and adverbial questions (1400–1700)

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

15

slope, as their measure of the growth rate. Here, for example, the slope would be ln(1.289)  =  0.254, where ‘ln’ means ‘natural logarithm of ’.9 While the slope is referenced in one or two places in this book, in general the more intuitive measure of the odds ratio is used to capture rates of change. This does not mean that synchronic rates of usage need to be expressed as odds, in the manner of Fig. 1.3, rather than as percentages (Fig. 1.2). The concept of the odds is only needed to enable a value to be assigned to the rate of change; for general purposes, including diachronic visualizations, we can use the familiar percentage scale. In reality, odds and percentages are just different ways of presenting the same underlying data, and it is to the latter that the logistic regression procedure is applied.10

1.2.3 Quantitative Decay and Failed Changes If an odds ratio in excess of 1 implies growth, an OR of less than 1 indicates decay (and an OR of exactly 1 implies no change at all). For example, if the decadal OR for structure X in context Y is 0.74, this means that the odds of X in context Y decay at the rate of 26% per decade, where 26% is simply 1–0.74, expressed as a percentage. One should be careful, however, not to fall into the trap of assuming that the downward curve delivered by an OR of 0.74 is the mirror image of the upward curve delivered by an OR of 1.26 (the latter implying growth of 26% per unit of time). This would be analogous to thinking that a 50% increase in an investment of £100 would be exactly reversed by a subsequent decrease of 50%. Given that the value of the investment after the increase is £150, a 50% decrease would leave the investor with only £75 pounds. For the investment to revert to its initial value, £50 must be lost from its post-­ increase value, implying a decline of one third or 33.3%. As a measure of decay, this latter figure equates to a “growth” ratio of 0.667, a value which is simply the reciprocal of the growth ratio of 1.5 corresponding to the 50% increase; that is to say, 0.667  =  1/1.5 (to three decimal places). Exactly the same principle holds for decay in language change. For example, if we have an upward curve with an OR of 1.25, implying growth of 25% per time unit, the equivalent downward curve must have an OR of 1/1.25 = 0.8, which is equivalent to a decay rate of 20% per time unit.

16 

I. E. Mackenzie

Being able to determine whether a decaying OR matches a previous upward-trending OR is useful for evaluating up-and-down quantitative events, called failed changes by Postma (2010). In the context of unemphatic declarative sentences, for example, do-support initially increased in frequency (cf. examples like Me thinke I doe heare a good manerly Begger at the doore, cited by Kroch 1989: 229) only to decline later and eventually disappear. Figure 1.4, again based on Ellegård’s data, illustrates the bell curve which events of this type characteristically describe. How to analyse this type of phenomenon is far from settled at present. However, one obvious question that can be asked is whether the growth rate represented by the left-hand flank of the bell and the decay rate corresponding to the right-hand flank are (approximately) equal or not. Symmetry in this regard might imply that the curve overall represents a single, ‘inherently failing’ change, as Postma (2010, 2017) has proposed. Conversely, asymmetry might suggest that the rise and decline correspond to separate changes, implying that the failure is ‘accidental’ rather than predetermined. Either way, assuming that change broadly corresponds to 20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0% 1395

1425

1455

1485

1515

1545

1575

1605

1635

1665

Fig. 1.4  Rise and fall of do-support in unemphatic affirmative declarative sentences (1390–1700)

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

17

exponential growth or decay in the odds, the OR provides a useful way of quantifying the speed of change on each side of the peak.

1.3 Overview of the Corpus The main corpus for the present study is constructed from Castilian prose texts covering the period 1250–1609, with an overall size of 14,340,000 words. Tokens of the structures being investigated are aggregated at twenty-year intervals, meaning that there are eighteen data points, the first covering the period 1250–1269 (identified as 1260) and the final one covering the period 1590–1609 (identified as 1600). This fine-­ grained segmentation of the data enables relatively precise quantitative curves to be constructed, which in turn assists with hypotheses as to how the syntactic changes were related to one another. Manuscript texts have been assigned to data points on the basis of their date of copy, using the bio-bibliographical database PhiloBiblon, hosted at http://bancroft. berkeley.edu/philobiblon/. In a number of cases, a given manuscript is assigned to two or more adjacent data points. In such cases the relevant data are aggregated on a weighted basis, e.g. 50% to each data point if assigned to two data points or 33.33% if assigned to three. The post-­ medieval printed texts are assigned to data points on the basis of their initial publication date. The texts in the first thirteen data points (1260–1500) have all been surveyed in their manuscript or incunabulum form, using the electronic transcriptions provided by the following outputs of the Hispanic Seminary of Medieval Studies (full details of which are given in the bibliography): 1. Prose works of Alfonso X el sabio. Digital library of Old Spanish texts. (Gago Jover 2011a) 2. Spanish legal texts. Digital library of Old Spanish texts. (Gago Jover 2011b) 3. Spanish medical texts. Digital library of Old Spanish texts. (Gago Jover 2011c)

18 

I. E. Mackenzie

4. Spanish chronicle texts. Digital library of Old Spanish texts. (Gago Jover 2011d) 5. Electronic texts and concordances of the Madison Corpus of early Spanish manuscripts and printings (O’Neill 1999) 6. Textos y concordancias electrónicos de documentos castellanos de Alfonso X (Herrera et al. 1999) The specific texts drawn from the above sources that are included in the corpus are listed in Appendix 2, with manuscript details given in parentheses after each entry. The majority of the texts in the last five data points (1520–1600) have been surveyed using the electronic editions provided by the Biblioteca virtual Miguel de Cervantes (hosted by the University of Alicante at http:// www.cervantesvirtual.com/). These editions range from direct transcriptions of the relevant princeps edition to machine-readable versions of modern scholarly editions. In selecting texts for the corpus from this source, care has been taken to ensure that any editorial intervention is limited at most to updating the orthography. As with the medieval texts referred to above, the specific early modern texts included in the corpus are listed in Appendix 2, together with details of the printed editions on which the relevant electronic editions are based.

Notes 1. According to Wright (2013: 31), the word romanz/romançe and its cognates were originally applied to (non-Latinate) ways of writing rather than to actual speech. Presumably, language in a more general sense would have been referred to using terms like lenguage, lengua or fabla ‘speech’. 2. In contrast, in the French domain, the ‘Middle’ instantiation of the language is relatively modern, being essentially the variety of French which post-dates the Old French period but is not yet the Classical French associated with the likes of Racine, Molière and Vaugelas. 3. The Poema de mio Cid, the first major literary work written in Old Spanish, is often presented as being a twelfth century text. If it is, it is likely to be a late twelfth century text. The sole surviving manuscript is actually from the fourteenth century, but folio 74r. states that ‘Per abbat

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

19

le escriuio enel mes de mayo En era de mill & .C.C xL.v. annos’ (‘Abbot Per wrote it down in the month of May, era of Caesar 1245 [i.e. 1207 AD]’). Modulo its poetic style, the syntax of the Poema de mio Cid is not appreciably different from what one finds in prose texts from the second half of the thirteenth century. 4. It should be noted that the majority of the manuscripts transcribed in the Orígenes del español are not technically from the Castile area. Assuming that Spanish is the Romance variety associated with Castile, it is not certain that early manuscripts from Aragon, say, can be regarded as embodying an embryonic form of Spanish. Pountain (2001: 19) makes an analogous point with respect to the San Millán glosses, which are from the Rioja area. On the other hand, Wright (2013: 32) advocates the exercise of latitude when applying the label ‘Spanish’ to the language of that period. 5. See (51) and (52) in Sect. 2.4.1 for additional illustrations. 6. ‘Probably the clause should not be seen as being a genuinely Spanish one but rather a Latin one.’ 7. It should be noted that the odds quoted by a professional bookmaker are, first and foremost, a conditional financial commitment rather than a disinterested estimate of relative probability. 8. Growth in the odds, in the statistical sense, equates in principle to a ‘shortening’ of the odds in everyday usage. For example, if the odds shorten from three-to-one to two-to-one, they actually increase from 1/3 or 0.333 (to three decimal places) to 1/2 or 0.5. 9. The decadal value given in the text equates to a slope of 2.54 on a century basis. Alert readers will note that this differs from Kroch’s own (century-based) estimate of the slope for the advance of do-support in this context, which is 3.62 (Kroch 1989: 225). The reason for the discrepancy is that Kroch’s estimate is based on just the first seven of Ellegård’s data points, rather than all eleven as here. 10. All the logistic regression estimates reported in this book employ maximum likelihood estimation, which is standardly available in statistical and mathematical software packages.

References Altmann, Gabriel, Haro Buttlar, Walter Rott, and Udo Strauss. 1983. A law of change in language. In Historical linguistics, ed. B.  Brainerd, 104–115. Bochum: Studienverlag Dr. N. Brockmeyer.

20 

I. E. Mackenzie

Batllori, Montserrat. 2010. La periferia izquierda del sintagma nominal: artículo ante posesivo en español medieval. In Actes du XXVe congrès international de linguistique et de philologie romanes, ed. Maria Iliescu et al., 419–430. Berlin: Walter de Gruyter. Benincà, Paola. 2006. A detailed map of the left periphery in medieval Romance. In Crosslinguistic research in syntax and semantics: Negation, tense and clausal architecture, ed. Raffaella Zanuttini et  al., 53–86. Washington, DC: Georgetown University Press. Camus Bergareche, Bruno. 1986. Cronología y extensión de un cambio en la expresión de la negación en español. Revista de Filología de la Universidad de La Laguna 5: 111–122. Gago Jover, Francisco. 2011a. Prose works of Alfonso X el sabio. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www. hispanicseminary.org/t&c/ac/index-en.htm. Last consulted: October 2018. ———. 2011b. Spanish legal texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/lex/index-en.htm. Last consulted: October 2018. ———. 2011c. Spanish medical texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/med/index-en.htm. Last consulted: October 2018. ———. 2011d. Spanish chronicle texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/cro/index-en.htm. Last consulted: October 2018. Herrera, María Teresa, María Nieves Sánchez, María Estela González de Fauve, and María Purificación Zabía. 1999. Textos y concordancias electrónicos de documentos castellanos de Alfonso X. CD-ROM. Madison: Hispanic Seminary of Medieval Studies. Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Menéndez Pidal, Ramón. 1926. Orígenes del español, 6th ed. Madrid: Espasa-Calpe. O’Neill, John. 1999. Electronic texts and concordances of the Madison Corpus of early Spanish manuscripts and printings. CD-ROM.  Madison: Hispanic Seminary of Medieval Studies.

  Preliminary Concepts: Old Spanish, How to Measure the Speed… 

21

Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. ———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Pountain, Christopher J. 2001. A history of the Spanish language through texts. London: Routledge. Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Wanner, Dieter. 1991. The Tobler–Mussafia law in Old Spanish. In Current studies in Spanish linguistics, ed. Héctor Campos and Fernando Martínez Gil, 313–378. Washington, DC: Georgetown University Press. Wright, Roger. 1994. Early Ibero-Romance. Twenty-one studies on language and texts from the Iberian Peninsula between the Roman Empire and the thirteenth century. Newark: Juan de la Cuesta. ———. 1999. Periodization and how to avoid it. In Essays in Hispanic linguistics dedicated to Paul M. Lloyd, ed. Robert J. Blake et al., 25–41. Newark: Juan de la Cuesta. ———. 2013. The prehistory of written Spanish and the thirteenth-century nationalist zeitgeist. In A political history of Spanish. The making of a language, ed. José del Valle, 31–43. Cambridge: Cambridge University Press. Yang, Charles D. 2000. Internal and external forces in language change. Language Variation and Change 12: 231–250.

2 Constituent Fronting: Focus, Discourse and Fashion

2.1 Introduction One of the most characteristic aspects of the Old Spanish syntax is the propensity for constituents that might normally be expected to occur post-verbally to be displaced or fronted to the left of the finite verb. This tendency was originally thought to be a remnant of the verb-final word order that is usually assumed to have been the unmarked pattern in Latin (see e.g. Lapesa 1981: 217; Menéndez Pidal 1926: 397–398; Alonso 1962: 81). That approach was robustly criticized by Martínez Gil (1989), who argued that many of the relevant claims were based on the study of verse texts such as the Poema del Cid and therefore had little direct applicability to linguistic usage at the vernacular level. In fact, owing to the relative ease with which digital versions of legal, medical and other prose texts from the medieval period can now be searched, it can be verified that the fronting of items such as the direct object, predicative adjectives, prepositional complements, adjuncts and infinitives was an integral part of Old Spanish and was not confined to poetry. On the other hand, the view that such structures represent a continuation of an older, verb-final word order does indeed appear to be misplaced. A variety © The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_2

23

24 

I. E. Mackenzie

of studies (e.g. England 1980; Rivero 1991; Fontana 1993; Danford 2002; Benincà 2006; Fernández Ordóñez 2008–2009; Cruschina and Sitaridou 2009; Sitaridou 2011; Mackenzie and van der  Wurff 2012; Fischer 2014) have instead converged on the view that the structures in question involve pragmatically motivated constituent reordering, of the kind that is now familiar from studies of the clausal left periphery in a wide variety of different languages. The present chapter analyses the medieval fronting structures from both a syntactic and discourse-pragmatic viewpoint, before examining their quantitative evolution from the Alfonsine period to the outset of the modern one. The findings and data adduced cast an interesting light on the typological status of Old Spanish and also call into doubt the notion, widely assumed in the literature, that the Spanish language has undergone significant parametric change in terms of its word order. It turns out, furthermore, that rather than declining steadily from the High Middle Ages onwards, fronting enjoyed an extended but transient increase at the end of the Middle Ages, with the result that early modern Spanish has, in this particular regard, a more ‘medieval’ character than the Spanish of the 1200s.

2.2 Descriptive Overview The principal fronting patterns are illustrated in examples (1a) to (5b) below (all from thirteenth-century manuscripts), the fronted constituent being highlighted in bold. For each syntactic category, the (a) example illustrates the construction in a root clause and the (b) example gives an illustration from an embedded clause. It is important to note from the outset (pace Wolfe 2015) that the relevant fronting phenomena are not confined to root environments, in contrast to analogous phenomena in languages like modern German or Old English, which are said to be subject to a V2 (‘verb-second’) word ordering constraint. In the examples below, the normal or unmarked position of the constituent in bold would be to the right of the main verb, both in Old Spanish and in Modern Spanish.

  Constituent Fronting: Focus, Discourse and Fashion 

25

1. Fronted object1: (1a) Et esta semeiança posieron al can destos menores. assi como a carauo. o de otros perriellos mas pequennos que allegan los ombres assi mas que los mayores (Libros del saber de astronomía, fol. 9v) ‘And they made this comparison to the small type of dog, such as a little hunting dog or other smaller dogs that such men talk about more than bigger ones.’ (1b) salieral a recebir contral mont. segund el tiempo que el sabie que uernie. ante que ellos el uezerro fiziessen. (General estoria I, fol. 213r) ‘He had gone out to meet him next to the Mount, at the time he expected him to return, before they made the (golden) calf.’

2. Fronted prepositional phrase (PP)2 (2a) Los que fazen peccados. ¶ de muchas maneras deuen seer penados. (Fuero Juzgo, fol. 55v) ‘Those who commit sins are subject to many punishments.’ (2b) Et si con esta agua lauaren pannos en que aya oro de que los no puedan alimpiar dotra guisa; laua los muy bien que no pierden niguna cosa de su fremosura. (Lapidario, fol. 86r) ‘And if with this water you wash cloths stained with gold which cannot be cleaned in another way, it washes them well and they lose nothing of their beauty.’

3. Fronted manner/quantity adverb: (3a) Ligera miente quiebran; & de color son muy blancas. (Lapidario, fol. 72v) ‘They break easily and they are very white in colour.’ (3b) & las que mucho salieren dela regla derecha; aquellas podremos nombrar enfermedades quando fuere sallido de toda ordenacion de natura. (Judizios de las estrellas, fol. 183v) ‘And those which depart a lot from the correct rule, these we can call illnesses in cases where the natural order is completely abandoned.’

26 

I. E. Mackenzie

4. Fronted predicative adjective/noun: (4a) ca buena es la tardança que faz la carrera segura. (Estoria de España I, fol. 28r) ‘For tardiness is good if it makes your path safe.’ (4b) Mas si costumbre fuesse que el padron estudiesse delante quando la eleccion fiziessen los clerigos [. . .] (Libro de las leyes, fol. 89v) ‘But if it was the custom that the patron was present when the clerics held the election . . .’

5. Fronted non-finite verb form: (5a) con todo esto guardar deuen los prelados que lo no reciban dellos con soberuia. ni con desden. (Libro de las leyes, fol. 114r) ‘Nevertheless, the prelates must ensure that they do not receive it from them with arrogance or disdain.’ (5b) & demas sobre todo. que podra acabar loque quisiere que es la mas preciada cosa que seer pueda. (Picatrix, fol. 25r) ‘And above all, he will be able to achieve whatever he wishes, which is the most precious thing there is.’

Structures such as the above – with the exclusion of (5a) and (5b) – also occur in modern Spanish, as is illustrated by (6) to (9) below. (6) Pero en el par de semanas algunos pequeños episodios le ocurrieron, algunas experiencias tuvo, algo le sucedió que lo obligaría a volver a la casa de la calle French. (Marco Denevi, Manuel de historia, Buenos Aires 1985) ‘But over the course of the two weeks, certain small episodes happened to him, he had certain experiences, something happened to him that made him return to the house in French Street.’

  Constituent Fronting: Focus, Discourse and Fashion 

27

(7) Porque de estos elementos está hecho el mundo de Alfredo González. (Manuel Peña Muñoz, Ayer soñé con Valparaíso, Santiago de Chile 1999) ‘Because these are the things of which the world of Alfredo González is made.’ (8) La comodidad es la primera ley del universo. Cómodamente fumás, cruzás las piernas, extendés la mano y levantás el vaso de whisky. (Ernesto Sábato, Sobre héroes y tumbas, Buenos Aires 1961) ‘Comfort is the first law of the universe. Comfortably, you smoke, cross your legs, hold out your hand and raise the glass of whisky.’ (9) Pero, por el momento, agradable era el descenso por la cuesta de Atocha [. . .] (Martín Santos, Tiempo de silencio, Barcelona 1962) ‘However, for the time being, the walk down the Calle de Atocha was pleasant.’

Thus what distinguishes the medieval situation from the modern one is not the possibility or impossibility of the structures in question but their rate of occurrence, which appears to have declined since the Middle Ages (although not in a uniform manner). The quantitative evolution of constituent fronting is discussed in detail in Sect. 2.5; however, to give substance to the point just made, summary data for the period 1250–1289 (covering the first two data points in the survey), together with comparator data from the post-1950s period, are given in Table 2.1.3 Table 2.1  Fronting rates per syntactic category in period 1250–1289 (with comparator data from post-1950s texts) Period

Objecta

PPa

Adverb

Predicative adj./ Non-finite noun verb form

1250–1289

6.1% (121/2000) 0.4% (8/2000)

8.1% (162/2000) 6.5% (129/2000)

2.8% (17/603) 3.8% (21/549)

11.3% (650/5757) 1.9% (21/1119)

Post-­1950 a

1.8% (320/17,415) N/A

Data for objects and PPs based on random sample (2000 tokens in each category)

28 

I. E. Mackenzie

2.3 Structural Analysis 2.3.1 O  ld Spanish Fronting Structures as Ā-Dependencies 2.3.1.1  S  imilarities Between Old Spanish Fronting and wh-Movement The Old Spanish fronting structures illustrated in (1a) to (5b) are in crucial ways analogous to interrogative structures (both modern and old) involving wh-movement. In the first place, analogously to when a wh-­ object or adjunct is fronted, any overt subject in an Old Spanish fronting structure must either be postverbal or it must be introduced to the left of the fronted item.4 Examples (10a) and (10b) illustrate these two placement options in a modern wh-structure, while (11) and (12) illustrate the same principle in the context of Old Spanish fronting structures. In each case, the subject is shown in bold typeface. (10a) ¿Qué tema aborda la película? ‘What theme does the film address?’ (10b) La película ¿qué tema aborda? ‘The film, what theme does it address?’

(11) Grand uerguenna ouieron los romanos del pleyto que mancino fiziera con los de Çamora assi cuemo desuso oyestes. (Estoria de España I, fol. 21r) ‘The romans were greatly ashamed of the treaty that Mancinus had made with the Zamorans, as was stated earlier.’

(12) Estos caualleros de sant yago. muchas obras fizieron en las tierras de Espanna. (Estoria de España II, fol. 296 r.) ‘Those knights of Santiago did many things in the land of Spain.

Examples (13) and (14) illustrate the same pattern in embedded clauses.

  Constituent Fronting: Focus, Discourse and Fashion 

29



(13) & si eneste medio algun iudizio diere el alcalde uala. (Fuero Real, fol. 11v) ‘And if in this situation the magistrate gave a judgment, it stands.’



(14) Bien ueo que este pueblo yerto es & de dura ceruiz. fascas de duro coraçon. (General estoria I, fol. 213r) ‘I can see that this nation is rigid and stiff-necked, that is to say, hard of heart.’

Secondly, the fronted item may be moved across a clausal boundary. For example, just as the wh-phrase qué coche in (15) originates as the object of the embedded verb compraron but ends up at the beginning of the matrix clause, so fronted esta penna in (16) is the object of the embedded verb ayan but appears to the left of the matrix verb mandamos.

(15) ¿Qué coche crees que compraron al final? ‘What car do you think they bought in the end?’



(16) & esta penna mandamos que ayan los obispos. & los sacerdotes. & los diaconos. (Fuero Juzgo, fol. 82v) ‘And this punishment we decree should apply to bishops, priests and deacons.’

Thirdly and finally, the fronted items in the Old Spanish structures are proclisis attractors, in the sense that the finite verb form to which they are preposed becomes incompatible with pronominal enclisis. A comparison with modern Spanish is unhelpful in this regard, because enclisis is no longer possible with finite verb forms (other than positive imperatives). However, Old Spanish wh-movement shows systematic proclisis if a weak object pronoun is present, as in (17) and (18) below.

(17) Muerto uengo de fambre. & si muriere que pro me tienen essas mayorias? (General estoria I, fol. 77r) ‘I am dying of hunger and if I die what good is this birthright to me?’

30 



I. E. Mackenzie

(18) Pues como me oyra el Rey ffaraon que es sennor; (General estoria I, fol. 151r) ‘In that case, how will the Pharaoh, who is a lord, listen to me?’

This placement pattern for clitics is replicated in the more generalized constituent fronting that was characteristic of Old Spanish, as is illustrated in (19) to (23).5 Fronted object:

(19) Et esta semeiança les dieron segund los quatros humores que se crian en el cuerpo dell omne. (Libro de ajedrez, dados y tablas, fol. 85v) ‘And they gave them the following representation in terms of the four humours that exist in the human body.’

Fronted PP:

(20) Mas por este nombre le llama sant bernaldo en la glosa. (General estoria IV, fol. 182v) ‘But Saint Bernard calls him by this name in the gloss.’

Fronted adverb:

(21) Ca mucho lo amauan yl preciauan por que assi los sacara de la seruidumbre de los Romanos & Regno diez annos. (Estoria de España I, fol. 168v) ‘For they greatly loved and appreciated him because he had in this way freed them from the servitude of the Romans, and he reigned for ten years.’

Fronted predicative adjective:

(22) Et dixoles meior me es amj de morir yo que non de celar a anthiogono la su muerte (General estoria V, fol. 135r) ‘And he said to them, “It is better for me to die than to conceal from Antigonus his own death.”’

  Constituent Fronting: Focus, Discourse and Fashion 

31

Fronted nonfinite verb form:

(23) la cosa que el omne no quiere fazer teniendo que es mala; guardar se deue quanto pudiere de no dar carrera poro la faga. (Libro de las leyes, fol. 45v) ‘As regards the thing which a man does not want to do because he knows it to be bad; he must refrain, as far as he can, from following a path which might lead him to do it.’

In the standard Generative analysis of wh-movement, the fronted interrogative phrase is analysed as having moved to a surface position in the left periphery of the clause, as is the finite verb if there is subject–verb inversion. The basic analytical schema is illustrated in (24) below, where ‘CP’ (‘Complementizer Phase’) denotes the full clause (i.e. including the left periphery) and ‘TP’ (‘Tense Phrase’) denotes the basic clause (i.e. excluding the left periphery). The strikethrough font indicates the base position of the relevant moved items, viz. the verb aborda and the wh-­ phrase qué tema.6

(24) ¿[CP Qué tema aborda [TP la película aborda qué tema]]?

According to this analysis, the fronted wh-phrase qué tema is in the (left) edge of CP, while the finite verb aborda is in the head of CP position, referred to as C or C0. This arrangement, given that the conventional Subject Position is assumed to be inside TP, to the right of C, delivers the surface verb–subject word order that is characteristic of wh-movement structures. Such structures are in fact analysed as being one particular instance – probably the prototypical instance – of a more general type of syntactic dependency, referred to as Ā-dependencies (Cinque 1990).7 The hallmarks of an Ā-dependency are (i) that it can cross a clause boundary, as when a wh-phrase is moved from inside an embedded clause and promoted to the beginning of the matrix clause, and (ii) that the fronted item has a determinate semantic or discourse-related property such as [+ interrogative] or [+ focus]. We have already seen – example (16) – that Old Spanish fronting dependencies can cross a clause boundary. In addition,

32 

I. E. Mackenzie

as is discussed in detail in Sect. 2.4 below, the item fronted in such structures turns out to be associated with a specific discourse-related property. Given these facts, and given the general parallel discussed above between Old Spanish fronting and wh-movement, it is highly plausible to analyse the former as giving rise to an Ā-dependency analogously to the latter.8

2.3.1.2  Applying the Analytical Framework To see explicitly how the proposed model works in the case of Old Spanish fronting structures, consider the earlier example (11), reproduced below with labelled bracketing as (25):

(25) [CP grand uerguenna ouieron [TP los romanos [VP ouieron grand uerguenna del pleyto que mancino fiziera con los de Çamora]]] assi cuemo desuso oyestes

As can be seen, the object grand uerguenna is analysed as moving to the edge of CP, while the finite verb ouieron has moved across the subject los romanos to the C position, exactly as is assumed to happen in wh-­ movement structures. The analysis is obviously consonant with the characteristic subject– verb inversion associated with Old Spanish fronting, given that los romanos in (25) is to the right of the surface position of ouieron. As regards cases in which the subject precedes the fronted item, these can be analysed as involving covert left-dislocation. In other words, the apparent subject is introduced as a left-dislocated topic, but this is obscured by the fact that the resumptive pronoun – the true syntactic subject – is silent, Spanish (both old and modern) being a pro-drop language. Assuming that left-dislocated constituents are external to the core CP, a sentence such as the earlier example (12) can be analysed as in (26) below, where pro represents the silent pronominal subject (which is co-referential with estos caualleros de sant yago).

(26) Estos caualleros de sant yago. [CP muchas obras fizieron [TP pro [VP fizieron muchas obras en las tierras de Espanna]]] ‘Those knights of Santiago, (they) did many things in the land of Spain.

  Constituent Fronting: Focus, Discourse and Fashion 

33

In this particular case, the scribe has actually inserted a full stop between the phrase estos caualleros de sant yago and the remainder of the sentence. This circumstance is consistent with the proposed analysis, given that full stops in the medieval manuscripts often indicate the possibility of an intonation break (Mackenzie 1997: 11). In general, however, there is no expectation that a left-dislocated “subject” should be demarcated by a specific punctuation mark. The possibility, in pro-drop languages, for an apparent preverbal subject to be in reality a left-dislocated topic is in fact quite widely recognized in the theoretical literature (see in particular Alexiadou and Anagnostopoulou 1998 and Barbosa 2009). The noted variation in weak pronoun placement in Old Spanish sentences that have a preverbal subject (see Sect. 3.2.3.4) can be interpreted as being a reflex of this structural ambiguity. Turning to the ‘long distance’ type of dependency illustrated by the earlier example (16), the basic analysis is indicated in (27) below:

(27) & [CP esta penna mandamos [TP mandamos [CP que ayan los obispos & los sacerdotes & los diaconos esta penna]]]

Here, the phrase esta pena is shown as being raised from object position in the embedded CP to the edge of the matrix CP, in the process crossing the boundary between the two clauses. As was noted above, the possibility of extraction across a clause boundary is a hallmark of the Ā type of dependency.9 Finally, the framework outlined enables the requirement for pronominal proclisis in Old Spanish fronting structures – see examples (19) to (23) above – to be incorporated into a broader theory of clitic placement in the medieval language. According to the descriptive generalization known as the Tobler–Mussafia Law (Tobler 1875; Mussafia 1888), weak object pronouns never occur sentence-initially in Old Romance, implying that they necessarily occur enclitically if the verb is the first element in the sentence. The generalization is correct for Old Spanish, examples such as (28) below being abundant in the medieval texts: (28) temen le quantos le ueen. et se acercan a el. (Judizios de las estrellas, fol. 11r) ‘It is feared by all those who see it and approach it.’

34 

I. E. Mackenzie

However, the descriptive power of the Tobler–Mussafia Law, as formulated above, is diluted to an extent by examples like (29), where the weak pronoun is again enclitic but the verb is not actually the first element in the sentence.

(29) Despues troxieron le otras mugeres Egipçianas otrossi mas dize Josepho que non quiso mamar njnguna njn llegar se a ellas. (General Estoria I, fol. 136v) ‘Afterwards they brought him other Egyptian women as well, but Joseph says that he refused to be breastfed by any of them or to go near them.’

The explanation for apparently unmotivated enclisis of this type appears to lie in the fact that the sentence-initial element is a circumstantial adverbial, determining the context of the main assertion rather than being part of it. In Spanish and many other languages, such adverbials occur naturally in sentence-initial position and are capable of being separated from the remainder of the sentence by an intonation break. Benincà (2006: 76) argues that such items are directly merged into a special field in the edge of CP which she labels ‘Frame’, rather than undergoing movement from a postverbal position. In the same spirit, but dispensing with the Frame construct, the assumption here will be that such items are simply external to the core CP structure, analogously to left-dislocated topics, as is shown schematically in (30):

(30) Despues [CP troxieron le otras mugeres Egipçianas otrossi] mas dize Josepho que non quiso mamar njnguna njn llegar se a ellas.

If it is further assumed that the domain of application of the Tobler– Mussafia Law is the CP rather than the sentence, examples like (29)/(30) come to be included within the generalization’s scope. In other words, both troxieron in (29)/(30) and temen in (28) are initial within the relevant domain for Tobler–Mussafia purposes, but this is now understood as being the CP rather than the sentence. On this basis, a revised version of the Tobler–Mussafia Law can be expressed as in (31):

(31) The weak pronoun is enclitic (on the finite verb) if and only if it would otherwise immediately follow a CP boundary.

  Constituent Fronting: Focus, Discourse and Fashion 

35

Given (31), the compulsory proclisis in Old Spanish fronting structures immediately falls into place. If such structures involve an Ā-dependency, with the fronted item occupying the (left) edge of CP rather than being external to CP, they do not satisfy the condition necessary for enclisis and hence any weak pronoun must be proclitic. From (31) we can also derive an explanation as to why the resumptive pronoun in Clitic Left Dislocation (CLLD) structures in Old Spanish is in principle enclitic (in main clauses), as in the example below:

(32) & aun esta ley guardan la alas uezes en la ley nueua. (General estoria I, fol. 238v) ‘And still, this law, it is sometimes retained in the new law.’

As was observed in note 1, the true object in CLLD is the resumptive pronoun (in the above case, la), while the associated preverbal phrase (esta ley) is CP-external. The resumptive pronoun would thus be CP-initial if it were proclitic; hence enclisis is expected under generalization (31).10 To sum up this section: Old Spanish fronting structures can be analysed as Ā-dependencies, in which the fronted expression is in the left edge of CP and the finite verb occupies C.11 It follows, under Chomsky’s hypothesis concerning the locus of the structural subject position, that any overt ‘true’ subject will be postverbal in a fronting structure of this kind. Apparent subjects that precede the fronted item are in reality dislocated topics whose resumptive subject pronoun is phonologically null or silent, given the pro-drop status of Old Spanish.

2.3.2 Old Spanish Fronting and Verb-Second Syntax Given the foregoing remarks, the Old Spanish fronting structures can be seen as being verb-second or V2 constructions. Following the standard analysis, a V2 structure can be defined as one in which the finite verb is in the C position and is preceded, within CP, by a single constituent (other than a clitic pronoun or the negation morpheme). Note that the existence in a language of V2 constructions does not necessarily entail that the language in question is a V2 language from the typological point of view, as are the continental Germanic languages, together with the mainland

36 

I. E. Mackenzie

Scandinavian ones. Modern English, for example, is not a V2 language per se, but it has V2 syntax in clauses in which the initial element is a fronted wh-phrase or negative phrase, as in (33) and (34) below:

(33) Which seminar can you attend?



(34) Never again will I shop in that supermarket.

Rizzi 1990 uses the term ‘residual V2’ to describe structures such as the above, implying that they are a residue of a prior phase during which the containing language was a V2 language. The term is appropriate for English, which is known to have previously been a V2 language. However, the label is more problematic for modern Spanish structures such as (24), repeated below, as it is not certain that Old Spanish was a V2 language; at least not if that classification is interpreted as implying the existence of a generalized requirement for the verb to be in second position (but see Fontana 1993 and Wolfe 2015 for the opposing view).

(24) ¿[CP Qué tema aborda [TP la película aborda qué tema]]?

One reason for being circumspect about treating Old Spanish as a language with generalized V2 syntax is that the range of initial constituents which trigger the V2 effect is much narrower in Old Spanish than it is in the Germanic languages, which are the prototypical exemplars of V2 syntax. A striking example of this difference can be seen when the first element in the sentence is an embedded adverbial clause. In this case in German, for example, the matrix clause is subject to a strict V2 effect, illustrated by the inversion of man and ist in (35), which ensures that ist is in second position overall:

(35) [Wenn man keine Träume mehr hat] ist man leer. (Holmberg 2015: 348) [if one no dreams anymore has] is one empty ‘If you have no dreams anymore, you’re empty.’

In contrast, initial adverbial clauses in Old Spanish do not trigger a V2 effect, in the sense that the subject and the finite verb in the matrix clause

  Constituent Fronting: Focus, Discourse and Fashion 

37

are not forced to invert. Thus in examples such as (36) and (37) below the matrix clause has SV order, which entails that the finite verb, shown in bold, is actually in third position overall.12

(36) & si fuere Saturno enel ascendente & Jupiter cayente del; aquel enemigo que guerrea aquella uilla. tornas della defendiendos la yente de la uilla del. (Judizios de las estrellas, fol. 94v) ‘And if Saturn is in the ascendant and Jupiter is in the descendant; that enemy who is fighting against that town turns away from it, the people of the town defending themselves against him.’



(37) E quando fue uençudo de Annibal en la batalla; este so fijo le saco della. (Estoria de España I, fol. 14v) ‘And when he was defeated by Hannibal in battle, this son of his rescued him.’

Postverbal placement of the matrix subject is nevertheless possible in Old Spanish after an adverbial clause, as examples like (38) below demonstrate:

(38) Et quando llegaron ael. omillos le Jacob. & saludol. & començo a rogar adios por el. (General estoria I, fol. 110r) ‘And when they came before him, Jacob knelt down and greeted him and began to pray to God for him.’

However, in this type of case, the “V2” word order is accidental rather than forced, because, in light of examples like (36) and (37), the subject Jacob could equally well be placed before the finite verb omillo(s). In other words, the fact that the subject is postverbal in (38) is unrelated to the presence of the adverbial clause in initial position. The lack of any connection between the postverbal placement of the subject in (38) and the presence of the quando-clause in initial position is implicitly confirmed by the fact that exactly the same placement option for the subject may be selected even in situations where nothing at all precedes the finite verb. This possibility is illustrated in (39) below, where guisaron (se) ‘prepared (themselves)’ is the initial element in the first sentence of the chapter and its subject, los Romanos, is postverbal.

38 



I. E. Mackenzie

(39) XV Guisaron se los Romanos & sacaron desi muy grand huest. & bastescieron la de armas & de todas las otras cosas que mester eran pora lid. (General estoria IV, fol. 242r) ‘(Chapter XV) The Romans prepared themselves and raised a very large army. And they provided it with arms and with all the other things needed for war.’

The foregoing divergence between Old Spanish and Germanic strict V2 languages is not limited to contexts in which the initial element is an adverbial clause; it arises whenever the preverbal element is a circumstantial adverbial of any kind. This fact is illustrated in (40) to (42) below, where the relevant adverbials are por esto ‘for this reason’, estonces ‘then’ and desi ‘afterwards’:

(40) & por esto el Rey deue entender que uence todas las cosas por que es rey. (Fuero Juzgo, fol. 3v) ‘And for this reason, the king must understand that he outranks all things because he is king.’

(41) Estonces alexandre començo a roer los dientes. & tornar la cabeça aunas & a otras partes. (General estoria IV, fol. 228v) ‘Then Alexander began to grind his teeth and to turn his head this way and that.’ (42) Desi los linages partieron sus suertes entre sus companas. (General estoria I, fol. 317v) ‘Afterwards the members of each clan divided their allocation among themselves.’

As can be seen, no V2 effect is triggered by these adverbial expressions and the subject precedes the finite verb, causing it to occupy third position. In contrast, in the equivalent structures in Germanic V2 languages the finite verb would have to come immediately after the clause-initial adverbial and, as a consequence, the subject would be postverbal (see Holmberg 2015, section 3.1.3). On the other hand, exactly as in sentences in which the preverbal element is an adverbial clause, the type of case illustrated by (40) to (42) is

  Constituent Fronting: Focus, Discourse and Fashion 

39

also compatible with postverbal placement of the subject, as is demonstrated by examples such as (43) and (44):

(43) e por esto dizen los sabios que el uso e la costumbre es la natura segunda. (Libro de las animalias, fol. 47r) ‘And for this reason, wise men say that use and custom are second nature.’



(44) & desi tomo ell Arçobispo iulian el libro & las razones con quel amparaua; [. . .] (Estoria de España I, fol. 187r) ‘Afterwards Archbishop Julian took the book and the arguments with which he defended it . . .’

The situation illustrated by these cases is analogous to the ‘accidental’ V2 word order which can occur in standard Arabic. The latter language is typologically a VSO language (Fassi Fehri 1993), which means that the word order will resemble that of a V2 structure whenever a non-subject constituent is placed to the left of the finite verb. Obviously, however, this V2-like surface order merely reflects the general verb–subject word order and does not imply the existence of a V2 constraint (see Holmberg 2015, section 2.2). Now without necessarily being a VSO language (but see Bossong 2006), Old Spanish clearly did have a predilection for VS(O) order, and indeed that order has not been entirely lost even in the modern language (see Zubizarreta 1998; Ordóñez 1997).13 The surface V2 ordering found in Old Spanish examples such as (38), (43) and (44) should be seen as resulting from the coincidence of VS(O) word order and the fortuitous presence of a clause-initial constituent, rather than as being the consequence of a generalized V2 constraint. Notice, furthermore, that in the examples in which the V2 word order is accidental rather than obligatory, any weak pronouns which may be present are enclitic in relation to the verb. This can be seen with omillos le in (38), where -s is a reduced form of reflexive se. In contrast, as was noted apropos of examples (19) to (23), the Old Spanish fronting structures (assumed here to instantiate ‘true’ V2) exhibit systematic proclisis if any weak object or reflexive pronouns are present. The following generalization can thus be advanced:

40 



I. E. Mackenzie

(45) True V2 structures in Old Spanish exhibit pronominal proclisis whereas accidental V2 structures exhibit enclisis.14

The pattern captured in (45) can be seen as a corollary of the fact that V2 syntax in Old Spanish is limited to structures in which an item is d ­ isplaced from its usual position inside the verb phrase to a position in the (left) edge of CP. By definition the displaced item is CP-internal and hence proclisis is mandatory, given the Tobler–Mussafia Law as interpreted in (31), repeated below:

(31) The weak pronoun is enclitic (on the finite verb) if and only if it would otherwise immediately follow a CP boundary.

Conversely, in accidental V2 structures, a CP boundary can be posited between the preverbal constituent and the finite verb. Any weak pronoun that was proclitic in relation to the verb would be adjacent to this boundary; hence enclisis is expected under (31). In light of the foregoing remarks, the most accurate assessment of Old Spanish appears to be that it has restricted rather than generalized V2 syntax. The label restricted V2 is intended to imply that although the language overall cannot be classified as a V2 language on a par with continental Germanic and mainland Scandinavian, the occurrence of V2 syntax is nevertheless more systematic and less sporadic than is the case with residual V2 syntax, as found, for example, in modern English. In modern diachronic syntax, it is customary to classify stages in the history of a language in terms of language-specific values, known as settings, of general syntactic variables known as parameters. According to Roberts (2007: 271), full V2 languages like modern German have a positive or ‘yes’ setting for the Verb Second parameter, requiring the finite verb to move to the C position in all finite main clauses, resulting in subject–verb inversion whenever the initial element in the clause is not itself the subject. Given the restricted nature of V2 in Old Spanish – and given that its V2 syntax is not limited to main clauses – it would be difficult to claim that Old Spanish also had a positive setting for the Verb

  Constituent Fronting: Focus, Discourse and Fashion 

41

Second parameter, at least as it is defined by Roberts. On the other hand, even in its restricted guise, the V2 syntax found in Old Spanish does appear to be emblematic of a particular type of language and, as such, it would make sense to analyse it as being the outward reflex of a specific parameter setting. Accordingly, we can tentatively posit a Restricted Verb Second parameter, shown below as (46), for which the setting in Old Spanish would be ‘yes’.

(46) Does the finite verb move to C whenever a constituent undergoes Ā-movement to the edge of CP?

Languages which have a positive setting for (46) will have V2 syntax in all cases in which a normally postverbal element is fronted to the preverbal field, but not when the initial element is a circumstantial adverbial or a left-dislocated topic (both of which, by hypothesis, are external to the core CP). As regards the position of the finite verb in accidental V2 structures, this can be inferred from what is known about VS(O) word order, given that accidental V2 is no more than VS(O) preceded by a circumstantial adverbial. According to the standard analysis (see e.g. Alexiadou and Anagnostopoulou 1998), the finite verb in VS(O) structures heads TP, the ‘Tense Phrase’, with the subject remaining in a ‘low’ position in the periphery of the verb phrase, as is shown schematically in (47).15

(47) [CP C [TP Verb [v*P Subject [VP Verb (Object)]]]]

The default assumption would be that the VS(O) component of accidental V2  in Old Spanish instantiates (47), in the general case at least.16 Given that the finite verb in modern Spanish is also assumed to head TP in surface structure – SV(O) order arising when the subject moves across the verb to the edge of TP  – the overall conclusion is that the basic ­position of the finite verb is the same in Old Spanish and in modern Spanish. The option of leaving the subject in a postverbal position has obviously declined in frequency (cf. Fontana 1993: 251–252), but this does not in itself diagnose any change in the position of the finite verb.

42 

I. E. Mackenzie

2.4 Fronting and Discourse-Pragmatics The previous section motivates an analysis of Old Spanish according to which the characteristic medieval fronting structures involve Ā-movement to the edge of CP, with the finite verb moving to C as a corollary. Chomsky (2008: 151) conceptualizes such movement in terms of a so-called edge-­ feature on C, which attracts the relevant item into the edge of CP. In Chomsky’s model, the edge feature is indiscriminate, in the sense that it can target any item ‘in its domain’ (ibid.), but in order for the resulting structure not to be interpreted as deviant, the item attracted into the CP edge must have a particular semantic or discourse-related property, such as being interrogative, a focus or a topic.17 Accordingly, the analysis of Old Spanish fronting set out in the previous section implies that the moved item can be expected to be pragmatically marked in some way. A number of authors (England 1980; Danford 2002; Cruschina and Sitaridou 2009; Sitaridou 2011; Eide and Sitaridou 2013) have indeed proposed that fronted constituents in Old Spanish tend to have either a topical/discourse-given status or a focal/emphatic one. The discussion below identifies a number of fairly robust patterns in the data from the 1200s that provide direct empirical support for this general approach and which also help to deconstruct the relevant pragmatic concepts, particularly the notion of focus insofar as it relates to fronting in Old Spanish. We can start by considering the quantitative data given in Table 2.2 below. This shows the proportions of the various different types of object and PP that make up the 1250–1289 samples referenced in the earlier Table 2.1. Table 2.2  Object and PP fronting in the 1250–1289 sample Type

Objects

PPs

(P +) demonstrative noun phrase (P +) demonstrative pronoun (P +) quantified noun phrase (P +) bare quantifier (P +) definite noun phrase (P +) possessive noun phrase Others Total

11.9% (10/84) 55.1% (59/107) 17.6% (12/68) 18.8% (3/16) 4.2% (29/691) 2.3% (2/86) 0.7% (7/946) 6.1% (121/2000)

18.1% (58/320) 38.7% (12/31) 15.6% (7/45) 7.7% (1/13) 5.6% (57/1025) 5.1% (13/257) 4.2% (13/310) 8.1% (162/2000)

  Constituent Fronting: Focus, Discourse and Fashion 

43

2.4.1 The Demonstrative Effect According to the table, the overall fronting rate for objects during the relevant period is 6.1%. Significantly, this rate increases to 11.9% for objects which are introduced by a demonstrative determiner, as in (1a) repeated below, and to 55.1% for objects consisting in a bare demonstrative pronoun, as in (48). Using the numerical data in the table, it can in fact be determined that the odds (see Sect. 1.2.2) of an object being fronted increase by a factor of 19.110 if the object is a demonstrative phrase or a bare demonstrative pronoun.18 This is a dramatic increase by any standards.

(1a) Et esta semeiança posieron al can destos menores. assi como a carauo. o de otros perriellos mas pequennos que allegan los ombres assi mas que los mayores (Libros del saber de astronomía, fol. 9v) ‘And they made this comparison to the small type of dog, such as a little hunting dog or other smaller dogs that such men talk about more than bigger ones.’



(48) Qvando esto quisieres saber; et ouieres la Era arabiga sabuda [. . .] (Cánones de Albateni, fol. 30r) ‘When you want to know this and you already know the Arabic date . . .’

Demonstrative expressions have two principal functions within discourse, viz. ostension and anaphora. The first use implies that the speaker/ writer and his or her audience are present in the same place at the same time, which in practice requires direct oral communication. In written texts, then, demonstratives are almost always anaphoric, in the sense that they are co-referential with an item mentioned in the preceding discourse.19 For example, the phrase esta semeiança in (1a) refers back to a discovery alluded to in the immediately preceding sentence, which is shown as (49) below:

(49) Et destas estrellas que son aqui dichas de suso fallaron los sabios enna figura del can menor. (Libros del saber de astronomía, fol. 9v) ‘And learnèd men found some of these stars mentioned above to be in the form of a small dog.’

44 

I. E. Mackenzie

Phrases involving demonstratives are thus highly likely to be discourse-­ given and the disproportionately high rates of fronting associated with demonstrative objects in Old Spanish can plausibly be taken as evidence that topic status – understood here as implying that the phrase in question represents ‘old’ or ‘presupposed’ information  – was a trigger for fronting. This picture is reinforced by the situation affecting prepositional phrases (PPs). As can be seen in Table 2.2, the overall fronting rate for this syntactic category in the 1250–1289 data is 8.1%. However, this rises to 18.1% if the complement of the preposition is a demonstrative noun phrase, as in the earlier example (2b) repeated below, and to 38.7% if the complement of the preposition is a bare demonstrative pronoun such as esto, as in (50).

(2b) Et si con esta agua lauaren pannos en que aya oro de que los no puedan alimpiar dotra guisa; laua los muy bien que no pierden niguna cosa de su fremosura. (Lapidario, fol. 86r) ‘And if with this water you wash fabrics stained with gold which cannot be cleaned in another way, it washes them well and they lose nothing of their beauty.’



(50) E si algunos contra esto fiziessen; cadrien en simonia. (Libro de las leyes, fol. 86r) ‘And if any persons contravened this, they would be guilty of simony.’

The data in Table 2.2 deliver an odds ratio (see note 18) of 4.216 for the fronting of demonstrative PPs vis-à-vis non-demonstrative PPs. This ratio is lower than the corresponding ratio for objects, but it is still indicative of a powerful demonstrative bias in the fronting data. That the effect is lower among PPs than it is for objects is probably to be expected, given that prepositional phrases are often adjuncts. Such items generally enjoy greater freedom of placement than objects and, as a corollary, quite often appear in preverbal position even if no specific discourse-related trigger is present. The strength of any discourse-related fronting effect is thus likely to be diluted by cases of fronting in which discourse-pragmatics is not a factor.

  Constituent Fronting: Focus, Discourse and Fashion 

45

The demonstrative trigger for fronting appears in fact to pre-date written Spanish, being noticeable already in the Latinate Ibero-Romance of the época de orígenes (see Sect. 1.1). This is illustrated in the two extracts below, dating from 1063 and from around 1030 respectively. In both cases, the fronted demonstrative object is highlighted in bold.

(51) Siquis . . . hanc meam donationem uoluerit refragare sit excomunicatus . . . et insuper regi terre persoluat çentum libras auri obriçi et uobis dampnum restituat duplatum quod fecit (Archivo Histórico Nacional de Madrid, Oña P-292; Menéndez Pidal 1926: 43) ‘If anyone seeks to oppose this gift of mine he should be excommunicated and, in addition, to the king of the land he must pay 100 pounds of pure gold and to you should reimburse the damage that he caused, doubled’



(52) In torre de Abolmondar, suo uarrio; jn torre de Abolazaba, suo uarrio, suas sernas et suo mulino; jn Mercatiello suo uarrio; jn Torriziella, suo uarrio et suas sernas et suo mulino; jnAbella nosa, suo uarrio et suas sernas et suo mulino. Toto jsto tenet Ruderico Godestioz qui a jsto toto (Archivo Histórico Nacional de Madrid, San Juan de la Peña, núm. 18; Menéndez Pidal 1926: 41) ‘In Torre de Abolmondar, its built-up area; in Torre de Abolazaba, its built-up area, its fields and its mill; in Mercadillo its built-up area; in Torrecilla, its built-up area and its fields and its mill; in Abellanosa its built-up area and its fields and its mill. All this is owned by Rodrigo Godestioz, who has all of this.’

Extract (51) involves a ‘sanctions’ clause which, under various guises, appears to have been common in documents recording land donations to monasteries. It is noticeable that the demonstrative object hanc meam donationem, which refers back to a list of the properties being donated, is placed immediately to the left of the finite verb, despite the fact that elsewhere in the document the object tends to follow the verb (e.g. Conçedo etiam solares et hereditates ‘I also grant land and fields’, which in the manuscript appears a few lines above the extract given here). Extract (52) is from a text recounting various events to do with land ownership in Espeja de San Marcelino (Soria province) and adjacent villages. As in the previ-

46 

I. E. Mackenzie

ous example, the demonstrative object, viz. toto jsto ‘all this’, is transparently coreferential with specific items in the immediately preceding discourse and this appears to be what triggers the preverbal placement. It is also interesting to note that a reordered version of the same phrase, viz. jsto toto, has a postverbal occurrence in the immediately following relative clause, indicating that fronting was a variable phenomenon, exactly as in post-1200 Spanish.

2.4.2 Definite Noun Phrases The discourse-given category is, of course, not limited to demonstrative phrases. Any definite noun phrase whose head noun replicates a noun used in the immediately preceding discourse is likely to fall into this category, and, like the demonstrative phrases considered above, such items could be fronted in Old Spanish. In (53) below, for example, the noun plazo within the fronted object el plazo ‘the term’ in the first relative clause in the apodosis directly resumes two occurrences of the same noun in the protasis.

(53) Si algun omne a plazo sabudo por iuyzio a que pague a otri alguna debda & non la pagare: al plazo. los alcaldes que el plazo dieren manden al merino que entregue de los bienes del debdor de mueble o de rayz a aquel que a de auer la paga. (Fuero Real, fol. 85v) ‘If some man has a term fixed by a judgment to pay another person some debt and does not pay it within the term, the magistrates who set the term must order the bailiff to surrender the goods of the debtor, either chattels or landed property, to that person who should have the payment.’

The same principle also holds for possessive objects, like sus maridos at the end of the extract in (54) below:

(54) Assi acaescio. Que cada una daquellas hermanas mato asu marido aquella noche. saluo ende ypermestra sola que non quiso matar asu marido. & dezir uos emos primero delas otras. desi contar uos emos desta. ¶ Las otras hermanas todas que sus maridos mataran [. . .] (General estoria I, fol. 314v)

  Constituent Fronting: Focus, Discourse and Fashion 

47

‘In this way it happened that each of those sisters killed her husband that night, except Hypermnestra alone who refused to kill her husband. And we shall speak first of the others. Afterwards we will talk about her. All the other sisters who had killed their husbands . . .’

As can be seen, the noun which occurs in the fronted object sus maridos is used twice in the immediately preceding context, making this object’s discourse status analogous to that of el plazo in (53). A less obvious subcategory of discourse-given noun phrase comprises phrases which are, in Cinque’s (1990: 87) words, ‘inferentially linked’ to the immediately preceding discourse. In this type of case, the anaphoric link is not signalled explicitly through identity of linguistic expression or through the use of an overtly resumptive element such as a demonstrative determiner. Rather, it is deduced on the basis of non-linguistic knowledge. Phrases that have this inferential link to the previous discourse, exactly like the overtly anaphoric phrases just discussed, have a tendency to be fronted in Old Spanish. This accounts for the preverbal placement of the object phrase el tuerto ‘the wrong’ in the legal formula illustrated in (55) below.

(55) E mandamos e deffendemos que ninguno non sea osado de ir contra este priuilegio pora crebantarlo ni pora minguarlo en ninguna cosa ca qualquier que lo fiziesse aurie nuestra ira e pecharnos ye en coto diez mill morauedis e a los que el tuerto recibiessen todo el danno doblado. (Alfonsine privilegio, 3rd July 1273, Guadalajara; Sevilla: Arch. Mun. Sec. I, c. 1, n. 17) ‘And we order and forbid anyone to dare to go against this privilege in order to break it or to diminish it in any way, because anyone who did would incur our anger and would pay us as a fine of ten thousand maravedies and to those who suffered the wrong, double the damage caused.’

The phrase el tuerto does not explicitly replicate language used in the preceding discourse. Nevertheless, it can be deduced to be anaphoric in view of the reference in the previous sentence to infringing the provisions of the priuilegio: entertaining the possibility of non-compliance implicitly references the concept of wrongdoing, and hence introduces into the context the notion of a tuerto or wrong committed against a person.20

48 

I. E. Mackenzie

An additional point to note about definite objects is that they can function as topics without actually being discourse-given, something which is not in principle possible with demonstrative objects, given that the latter manifest their anaphoric character overtly. Thus in (56) below, the fronting of the objects in bold is not due to prior mention (explicit or implicit), but rather to the topic–comment structure, the objects being introduced as topics about which a comment is subsequently made: (56) Toda la tierra desgastaron los enemigos. Las casas hermaron. Los omnes mataron. Las cibdades quemaron. Los arbores. Las uinnas & quanto fallaron uerde cortaron. (Estoria de España I, fol. 193v) ‘The enemies let all the land go to waste. They depopulated the towns. They killed the men. They burned the cities. They cut down the trees, the vines and whatever they found that was green.’

Examples involving body parts, such as (57), can presumably be assimilated to the same category: (57) Los oios auie muy grandes. & lo que es grand marauilla auie los tan claros que ueye de noch a lo lobrego. (Estoria de España I, fol. 69v) ‘His eyes were very big, and what was astonishing is that they were so light that he could see in the dark at night.’

As was shown in Table 2.2, definite objects in the 1250–1289 sample have a fronting rate of 4.2% (29/691) while the corresponding rate for PPs is 5.6% (57/1025). According to these data, definiteness per se is not a trigger for fronting, given that the rates in question are lower than the rates for the relevant syntactic categories overall (6.1% for objects and 8.1% for PPs). However, the figures just given are based purely on grammatical form and take no account of the discourse–pragmatic status of the item in question. A sub-classification of the definite objects and definite PPs in the sample reveals that the fronting rates increase to 7.9% (17/216) and 11.4% (38/333) respectively if the item in question is a topic (either in the discourse-given sense or in the sense of being the first element in a topic– comment structure). Based on these figures, the fronting odds ratios for topical definite objects and PPs in the 1250–1289 sample are 1.380 and 1.603 respectively, implying that the combination of definiteness and topi-

  Constituent Fronting: Focus, Discourse and Fashion 

49

cality increases the odds of fronting by 38% in the object category and by 60.3% among PPs. These figures point to a modest but clear fronting effect for both types of item, albeit one which is weaker than when the topicality of the relevant phrase is anchored in its demonstrative form (see Sect. 2.4.1). An explanation for this discrepancy can perhaps be derived from the assumption that, in their anaphoric use, demonstrative determiners retain an element of their indexical character, encouraging the placement of demonstrative phrases in positions that are maximally proximate to the item with which they are co-referential; that is, in positions at or near the beginning of their containing clause. It would then be expected that speakers availed themselves of the preverbal field rather more frequently in the case of demonstrative phrases than in that of phrases introduced by the definite article, which can be assumed by the thirteenth century not to retain any residue of its earlier indexical value (cf. Chap. 4).

2.4.3 The Quantifier Effect A further important trend that can be inferred from the data in Table 2.2 is highlighted in Table 2.3, which relates to the fronting of constituents that either contain or consist of a quantifier expression such as muchos, algo, algunos etc. Examples of this type of fronting are given in (58) to (60) below: Quantified object:

(58) Quando algo dieres afazer. non demandes que se faga la obra ayna. mas demandad que se faga buena. (General estoria IV, fol. 234r) ‘When you stipulate something to be done, do not ask for the task to be done quickly but ask for it to be done well.’

Table 2.3  Quantified objects and PPs in the 1250–1289 sample Quantified Objects PPs a

Non-quantified

Fronted

Total

Rate

Fronted

Total

Rate

Odds ratioa

15 8

84 58

17.9% 13.8%

106 154

1916 1942

5.5% 7.9%

3.712 1.858

Quantified versus non-quantified

50 



I. E. Mackenzie

(59) Ca mucho conorte toman los omnes usando de las costumbres que uso & amo. (Estoria de España II, fol. 257r) ‘For a man derives much comfort from doing the things he has always done and loved.’

Quantified PP:

(60) E si en alguna cosa erraren; deue los ensennar cuemo han de fazer. segund manda santa eglesia. (Libro de las leyes, fol. 115v.) ‘And if they err in something, he must show them how they must proceed, in accordance with church ordinances.’

As can be seen from the odds ratio column in Table 2.3, quantified objects in the sample are over three and half times more likely to front than other objects, while quantified PPs are nearly twice as likely to front than other PPs. (The gap between the percentage rates for the two categories of constituent is narrower, viz. 17.9% versus 13.8%, but those rates do not take into account the fact that PPs overall have a higher fronting rate than objects.) The data in the sample can thus be regarded as supporting the view that quantifiers were associated with a fronting effect in thirteenth-­century Spanish. While the quantifier effect on fronting is not as strong as the demonstrative effect reported in Sect. 2.4.1, for which the OR was 19.110 in the case of objects and 4.216 for PPs, it is still a robust one, suggesting that quantifier fronting represents an important aspect of Old Spanish usage (cf. Mackenzie 2010; Mackenzie and van der Wurff 2012). Quer (2002) has in fact discussed this type of fronting in relation to modern Spanish and Catalan, proposing (p. 261) that it is linked to a so-­called ‘focus-affected reading’. This term alludes to the phenomenon whereby accenting a quantifier’s nominal complement causes this latter element to exchange semantic roles with the clausal predicate. This is illustrated in (61) below, in which cooks is the complement of the quantifier few and applied is the clausal predicate.

(61) Few COOKS applied.

  Constituent Fronting: Focus, Discourse and Fashion 

51

Accenting the word cooks, represented here by capitalization, results in the sentence being interpreted as asserting that few applicants were cooks (with cooks implicitly the clausal predicate), as opposed to the default reading according to which few cooks were applicants. According to Quer (ibid. 263), col·legues in (62) below is like COOKS in the English example, becoming interpreted as the clausal predicate rather than as the restrictor expression within the quantifier phrase (QP) to which it belongs syntactically. (62) Pocs col·legues hi ha convidat, a la festa, la Mireia. ‘Few of the people that Mireia has invited to the party are colleagues.’

Quer’s approach is plausible for the type of case illustrated by (62); that is, when the quantifier is in the plural and its complement is a noun which is actually capable of being interpreted as the clausal predicate. However, the focus-affected reading is unnatural if the quantifier’s complement is the dummy noun cosa as in (60): it would be perverse to analyse the si-clause in that example as saying ‘if one of the things they err in is a thing’. The focus-affected analysis is similarly implausible if the quantifier has a singular abstract noun as its complement, such as conorte in (59). The intended effect of this latter example is surely the one captured by the translation given earlier and not by the reading in which conorte is interpreted as the clausal predicate, which might be paraphrased as ‘Much of what a man derives from doing the things he has always done and loved is comfort’. Finally, when the quantifier is a bare pronoun like algo in (58), it is hard to see how the focus-affected analysis can even get off the ground. What these remarks highlight is that the examples in question, i.e. (58) to (60), lack a focus-bearing nominal complement. A plausible assumption therefore, and one which preserves Quer’s basic insight that quantifier fronting in principle involves focus, would be that in this specific case it is the quantifier (Q) itself which is in focus, rather than the QP headed by the quantifier. From that perspective, (59) can be regarded as foregrounding the magnitude associated with mucho, while the existential

52 

I. E. Mackenzie

quantifier embodied in algo in (58) and alguna cosa in (60) can be analysed as receiving a ‘strong’ interpretation, hence as having an equivalent quantificational force to English stressed SOME [ˈsʌm] as opposed to unstressed some [səm]/[sm]. This latter assumption is given additional credence by the fact that fronted algo in modern Spanish typically corresponds to stressed SOMETHING in English, as is shown in (63) below.21

(63) Algo debe saber. ‘He/She must know SOMETHING.’

Furthermore, the core notion built into the proposed analysis, viz. the possibility of maximally narrow focus on the quantifier itself, appears to be overtly realized in the ‘extraposed’ pattern illustrated by (64):

(64) Maldita sea la sanna del traydor Julian. ca mucho fue perseuerada. (Estoria de España I, fol. 192 r.) ‘Damn the anger of the traitor Julian, for it was very long-lasting’

Here, the fact that only the quantifying adverb mucho is fronted, while its adjectival complement is left stranded to the verb’s right, indicates that the focus must be on the quantifier alone and not on its containing phrase. Combining, then, the possibility for maximally narrow focus on Q with Quer’s overall approach, we arrive at the generalization in (65):

(65) A fronted quantifier expression, Q or QP, attracts focus onto Q or onto Q’s nominal complement.

Given (65), the fronting effect associated with both Q and QP in Old Spanish implies that the preverbal position to which objects, PPs, adverbs etc. could move was associated not just with the discourse-given feature identified in Sects. 2.4.1 and 2.4.2, but also with focus. Importantly, however, the focus in question is not the familiar ‘presentational’ or ‘new information’ type of focus in which the focused constituent supplies the answer to a context-question such as Who wrote that novel? or What did the protestors do? (see Lambrecht 1994). Rather, quantifier-related focus falls into the less well studied category of affective, emphatic or mild

  Constituent Fronting: Focus, Discourse and Fashion 

53

focus (see Hernanz 2001, 2006; Gallego 2007; Leonetti and Escandell Vidal 2010), where the principal effect is to foreground an item rather than to present it as new information. Furthermore, this latter type of focus appears to be capable of being very narrowly targeted, attaching, where necessary, to the quantifier in isolation rather than to its containing phrase. In contrast, new-information focus, when it applies to a phrase-level constituent, is assumed to apply to the whole of that constituent.22

2.4.4 Predicate Fronting The foregoing analysis, according to which some of the fronting in Old Spanish causes the relevant constituent to receive affective focus, can be extended to the fronting of predicative elements such as adjectives or participles. This can be seen from consideration of examples like (66) and (67) below, which are cited by Sitaridou (2011: 175–176) who in fact analyses them as involving new-information focus rather than affective focus.

(66) & los qui se gozaron con el to derribamiento penados seran por ello. (General estoria IV, fol. 106v [sic.]) ‘And those who rejoiced in your fall will be punished for it.’ (My translation)



(67) Quando esto oyo ysaac espantado fue fiera mientre. (General estoria I, fol. 79v) ‘When Isaac heard this, he was deeply shocked.’ (My translation)

It can be noted first of all that the subject of the root clause is, in both cases, the clausal topic as well. In (66), the root subject is the long phrase los qui se gozaron con el to derribamiento, which serves to set up the topic for a subsequent comment. In (67) the root subject is the phonologically null subject of fue (which is coreferential with ysaac, the overtly expressed subject of the sentence-initial quando-clause). In view of its discourse-­ given status and null phonological realization, it can be regarded as being a ‘ratified topic’ in the sense of Lambrecht and Michaelis 1998.23 The

54 

I. E. Mackenzie

root clauses in (66) and (67) thus have a fairly transparent topic–comment information structure, the item bearing the new-information focus being in each case the predicate. That is to say, the two examples involve predicate focus, in the sense of Lambrecht 1994. At first glance, this analysis seems to be in line with Sitaridou’s contention that the fronted element bears the new-information focus. However, it is well known that predicate focus is the default information structure in topic–comment sentences which have an unmarked word order. In other words, leaving penados and espantado in situ, as in (69) and (70) below, would also deliver predicate focus, implying that these items do not actually need to be fronted in order to be interpreted as focal (in the new-information sense):

(69) & los qui se gozaron con el to derribamiento seran penados por ello.



(70) Quando esto oyo ysaac fue espantado fiera mientre.

The impression that the fronting of penados and espantado in (66) and (67) is superfluous, as far as new-information focus is concerned, is reinforced by the fact that the manner adverb fiera mientre in (67) is left in situ. This item modifies espantado and hence delivers part of the comment which is made in respect of the topic; by the same token, it forms part of the clause’s new-information focus. If fiera mientre can remain in a postverbal position and still contribute to the new-information focus, it seems inconsistent to argue that espantado is fronted in order to be interpreted as expressing such focus. These considerations suggest that the fronting operation illustrated in (66) and (67) must do something other than assign new-information focus to the moved items. At this historical distance it is impossible to be certain what this additional effect was. However, the fronting of the predicative element in these medieval examples appears to be structurally analogous to the fronting exhibited by modern examples such as (71) below (retrieved using Corpus del español: Web/Dialects), given that in all three cases an adjectival predicative element occupies a surface position to the left of a copula.

  Constituent Fronting: Focus, Discourse and Fashion 

55

(71) Harto estoy de ver este tipo de cosas en los institutos. ‘I’m tired of seeing this type of thing in secondary schools.’ (http://tiscar.com/2007/03/27/en-defensa-del-derecho-a-la-informacion/)

Here the fronted predicative adjective harto can be broadly characterized as being emphatic, analogously to items like algo in (63) or mucho in (64), which were analysed as bearing affective focus. Accordingly, given the structural parallel between harto in (71) and penados and espantado in (66) and (67), it seems reasonable to assume that the fronting operation in these latter examples also delivers affective focus. Notice that the affective focus resulting from the fronting operation does not alter the role that the fronted item has in terms of the sentence’s new-information focus structure. That is to say, affective focus and new-­ ­ information focus are in principle orthogonal to one another. Indeed, it was already observed at the end of Sect. 2.4.3 that affective focus is non-­isomorphic with new-information focus, in the sense that it maps onto potentially finer-grained syntactic units. One consequence of this is that the presence of affective focus may cause the constituent bearing the new-­information focus to become syntactically discontinuous. This was the case in (67), where part of the predicate is to the left of the copula in the surface structure and part of it remains to the copula’s right. The modern example (71), which fronts the predicative adjective harto but leaves behind its complement, can be regarded as a further illustration of this possibility, the resultant ‘extraposed’ type of structure arguably being peculiar to affective focus. In general, therefore, the fronting of predicative elements appears to be analogous to the quantifier-related fronting discussed in Sect. 2.4.3. The one common exception to this tendency is when the predicative element is an infinitival verb form, in which case fronting appears to correspond more often than not to a discourse-given interpretation. In (72) below, which illustrates a fairly recurrent pattern in the legal texts, the fronted infinitive soterrar (se) ‘be buried’ resumes an occurrence of the same item in the immediately preceding rubric (enclosed between braces), implying that it is topical rather than focal.

56 



I. E. Mackenzie

(72) {Ley..viija. En qual eglesia se deue cada uno soterrar.} Soterrar se deue cada un omne en el cimiterio daquella eglesia onde es parrochiano. ‘Law VIII. In which church each person must be buried. Each man must be buried in the cemetery of that church where he is a parishioner.’ (Libro de las leyes, fol. 84r)

Summing up Sect. 2.4 overall, the fronting of (non-interrogative) constituents in Old Spanish appears to result in the fronted item receiving either a topical interpretation (primarily in the discourse-given sense) or an affective focal interpretation. Phrasal categories, i.e. objects and PPs, may receive either type of interpretation, depending on their internal composition, whereas the effect of fronting in the case of predicative elements is normally that the fronted item is affectively focused. That fronting in Old Spanish correlates with the displaced item receiving a pragmatically marked interpretation is consistent with the assumption that the Old Spanish fronting structures are Ā-dependencies, i.e. that the items in question are moved from a VP-internal position to a landing site in the edge of CP.

2.5 Q  uantitative Evolution of Fronting Structures Turning now to the evolution of the foregoing patterns, it is undeniable that constituent fronting is a much less productive operation in modern Spanish than it was in Old Spanish (see Table 2.1 at the end of Sect. 2.2). On the other hand, the historical trajectory of this aspect of Spanish syntax since the thirteenth century has not been unidirectional, as can be seen from the ‘up-and-down’ curves in Fig. 2.1 (see Table A.1 in Appendix 1 for the corresponding numerical data). Rather than decreasing as Old Spanish evolves into early modern Spanish, the rate of fronting for each type of constituent actually increases, peaking in 1440 in the case of prepositional phrases, objects and non-­ finite verb forms, in 1460 in the case of predicative adjectives and nouns,

  Constituent Fronting: Focus, Discourse and Fashion 

20%

57

Prepositional phrase Predicative adj./noun Adverb

15%

Object Non-finite verb

10%

5%

0% 1260 1290 1320 1350 1380 1410 1440 1470 1500 1530 1560 1590 Fig. 2.1  Evolving rates of constituent fronting (1250–1609)

and in 1500 in the case of adverbs. This, together with the subsequent decline of fronting, delivers a series of bell curves which can be classified as ‘failed changes’ (see Sect. 1.2.3), although the alternative label of ‘transient state’ may be more appropriate in this particular case, given that the changes consist in shifts in the rates of occurrence of a set of existing constructions rather than the introduction of entirely new syntactic possibilities. I return to the ‘up-and-down’ nature of these curves in a moment, after considering the initial increase.

2.5.1 The Increase A key point to note about the increase is that it does not reflect any change in the underlying syntax. In Sect. 2.3.2 it was proposed that the syntactic properties of the fronting patterns associated with thirteenth-­ century Spanish could be characterized in terms of a ‘yes’ setting for the Restricted Verb Second parameter shown (in question format) in (46), repeated below:

58 



I. E. Mackenzie

(46) Does the finite verb move to C whenever a constituent undergoes Ā-movement to the edge of CP?

Now despite the quantitative movement shown in Fig. 2.1, the setting in respect of (46) does not appear to undergo any change across the medieval period. For exactly the same constraints as were apparent in the thirteenth century remain in place at the end of the Middle Ages. That is to say, the subject may not intervene between the fronted item and the finite verb; the fronted item must be either topical or affectively focused; and any weak object or reflexive pronoun must be proclitic. Examples (73) to (77) below, all from the period 1490–1509, illustrate this continuity with respect to the thirteenth century (the relevant portions of the examples are highlighted in bold). (73) abastar deue nuestra amistad mientre que la necessidad la forço estar sin sospecha. (Exemplario contra los engaños y peligros del mundo, fol. 69v) ‘Our friendship will endure as long as necessity has forced us to supress our suspicions.’

(74) y por esto dize que conuiene de añadir cautela enlas cosas gruessas y viscosas [. . .] De la viscosidad deze el Guillermo plazentino en el libro primero capitulo.cxlviij. assi [. . .] (Cura de la piedra, fol. 25r) ‘And for this reason, he says it is prudent to exercise caution with fatty and viscous substances . . . Of viscosity, Guillermo of Plasencia says the following in book 1, chapter CXLVIII . . .’



(75) Quarto es de entender que el causon mas peligroso es en los viejos: que fuerte es el fuego que no se amata conel agua. (Lilio de medicina, fol. 7r) ‘Fourthly, it should be understood that a sudden high fever is most dangerous in elderly people, because a fire that cannot be extinguished with water is strong indeed.’

(76) esta medicina le deuen fazer dos vezes al dia fasta que sea el cuero soldado: y en tanto guarden no echen la silla. (Libro de albeitería, fol. 30v) ‘This treatment should be administered to it twice per day until the skin has closed up; meanwhile, ensure the animal is not saddled up.’

  Constituent Fronting: Focus, Discourse and Fashion 

59

(77) humilmente te ruego que guies a mi sempronio: en manera que conuierta mi pena & tristeza en gozo: (Comedia de Calisto y Melibea, fol. 7r) ‘I humbly ask you to guide my Sempronio, in such a way that he turns my grief and sadness into joy.’

As examples like these show, the fronting structure as manifested at the outset of the early modern period is in all significant respects identical to the fronting structure as manifested in the texts of the Alfonsine period: at no time does it cease to involve V2 syntax. The growth in fronting captured in Fig. 2.1 therefore relates purely to the frequency of use and does not diagnose any underlying structural development. In syntactic terms, the observable growth in the frequency of use can be conceptualized as an increase in the activeness of C’s edge feature (see beginning of Sect. 2.4). However, the concept ‘activeness of C’s edge feature’ implicitly references a continuous variable whereas, according to Roberts (2007: 295), changes to the grammatical system can never be ‘clines, continua, squishes [or] the like’. In light of these considerations, it seems likely that the late medieval increase in V2 structures was primarily a stylistic tendency. This would be consistent with the fact that the fronting operation correlates with the displaced constituent receiving a pragmatically marked interpretation (see Sect. 2.4). As a consequence of this latter fact, V2 structures in the context of a language variety like Old Spanish lend themselves to expressive strategies; accordingly, it is easy to see how a vogue for V2 syntax could have arisen out of a desire for enhanced expressivity. With that in mind, it is interesting to compare the rates of growth in fronting across the five contexts represented in Fig. 2.1. As per the discussion in Sect. 1.2.2, the underlying growth model is assumed to be exponential growth in the relative probability or odds, meaning that rates of growth can be expressed as odds ratios (ORs).24 The OR is the factor by which the odds of the structure increase per unit of time, which here (as elsewhere in this book) is the decade. The first column of numbers in Table 2.4 shows the ORs corresponding to the growth in fronting during the period of increase in each of the five contexts under consideration. In each case the OR has been estimated by applying logistic regression analysis to the relevant data in Table A.1 in Appendix 1.

60 

I. E. Mackenzie

Table 2.4  Rates of increase and decline in constituent fronting Context Prepositional phrase Pred. adjective/ noun Adverb Object Non-finite verb form

Rising OR (per decade)

Decaying OR (per decade)

Reciprocal of rising OR

Discrepancy

1.086

0.970

0.921

5.3%

1.127

0.947

0.887

6.7%

1.112 1.103 1.188

0.906 0.959 0.884

0.899 0.907 0.842

0.7% 5.8% 5.0%

The highest OR in the table is the one corresponding to the fronting of non-finite verb forms, viz. 1.188, which equates to growth in the odds of 18.8% per decade. This particular OR seems to be something of an outlier in comparison to the ORs for the four other contexts shown, which range from 1.086 to 1.127. Among this latter group of contexts, the median OR is 1.108, from which the individual ORs deviate by 2% or less. In contrast, the OR for the fronting of non-finite verb forms deviates by 7.2% from this median value. Thus while the rates of increase in the four contexts of objects, PPs, adverbs and predicative adjectives/ nouns are very close to one another, the rate in the context of non-finite verb forms is appreciably faster. Now, according to the Constant Rate Effect of Kroch 1989 (see Sect. 1.2.1), a syntactic change, as externally manifested, will evolve at approximately the same rate in each of the various contexts in which the relevant structure occurs. The present case, as was observed above, does not involve a syntactic change in the strict sense of a modification of the grammar. However, it is comparable to the type of case in which structural change actually occurs in the sense that the observable phenomenon being measured is the same, namely change in the rate of use of a syntactic structure. This latter process reduces essentially to the spread of innovative linguistic behaviour through the speech community, a phenomenon whose basic dynamic presumably does not vary radically from one case to another. There is then no obvious reason to suppose that the principle of (approximate) cross-contextual uniformity in the rate of change – i.e. the Constant Rate Effect – should not hold fairly generally, including in the type of case examined here, in

  Constituent Fronting: Focus, Discourse and Fashion 

61

which quantitative movement is not accompanied by structural change. From that perspective, a plausible construction to put on the rates of increase shown in Table 2.4 is that the advance of fronting in the contexts of objects, PPs, adverbs and predicative adjectives/nouns corresponds to a single change process, which in turn is non-identical with the change process represented by the advance of fronting in the context of non-­ finite verb forms. This interpretation of the data implies that, within the context of the general increase in the use of V2 syntax that occurred towards the end of the Middle Ages, the growth in the fronting of non-finite verb forms was at least partly sui generis. It was in fact already implied at the end of Sect. 2.4 that the fronting of infinitives, which account for the majority of fronted non-finite verbs forms, is slightly anomalous in terms of the medieval fronting patterns overall. This follows from the fact that fronted infinitives in the thirteenth-century manuscripts are typically discourse-­ given, whereas the usual effect of fronting a predicative element is focal/ emphatic. Moreover, as was noted in connection with example (72), fronted infinitives commonly occurred in a resumptive capacity immediately after a legal rubric, suggesting that fronting in this particular case may have been to an extent formulaic. Therefore, if the late medieval increase in fronting overall reflects a desire for greater expressivity, in the specific case of non-finite verb forms the increase may have been assisted by an imitative tendency, modelled on the pattern enshrined in the Alfonsine legal texts. It is also possible that the fronting of non-finite verb forms may have benefitted from the late medieval vogue for Latinate constructions alluded to by Pountain (2001: 120) and others. For while this type of fronting occurs as a reflex of V2 syntax in a variety of languages, it is also reminiscent of Latin word order and could, as a consequence, have become identified by speakers as a Latinism.

2.5.2 The Bell Curves To go back to the bell-shaped nature of the curves in Fig. 2.1, the interesting question is whether their quantitative profile is consistent with the ‘accidental failure’ model or with the ‘inherent failure’ model advanced

62 

I. E. Mackenzie

by Postma (2010). In the latter model, the innovative trend is seen as being doomed from the outset to be reversed, due to the existence of some general syntactic principle with which the upward-trending structure is incompatible. The interaction of these two opposing forces yields a symmetric bell curve, modelled originally as S × (1−S), where S is a logistic function of time (see Sect. 1.2.2). In later work (Postma 2017), S × (1−S) is taken to be an approximation, but the assumption of symmetry remains unchanged. In contrast, in the accidental failure model we simply have a change or quantitative shift which turns out not to be permanent. The rising and declining flanks of the corresponding bell curve are thus not assumed to be related and there is no reason to expect the curve overall to be symmetrical. Visual inspection of the bell curves in Fig. 2.1 is not very conclusive as regards the issue of symmetry. However, we can also compare, for each bell, the upward-trending OR corresponding to the left-hand flank with the decaying OR corresponding to the right-hand flank. As was observed in Sect. 1.2.3, the downward equivalent of an upward curve with an OR of value x has an OR of 1/x. In other words, the downward curve’s OR must be equal to the reciprocal of the upward curve’s OR. Thus if the bells in Fig.  2.1 are to be deemed symmetrical, the right-hand ORs should be approximately equivalent to the reciprocals of the left-hand ORs. For each curve in Fig. 2.1, Table 2.4 shows the OR corresponding to the decaying, right-hand flank, together with the reciprocal of the OR for the rising (left-hand) flank and the discrepancy between these two values, expressed as a percentage. In light of these numerical values, the picture becomes rather clearer. In only one of the contexts, that of adverbs, can the declining flank be regarded as being an approximate mirror image of the rising flank, the OR of the former deviating from the reciprocal of the latter’s OR by only 0.7%. In the other cases, the corresponding discrepancy is 5% or more, implying that the rate of decline was appreciably different from the initial rate of increase. Notice, furthermore, that in each case the discrepancy consists in the rate of decline being lower than the rate of increase rather than higher. This is indicated by the fact that the decaying OR is always closer to the value 1 than is the reciprocal of the corresponding increase, an OR of 1 implying no change at all (see Sect. 1.2.3). On balance, then, fronting in the later Middle Ages appears

  Constituent Fronting: Focus, Discourse and Fashion 

63

to have increased at a faster rate than it declined in the early modern period. This asymmetry suggests that we should see the rise and decline of fronting as an accidental failure rather than as an inherent one. It is worth recalling at this juncture that the increase in fronting is non-­ structural, in the sense that its occurrence does not imply any change in the parameter settings of the Spanish language of the period. This implies that the phenomenon was essentially stylistic, with the fluctuation in the rate being regulated primarily by the internal dynamics of the propagating population. From that perspective, the case is analogous to when a meme such as a rumour is spread across a given community. Using a variant of the famous Susceptible-Infected-Removed (SIR) epidemiological model stemming from Kermack and McKendrick 1927, Piqueira (2010: 3–5) finds that the evolution of a rumour in a population is a function of two variables, viz. the probability with which ‘ignorants’ become ‘spreaders’ and the probability with which spreaders become ‘stiflers’. In the case in which these two probabilities are approximately equal, the evolution of the spreader population describes a symmetrical bell curve, which Postma (2017: 90) identifies with the bell curve of an inherently failing change, previously modelled by S × (1−S). Conversely, if the probability of an ignorant becoming a spreader is significantly higher than the probability of a spreader becoming a stifler, the spreader population describes an asymmetric bell curve, increasing rapidly and then declining relatively slowly (Piqueira 2010: 5, Figure 3). The latter pattern is obviously the one which is closest to the present case. Therefore, if the analogy with rumour propagation is on the right lines, the asymmetric nature of the bell curves in Fig. 2.1 may perhaps be interpreted as diagnosing a situation in which the rate of recruitment to the spreader population was greater than the rate of recruitment to the stifler population. The spreaders in this analogy would be those speakers who have acquired the fronting tendency and who, for reasons of prestige, influence, participation in tight social networks or whatever, cause other speakers to adopt it; while the stiflers would be speakers who have been exposed to the tendency but either failed to adopt it or abandoned it.25 Piqueira (2010: 2) observes that ‘spreading decays due to a forgetting process or because spreaders learn the rumour has lost its new value’. Accordingly, the apparent imbalance between the rates of recruitment to

64 

I. E. Mackenzie

and loss from the fronting paradigm suggests that any new or expressive value attaching to it was dissipated relatively slowly.

2.6 Conclusion The chapter has examined the emblematic fronting structures of Old Spanish from both structural and quantitative perspectives. The structures in question appear to be fairly typical instances of Ā-movement, with the fronted item occupying the edge of CP at surface structure – where it receives a topical or focal interpretation  – and the finite verb moving to C0. The fact that the verb ends up in C0 means that there is obligatory subject–verb inversion if an overt subject is present. Apparent subjects which precede both the finite verb and the fronted item are in fact left-dislocated topics, resumed by a phonologically null subject. Given the foregoing characterization, the fronting structures must be deemed to involve V2 syntax. However, the view that Old Spanish was a V2 language per se appears to be misplaced. Owing to the prevalence of VS(O) order – a pattern which remains an option even in the modern language – Old Spanish often exhibits ‘accidental’ V2 word order, which arises whenever a VS(O) unit is preceded by a circumstantial adverbial or a left-dislocated topic. The fortuitous nature of the V2 order in this latter type of case can be inferred from the facts (i) that subject–verb inversion is not obligatory and (ii) that the relevant structures exhibit enclisis rather than proclisis if a weak pronoun is present. The view that Old Spanish was a V2 language is likely to reflect a failure to distinguish true V2 structures (fronting of objects, prepositional complements, predicative adjectives etc.) from accidental ones. Somewhat unexpectedly, fronting increases in frequency towards the end of the Middle Ages, with the paradoxical result that early modern Spanish is more ‘medieval’ in this respect than is the Spanish of the Alfonsine period. Although the relevant parameter setting remains unchanged, the increase in fronting mimics a parametric change, in the sense that the growth rate is approximately uniform across four out of the five contexts in which the shift is manifested. This suggests that the Constant Rate Effect of Kroch 1989 has a more general applicability than

  Constituent Fronting: Focus, Discourse and Fashion 

65

is usually assumed. The subsequent decrease in fronting appears to be asymmetrical with the increase. Given that the fluctuation in the rate of fronting is stylistic rather than structural, it is plausible to suppose that the associated growth and decay can be modelled analogously to the spread of cultural memes such as rumours. Accordingly, and in light of Piqueira’s SIR-based mathematical model of rumour propagation, the asymmetric profile of the fronting-related bell curves is plausibly interpreted in terms of two variables, viz. the rate of recruitment to the spreader population and the rate of loss from that population. Specifically, the flatter right-hand flanks of the relevant bell curves imply that the value of the first variable was higher than the value of the second one, suggesting in turn that the expressive power of fronting was dissipated relatively slowly.

Notes 1. As used here, the term ‘fronted object’ excludes Clitic Left Dislocated (CLLD) structures in which a preverbal phrase appearing to be the object is resumed by a co-referential clitic. An example of CLLD from thirteenth century Spanish is provided by (i) below, where le resumes este nombre (the use of le rather than lo being an instance of medieval leísmo). (i) Y este nombre tomaron le del latin. (Estoria de España I, fol. 21v) ‘And this name, they took it from Latin.’ For CLLD, I take the verb’s syntactic object to be the resumptive clitic and not the preverbal phrase, as is implied by the analysis in Cinque (1990). Accordingly, the preverbal phrase should be treated as being external to the clause proper, analogously to the proper name Trump in the Left Dislocation structure in (ii) below. (ii) Trump, nobody believes that inveterate liar. As is discussed in more detail in Sect. 2.3.1, Old Spanish systematically distinguishes between CLLD and object fronting in the strict sense, showing (near) mandatory pronominal enclisis in the first case and mandatory proclisis in the latter.

66 

I. E. Mackenzie

2. This category excludes sentence-initial circumstantial elements, which can usually be separated from the remainder of the clause by a comma intonation. 3. The comparator data are drawn from a relatively small corpus (1,282,000 words) of mid- to late twentieth century novels, as follows: Ayer soñé con Valparaíso (Manuel Peña Muñoz), Casa de campo (José Donoso), Cien años de soledad (Gabriel García Márquez), Debajo de la cama (Mabel Pedrozo Cibilis), Delirios y certezas (Chiquita Barreto Burgos), El camino (Miguel Delibes), El laberinto (Augusto Casola), El lado de la sombre (Adolfo Bioy Casares), Hijo de ladrón (Manuel Rojas), La muerte de Artemio Cruz (Carlos Fuentes), La reivindicación del conde Don Julián (Juan Goytisolo), La tía Julia y el escribidor (Mario Vargas Llosa), Los hombres de a caballo (David Viñas), Los pies de barro (Salvador Garmendia), Manuel de historia (Marco Denevi), Paredes, un campesino extremeño (Patricio Chamizo), Sobre héroes y tumbas (Ernesto Sábato), Tiempo de silencio (Martín Santos). 4. As regards wh-fronting, this principle is partially relaxed for cómo ‘how’, at least in speech. There also appears to be a degree of dialectal variation, with speakers of Caribbean Spanish, for example, exhibiting non-inversion with fronted wh-objects under certain circumstances (see Zagona 2002: 71–72). 5. The examples are from main clauses, given that pronominal enclisis is almost completely excluded from finite embedded clauses in Old Spanish. For exceptions to this principle, see Castillo Lluch (1996: 142– 196) and Granberg (1988). 6. Technically, the items in strikethrough font represent phonologically null copies of the moved constituents. 7. The label ‘Ā’, pronounced ‘A-bar’, is at first glance rather opaque. The term was in fact coined as the obverse of ‘A-movement’ (for ‘argument movement’), the paradigm instance of which is the promotion of the thematic object into the preverbal subject, as in passive clauses. 8. Fontana (1993) appears to be the first linguist to explicitly characterize Old Spanish fronting as involving an Ā-dependency. Most subsequent authors appear to assume at least tacitly that an Ā-dependency is involved (see e.g. Benincà 2006; Sitaridou 2011; Poole 2013). 9. In a more detailed representation, the fronted object in (27) would have to be shown as moving first to the edge of the embedded CP, given that ‘long-distance’ Ā-movement is assumed to involve movement from one CP edge to another.

  Constituent Fronting: Focus, Discourse and Fashion 

67

10. The presence of an additional, CP-internal item between the left-dislocated phrase and the finite verb will trigger proclisis, however, as per generalization (31). The commonest instance of this is when the clause is negated, the negation marker no(n) then preventing the clitic from being CP-initial. Two other possibilities are illustrated by (i) and (ii) below, highlighted by Danford (2002) and Bouzouita (2008a) respectively. (i)

E esto los gentiles lo fizieron que fueron muy sabios omnes en estos saberes & en todos los otros. (General estoria I, fol. 27v) ‘And this, the gentiles did it – they were very wise in these matters and all others.’ (ii) asus mugieres e asos uassallos e todos los mayores de toda su tierra todos los catiuo e los leuo a babilonia. (Fazienda de ultramar, fol. 59r) ‘His women and his vassals and all the elders of all the land, he captured them all and he took them to Babylon.’ While both examples have CLLD, the relevant verbs, viz. fizieron in (i) and catiuo in (ii), are preceded not just by the left-dislocated “object” but also by the subject los gentiles in (i) and by the quantifier todos in (ii). Both of these latter items can be assumed to be CP-internal: los gentiles plausibly occupies the standard preverbal subject position (rather than being introduced as a left-­dislocated topic) and todos is a floated quantifier raised from a postverbal position (cf. catiuo los todos). The resumptive pronoun lo/los is thus prevented from being CP-initial, triggering proclisis as per generalization (31). 11. Ā-movement can also be posited for the much rarer operation of object movement to the middle field, illustrated by the position of el oro in (i) below. (i)

Et desta piedra usan mucho los orebzes; o aquellos que quieren el oro apurar. (Lapidario, fol. 6r) ‘And this stone is much used by goldsmiths; or those who want to purify gold.’

Chomsky (2008: 150) defines Ā-movement as movement either to the CP edge or to the edge of v*P, the transitive verb phrase. Finite

68 

I. E. Mackenzie

quieren in (i) can plausibly be regarded as selecting v*P as its complement, implying that the phrase el oro is indeed in the edge of v*P. 12. The coordinating conjunction variously written as &, et, y or e can be assumed not to count as a position-holder for V2 purposes (cf. MeyerLübke 1897: 315). 13. The VS(O) pattern should not be confused with VS# order, where # indicates the end of the clause. The latter pattern is associated with focus on the subject and is fully productive in modern Spanish. See Belletti (2004) for an analysis of VS# in Italian, which is analogous in this regard to Spanish. 14. This generalization is subject to analogous provisos to those mentioned in note 10 in respect of Clitic Left Dislocation. 15. The label ‘v*P’ in (47) designates the extended (transitive) verb phrase, i.e. the verb phrase proper together with its periphery. 16. For the specific case of VS(O) order with pronominal enclisis, the analysis put forward by Rivero (1993) can be adopted. Accordingly, the finite verb should be analysed as undergoing long head movement to C as a last resort, in order to prevent the clitic from occupying an illicit CP-initial position. 17. Note that ‘topic’ in this sense refers purely to the pragmatic status of the relevant phrase and does not in itself imply left-dislocation. The absence of any necessary connection between the two notions is illustrated in (i) and (ii) below. The phrase the film is a topic in both examples, but it is left-dislocated only in (ii). (i) The film I found really tedious. (ii) The film, I found it really tedious. 18. In the terminology introduced in Sect. 1.2.2, the figure of 19.110 is the odds ratio (for fronting) between the demonstrative objects and the remainder of the objects in the sample. 19. England (1980: 7) observes that they ‘refer almost without exception to an element which has already been mentioned’. 20. Analogous considerations apply to similar Latinate formulas from the pre-1200 period, such as et uobis dampnum restituat duplatum quod fecit ‘and to you should reimburse the damage he caused, doubled’ in (51). It seems likely that the preverbal placement of the object dampnum results from the fact that this item is covertly anaphoric, the concept of damage being introduced by the previous talk of acting against the deed of gift (Siquis . . . hanc meam donationem uoluerit refragare).

  Constituent Fronting: Focus, Discourse and Fashion 

69

21. Leonetti and Escandell Vidal (2009, 2010) have proposed that examples like (63) involve so-­called verum focus, i.e. the type of focus illustrated by English examples like (i) and (ii) below: (i) Liverpool DID score in the last minute. (ii) A python CAN eat a man whole. In examples like these, the effect of stressing the finite verb is to confirm the truth of the proposition, which implicitly has been called into question. The corresponding device in Spanish is the (stressed) particle sí ‘yes’, often used in conjunction with the complementizer que ‘that’. For example, the direct translation of (ii) into Spanish would be as in (iii): (iii) Un pitón SÍ que puede comer a un hombre entero. The fronting operation considered in the text does not in general seem to have the same effect as either English-style auxiliary stressing or the use of the particle sí in Spanish. Insofar as examples like (63) seem to insist on the sentence’s overall truth, this appears to result from a conversational implicature that attaches to the emphatic occurrence of the quantifier algo. This latter item, unlike proportional quantifiers such as mucho ‘much’ and poco ‘few’ or the universal quantifier todo ‘every/all’, has a purely existential meaning and in that sense is relatively empty of semantic content. The emphasis which comes from fronting this item is thus easily viewed as being transferred to the verb phrase generally, leading to an impression that it is the sentence overall, or its truth, which is being emphasized rather than just the quantifier. 22. Affective focus should also be distinguished from contrastive focus, which involves the exhaustive ruling out of competing alternatives and is often associated with a strong pitch accent on the focused constituent, as in (i) below: (i) El PEUGEOT vendimos (y no el Seat). ‘It was the Peugeot we sold (not the Seat).’ For this additional type of focus, see Drubig (2003) and È Kiss (1998). 23. Lambrecht and Michaelis (1998) use the term ‘ratified topic’ to refer to a topic which is already salient in the discourse. Ratified topics are usually manifested as unaccented or phonologically null constituents.

70 

I. E. Mackenzie

24. ORs can be converted to slopes (in the sense of Kroch 1989) by taking the natural logarithm of the OR. 25. The ignorants would be speakers who have not yet been exposed to the fronting tendency.

References Alexiadou, Artemis, and Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory 16: 491–539. Alonso, Martín. 1962. Evolución sintáctica del español. Madrid: Aguilar. Barbosa, Pilar. 2009. Two kinds of subject. Studia Lingüística 63 (1): 2–58. Belletti, Adriana. 2004. Aspects of the low IP area. In The structure of CP and IP.  The cartography of syntactic structures, ed. Luigi Rizzi, vol. 2, 16–51. New York: Oxford University Press. Benincà, Paola. 2006. A detailed map of the left periphery in medieval Romance. In Crosslinguistic research in syntax and semantics: Negation, tense and clausal architecture, ed. Raffaella Zanuttini et  al., 53–86. Washington, DC: Georgetown University Press. Bossong, Geor. 2006. La sintaxis de las Glosas Emilianenses en una perspectiva tipológica. In Actas del VI Congreso Internacional de Historia de la Lengua Española, ed. José Jesús de Bustos and José Luis Girón, vol. 1, 529–544. Madrid: Arco Libros. Bouzouita, Miriam. 2008a. At the syntax-pragmatics interface: Clitics in the history of Spanish. In Language in flux: Dialogue coordination, language variation, change and evolution, ed. Robin Cooper and Ruth Kempson, 221–263. London: College Publications. Castillo Lluch, Mónica. 1996. La posición del pronombre átono en la prosa hispánica medieval. Doctoral dissertation, Universidad Autónoma de Madrid. Chomsky, Noam. 2008. On phases. In Foundational issues in linguistic theory, ed. Robert Freidin et al., 133–166. Cambridge, MA: MIT Press. Cinque, Guglielmo. 1990. Types of Ā-dependencies. Cambridge, MA: MIT Press. Cruschina, Silvio, and Ioanna Sitaridou. 2009. From Modern to Old Romance: The interaction of information structure and word order. Paper presented at XI Diachronic Generative Syntax Conference (DiGS), August, Universidade de Campinas.

  Constituent Fronting: Focus, Discourse and Fashion 

71

Danford, Richard K. 2002. Preverbal accusatives, pronominal reduplication and information packaging: A diachronic analysis of Spanish. Doctoral dissertation, Ohio State University. Drubig, Hans Bernhard. 2003. Toward a typology of focus constructions. Linguistics 41 (1): 1–50. Eide, Kristine G., and Ioanna Sitaridou. 2013. Contrastivity and information structure in the old Ibero-Romance languages. In Information structure and syntactic change in Germanic and Romance languages, ed. Kristin Bech and Kristine G. Eide, 377–412. Amsterdam: John Benjamins. England, John. 1980. The position of the direct object in Old Spanish. Journal of Hispanic Philology 5 (1): 1–23. Fassi Fehri, Abdelkader. 1993. Issues in the structure of Arabic clauses and words. Dordrecht: Kluwer. Fischer, Susann. 2014. Revisiting stylistic fronting in Old Spanish. In Left sentence peripheries in Spanish: Diachronic, variationist and comparative perspectives, ed. Andreas Dufter and Álvaro S. Octavio de Toledo, 53–76. Amsterdam: John Benjamins. Fontana, Josep. 1993. Phrase structure and the syntax of clitics in the history of Spanish. Doctoral dissertation, University of Pennsylvania. Gallego, Ángel. 2007. Phase theory and parametric variation. Doctoral dissertation, Universitat Autònoma de Barcelona. Granberg, Robert. 1988. Object pronoun position in medieval and early modern Spanish. Doctoral dissertation, University of California, Los Angeles. Hernanz, María Lluisa. 2001. ¡En bonito lío me he metido!: notas sobre la afectividad en español. Moenia 7: 93–109. ———. 2006. Emphatic polarity and C in Spanish. In Studies in Spanish syntax, ed. Laura Brugè, 105–150. Venice: Università Ca’ Foscari. Holmberg, Anders. 2015. Verb second. In Syntax – Theory and analysis. An international handbook, ed. Tibor Kiss and Artemis Alexiadou, chapter 12. Berlin: Walter de Gruyter (e-book). Kermack, William O., and Anderson G. McKendrick. 1927. Contributions of mathematical theory to epidemics. Proceedings of the Royal Society Series A 115: 700–721. Kiss, Katalin É. 1998. Identificational focus versus information focus. Language 74: 245–273. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244.

72 

I. E. Mackenzie

Lambrecht, Knud. 1994. Information structure and sentence form. Topic, focus, and the mental representations of discourse referents. Cambridge: Cambridge University Press. Lambrecht, Knud, and Laura A. Michaelis. 1998. Sentence accent in information questions: Default and projection. Linguistics and Philosophy 21 (5): 477–544. Lapesa, Rafael. 1981. Historia de la lengua española, 9th ed. Madrid: Gredos. Leonetti, Manuel, and Victoria Escandell Vidal. 2009. Fronting and verum focus in Spanish. In Focus and background in Romance languages, ed. Andreas Dufter and Daniel Jacob, 155–204. Amsterdam: John Benjamins. ———. 2010. Las anteposiciones inductoras de foco de polaridad. In La renovación de la palabra en el bicentenario de la Argentina: los colores de la mirada lingüística, ed. Victor M.  Castel and Liliana Cubo de Severino, 733–743. Mendoza: Editorial FFyL, UNCuyo. Mackenzie, David. 1997. A manual of manuscript transcription for the dictionary of the Old Spanish Language. (5th edition, revised and expanded by Ray Harris-Northall). Madison: Hispanic Seminary of Medieval Studies. Mackenzie, Ian. 2010. Refining the V2 hypothesis for Old Spanish. Bulletin of Hispanic Studies 87 (4): 379–396. Mackenzie, Ian, and Wim van der Wurff. 2012. Relic syntax in Middle English and Medieval Spanish: Parameter interaction in language change. Language 88 (4): 846–876. Martínez Gil, Fernando. 1989. Las inversiones del orden de palabras en el Romancero. Hispania 72 (4): 895–908. Menéndez Pidal, Ramón. 1926. Orígenes del español, 6th ed. Madrid: Espasa-Calpe. Meyer-Lübke, Wilhelm. 1897. Zur Stellung der tonlosen Objektspronomina. Zeitschrift für romanische Philologie 21 (3): 313–334. Ordóñez, Francisco. 1997. Word order and clause structure in Spanish and other Romance languages. Doctoral dissertation, CUNY. Piqueira, José Roberto. 2010. Rumor propagation model: An equilibrium study. Mathematical Problems in Engineering, 631357. https://www.hindawi.com/ journals/mpe/2010/631357/ Poole, Geoffrey. 2013. Interpolation, verb-second and the low left periphery in Old Spanish. Iberia: International Journal of Theoretical Linguistics 5 (1): 69–98. Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins.

  Constituent Fronting: Focus, Discourse and Fashion 

73

———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Pountain, Christopher J. 2001. A history of the Spanish language through texts. London: Routledge. Quer, Josep. 2002. Edging quantifiers: On QP-fronting in Western Romance. In Romance languages and linguistic theory 2000, ed. Claire Beyssade et al., 253–270. Amsterdam: John Benjamins. Rivero, María Luisa. 1991. Clitic and NP climbing in Old Spanish. In Current studies in Spanish linguistics, ed. Héctor Campos and Fernando Martínez Gil, 241–282. Washington, DC: Georgetown University Press. ———. 1993. Long head movement versus V2, and null subjects in Old Romance. Lingua 89: 217–245. Rizzi, Luigi. 1990. Speculations on verb-second. In Grammar in progress: Essays in honour of Henk van Riemsdijk, ed. Joan Mascaró and Marina Nespor, 375–386. Groningen: Foris. Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Sitaridou, Ioanna. 2011. Word order and information structure in Old Spanish. Catalan Journal of Linguistics 10: 159–184. Tobler, Adolf. 1875. Review of Jules Le Coultre, De l’ordre des mots dans Crestien de Troyes. Göttingische Gelehrte Anzeigen 34: 1057–1082. Wolfe, Sam. 2015. The nature of Old Spanish verb second reconsidered. Lingua 164: 132–155. Zagona, Karen. 2002. The syntax of Spanish. Cambridge: Cambridge University Press. Zubizarreta, María Luisa. 1998. Prosody, focus and word order, Linguistic inquiry monograph 33. Cambridge, MA: MIT Press.

3 Clitic Linearization: A Tale of Successful and Failed Changes

3.1 Introduction Clitic linearization – the linear position of weak personal pronouns in relation to the verb – is an important variable in the history of Spanish, the transition from the high medieval system to the modern one having generated a wide array of linguistic studies. While the medieval corpus presents a limited number of examples in which the weak pronoun could be non-adjacent to its anchoring verb, the principal locus of interest is the opposition between proclisis, in which the pronoun occurs immediately to the left of its verb, and enclisis, where the pronoun is positioned to the verb’s immediate right. As in the majority of other Romance languages, the proclisis–enclisis alternation in the High Middle Ages was primarily a function of syntax, perhaps representing the conventionalization of a previous stage at which phonology played a critical role. Over the course of the later Middle Ages and early modern period, the alternation comes to be regulated primarily by verbal morphology. In the quantitative dimension this systemic reorganization is manifested as a constantly evolving ratio between proclisis and enclisis, which follows differing trajectories according to context before eventually attaining a © The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_3

75

76 

I. E. Mackenzie

steady state in the early modern period. In the fundamental context of finite main clauses, the advance of proclisis across the Middle Ages is steady and relentless, as is normal in cases of successful change; elsewhere, however, we find bell curves corresponding to transient states or failed changes. The interrelationship between the evolution in the various contexts provides an important window onto the geometry of syntactic change, with implications for both the Constant Rate Effect (see Sect. 1.2.1) and for the way in which successful changes interact with unsuccessful ones.

3.2 An Early Synchronic Overview As is discussed in detail in Sect. 3.3 below, the ratio of proclisis to enclisis undergoes important changes across the Middle Ages. Accordingly, any generalizations about the factors governing synchronic variation must be anchored to a specific period. The present section gives an overview of the situation at the first two data points in the corpus, covering the period 1250–1289.1

3.2.1 Comparison of the Different Clause-Types Initially, Old Spanish main clauses can be grouped with (post-­ prepositional) infinitival clauses, in that both proclisis and enclisis may be found in the two contexts. It should be noted, however, that the alternation in clitic placement appears to have been quite tightly constrained in the first context (see Sect. 3.2.2) but to have been relatively random in the second. The broad parallel between the two contexts is illustrated in (1) to (4) below, the relevant clitic and verb being highlighted in bold in each case. Proclisis (1) & assi los faras por orden fatal seseno. (Cánones de Albateni, fol. 24r) ‘And you will do them in this way, in order, until you reach the sixth one.’

  Clitic Linearization: A Tale of Successful and Failed Changes 

77

(2) Ca estos siempre punnan de los embargar que se no saluen. (Libro de las leyes, fol. 78r) ‘For they continually strive to prevent them from being saved.’

Enclisis (3) & llaman los todos comunalmientre cauallos. mas los sus nombres derechos son caualleros. (Libro de ajedrez, dados y tablas, fol. 3r) ‘And everyone usually calls them horses, but their correct name is “knight”.’ (4) esto significa que los enemigos del Rey se moueran. et uenran contra las uillas del Rey por acercar las. & conquerir las. (Libro de las cruzes, fol. 41v) ‘This means that the enemies of the king will mobilize and they will attack the king’s towns in order to surround them and to conquer them.’

As regards bare infinitives, found most obviously in clitic climbing structures, these lack proclisis altogether in the thirteenth century, a situation which appears to continue throughout the medieval period. In other words, the ‘medial’ clitic position illustrated by the bold portion of (5) below should be understood as involving enclisis on the finite verb rather than proclisis on the infinitive. (5) pero sil uyan acorrer con aquellas cosas que son pora esto; pueden le estorcer de muerte. (Lapidario, fol. 20r) ‘But if they attend to him with those things that are suitable for this purpose, they can save him from death.’

The generally enclitic status of weak pronouns in the above medial position can be inferred from what happens under negation. As is observed in Sect. 3.2.2 below, a finite verb negated by no(n) has obligatory proclisis if a weak pronoun is present. This means that the string ‘no(n)–VFIN–clitic–VINF’ (i.e. with no(n) negating VFIN, the finite verb) is only possible if the pronoun is anchored, proclitically, to the infinitive (VINF). The string thus serves as a diagnostic (arguably the sole diagnostic) for that specific structure. Now Mackenzie (2017: 138, note 8) reports finding only one token of this string in the entire period 1250–1600, as against 6878 tokens for the alternative

78 

I. E. Mackenzie

string ‘no(n)–clitic–VFIN–VINF’. Moreover, the solitary token of ‘no(n)– VFIN–clitic–VINF’ occurs in the late and somewhat standardized incunabulum of the Siete Partidas (Seville 1491), which is not the ideal locus for the sole token purporting to attest to the existence of a specific syntactic structure in the medieval language. There is therefore no real evidence that the medial clitic in the clitic climbing structure illustrated by (5) was anything other than enclitic (on the finite verb). Note that this only applies to bare infinitivals. In the case in which the subordinate infinitive was governed by a preposition, proclisis on the infinitive was available. This is demonstrated by the fact that, unlike in the bare infinitival case just discussed, the clitic could remain in the medial position under negation: (6) & aun los sos pastores non dexauan de les fazer el so mal por tod esto. (General estoria IV, fol. 148r) ‘And even their shepherds did not stop doing them harm for this reason.’

Placement of the clitic to the left of the preposition, as in (7) below, should then be assumed to indicate enclisis on the finite verb, exactly as in the prepositionless example (5). (7) & estado yo enel mont los quarenta dias como dix. oyo me el sennor esta uez por ti. & dexo te de destruyr. (General estoria I, fol. 325v) ‘And after I had been on the mountain for forty days, as I said, the Lord heard me this time on your behalf and stopped destroying you.’

Turning now to embedded finite clauses, these have systematic proclisis from at least the beginning of the thirteenth century. An illustration is given in (8) below. (8) Et por ende amester que qui este libro oyere & enell leyere. que pare mientes atodas estas cosas & ques tenga bien con dios & quel ruegue el pida merçet quel guie & quel endereçe enello. (Picatrix, fol. 25r) ‘And therefore it is necessary that whoever listens to this book and reads in it, must think about all these things and should follow God and should pray to him and ask him for grace that he may guide him and assist him in this.’

  Clitic Linearization: A Tale of Successful and Failed Changes 

79

While (8) illustrates the norm for subordinate finite environments, including embedded coordination, the latter context occasionally exhibits enclisis, as example (9) demonstrates:

(9) Despues de tod esto mando que tomasse esse sacerdot toda la otra sangre que fincaua. & quela sacasse fuera & esparziesse la al pie dell altar o quemauan los sacrifficios que eran en la entraba dela tienda. (General estoria I, fol. 228v) ‘After all of this he instructed that the priest collect the remaining blood and that he take it outside and pour it at the base of the altar where offerings that were at the doorway of the tent were burned.’

Given the overwhelming predominance of proclisis in embedded finite clauses, the enclisis observed in examples such as (9) is likely to be either ‘off-grammatical’ (Postma 2010) or stylistic in nature, perhaps reflecting interference from main clause &, which in the thirteenth century exhibits a very strong correlation with enclisis (see the discussion in Sect. 3.2.3.2 below). As regards gerundial clauses, these show only enclisis in the thirteenth century, except under negation, in which case proclisis is nearly categorical:

(10) El segundo iuego dar la xaque con el Roque prieto que esta en la casa del alffil prieto poniendo lo en la tercera casa del alffil blanco & tomar lo a el Rey blanco por fuerça (Libro del ajedrez, dados y tablas, fol. 42r) ‘The second move will put it [the white king] in check with the black rook, which is in the black bishop’s square, by moving it to the white bishop’s third square, where the white king is obliged to take it.’



(11) Por camio. & por uendida puede otrossi passar. no lo camiando. ni lo uendiendo. Por si apartadamientre. mas de buelta con todas las otras cosas que en aquel logar ouiesse. (Libro de las leyes, fol. 90v) ‘It may also be transferred in a swap or a sale, without swapping it or selling it separately, but rather as part of all the other things that there are in that place.’

80 

I. E. Mackenzie

3.2.2 The Basic Regulating Principle in Finite Contexts Given the foregoing remarks, it can be seen that the principal loci for variation in terms of the proclisis–enclisis dichotomy were main clauses and post-prepositional infinitivals. (Recall that bare infinitivals did not constitute a variation context, as they lacked proclisis altogether.) Setting aside discussion of post-prepositional infinitivals for the moment, it is usually assumed that the distribution of enclisis and proclisis in main clauses fell under the descriptive principle known as the Tobler–Mussafia Law (Tobler 1875; Mussafia 1886). In Chap. 2, this principle was given the more precise syntactic characterization shown as generalization (12) below (= (31) from Chap. 2), in which ‘CP’, standing for ‘Complementizer Phrase’, is the standard Generative label for the full clause.

(12) The weak pronoun is enclitic (on the finite verb) if and only if it would otherwise immediately follow a CP boundary.

The prototypical illustrations of this generalization are delivered by examples such as (13) to (17) below, the relevant portion in each case being highlighted in bold.

(13) XXXII Pusieron te por mantenedor; non te quieras exaltar. (General estoria IV, fol. 269r) ‘Chapter 32. You were placed here as a defender; do not succumb to rage.’



(14) mas despues retienen bien so ceuo e tuellen lo bien. (Libro de las animalias, fol. 85r) ‘But afterwards they retain their food and they excrete it well.’



(15) Et si la mezclaren con mestranto; faze se della emplastro muy bono pora la ferida del alacran. (Lapidario, fol. 72r) ‘And if it is mixed with wild mint, it makes a very good poultice for a scorpion sting.’

(16) Non les pudo dar la tierra queles prometiera. (General estoria I, fol. 213r) ‘He was unable to give them the land he had promised them.’

  Clitic Linearization: A Tale of Successful and Failed Changes 



81

(17) Ca el juyzio que es cosa muy derecha manifiestamientre se deue dar. & no en encubierto. (Libro de las leyes, fol. 48v) ‘For a trial, which is a very just event, must take place openly and not in obscurity.’

In (13) and (14) the relevant verb  +  clitic complex can be analysed as being immediately adjacent to the left boundary of its containing CP, as is shown in (18) and (19) below; preverbal positioning of the clitic would thus be at variance with generalization (12).

(18) [CP pusieron te por mantenedor]; [CP non te quieras exaltar]



(19) mas despues [CP [CP retienen bien so ceuo] e [CP tuellen lo bien]]2

In example (15), the enclisis-triggering CP boundary is the right-hand one belonging to the embedded si-clause, as is shown in (20) below.

(20) et [CP [CP si la mezclaren con mestranto]; faze se della emplastro muy bono pora la ferida del alacran]

As regards the proclitic examples (16) and (17), these are structures which unambiguously manifest a CP-internal item that prevents the clitic from being adjacent to any CP boundary. In (16) the item in question is the negation marker, standardly assumed to occupy a position in the edge of TP, the ‘Tense Phrase’ or basic clause, which itself is nested inside CP. The item which has the same effect in (17) is the manner adverb manifiestamientre, which would normally occur verb phrase-internally but here has undergone Ā-movement to the CP edge (see Sect. 2.3.1). The relevant boundaries are shown in (21) and (22) below (in the latter, I assume the apparent subject el juyzio to be a clause-external topic (see Sect. 2.3.1.2)).

(21) [CP [TP Non les pudo dar la tierra queles prometiera]]



(22) Ca el juyzio que es cosa muy derecha [CP manifiestamientre se deue dar]. & no en encubierto

It is worth noting, in passing, that a CP boundary which triggers enclisis is often (though not always) associated phonologically with a comma-­

82 

I. E. Mackenzie

style intonation break, or at least the possibility of such a break. In example (15)/(20), this is actually reflected in the manuscript by the presence of a semicolon after the clause si la mezclaren con mestranto. In addition, sentence-initial pusieron te in (13)/(18) can be assumed to follow a prosodic break and (14)/(19) reads very naturally with a break either just before or just after the coordinating conjunction e. It turns out, in fact, that one of the strongest clitic-related correlations in the early manuscripts is between the presence of an orthographic stop (e.g. a semicolon, a colon or a full stop) and enclisis. In the 1250–1289 data, for example, while the rate of enclisis in main clauses overall is 75.4% (21,327/28,275), the rate in cases in which the clitic would otherwise immediately follow a written stop of some kind is 98.4% (3448/3503). Comparing these figures, it can be calculated that the presence of the stop increases the odds of enclisis (in the 1250–1289 data) by a factor of 24.17. This points to a very strong pause-related effect, if, as seems plausible (cf. Mackenzie 1997: 11), written stops in the manuscripts typically diagnose intonation breaks (or the possibility thereof ). However, given Kuchenbrandt’s (2016) finding that clitic linearization in Old Spanish was not primarily driven by phonology, it would probably be unwise to view the pause-­ related effect as evidence that the weak pronoun was phonologically enclitic. The effect is more plausibly understood as a by-product of the fact that CP boundaries are very often manifested prosodically.

3.2.3 Variation Versus Syntactic Ambiguity 3.2.3.1  Counter-Systemic Instances of Proclisis The previous section alludes to a governing principle, identified in the philological literature as the Tobler–Mussafia Law and expressed here as generalization (12), which appears to have regulated the distribution of proclisis and enclisis in finite clauses in the High Middle Ages. Even in the thirteenth century, however, examples of proclisis are forthcoming in contexts in which, given the generalization in (12), enclisis is expected. This implies that a state of synchronic variation was already in existence, a fact recognized by Fontana

  Clitic Linearization: A Tale of Successful and Failed Changes 

83

(1993) and Bouzouita (2008a) among others. Two early examples of proclisis which are impossible to reconcile with (12) are given below.

(23) Los otros assolauan lo que podien & quemauan sus casas por que non fallassen y njnguna cosa los enemjgos. Los matauan sus fijos mismos & sus mugieres por que non fuessen sieruos delos judios. (General estoria V, fol. 136v) ‘The others raised to the ground whatever they could and burned their houses so the enemy would find nothing there. They killed their own children and wives to prevent them becoming enslaved by the Jews.’



(24) Mas depues que fuere babtizado; lo deuen aun ungir otras dos uezes con crisma. (Libro de las leyes, fol. 6v) ‘But after he is baptized, they must still anoint him with chrism two more times.’

In (23), proclitic los is sentence-initial, hence necessarily adjacent to the left boundary of its containing CP. In (24), proclitic lo immediately follows the right-hand boundary of the embedded CP depues que fuere babtizado. Neither example offers any scope in terms of possible syntactic ambiguity and it simply has to be accepted that these particular clitic tokens are not subject to the principle expressed here as (12); that is to say, they are not Tobler–Mussafia clitics. Instead, the unconstrained ­proclisis of these examples foreshadows the situation found in modern Spanish, in which weak pronouns in (tensed) finite clauses are necessarily or, as I will say, inherently proclitic. Examples like (23) and (24) thus point to an incipient state of sociolinguistic variation, in which an underlying variable – call it ‘weak pronoun’ – has both a conservative variant and an innovative variant. The former can be identified with the then dominant Tobler–Mussafia clitic, known for its capability to be proclitic or enclitic depending on the precise context, while the latter can be identified with the inherently proclitic item which is found in all finite clauses (other than positive imperatives) in the modern language. One should, however, be careful to draw a distinction between true synchronic variation, which implies sociolinguistic competition between

84 

I. E. Mackenzie

two equivalent but competing variants (in this case, the two variants of the weak pronoun just identified), and variation which is actually the reflex of syntactic ambiguity. The latter phenomenon is not uncommon in the context of medieval clitic linearization, given that the distribution of CP boundaries is not always obvious from surface strings of words. Sections 3.2.3.2, 3.2.3.3, and 3.2.3.4 below examine some of the most important cases in which such ambiguity arises, cases which if considered purely in terms of surface form may lead to an overestimation of the variability of medieval clitic placement.

3.2.3.2  Coordinate Structures A prototypical instance of the relevant ambiguity arises in connection with coordination. For example, looked at purely in terms of surface form, the clitic placement after the second & in the coordinated structures below appears to vary randomly, (25) exhibiting enclisis and (26) proclisis (in both cases the relevant & links two main clauses).

(25) & si las fizieren no ualen. & puede las el obispo desfazer. (Libro de la leyes, fol. 87v) ‘And if he makes any [donations of church property], they are invalid, and the bishop can undo them.’



(26) & a las uezes otrossi maltroxieron los de carthago a los Romanos & los uencieron. (General estoria IV, fol. 254v) ‘And sometimes, as well, the Carthaginians attacked the Romans and defeated them.’

Recall, however, from generalization (12), that medieval clitic linearization is, in principle, a function of the locus of the CP boundary. Therefore, given that coordinating conjunctions are not restricted to conjoining specific syntactic categories of constituent, clitic placement after such a conjunction can be expected to be dependent on whether the coordinator selects CP as its complement or whether it selects some smaller unit, with enclisis arising in the first case and proclisis in the second. Rather than showing random variation, the two examples (25) and (26) appear to illustrate precisely this pattern.

  Clitic Linearization: A Tale of Successful and Failed Changes 

85

Looking at (25) first, it can be observed that the coordinate clause puede las el obispo desfazer is a complete clause, hence an instance of CP. The enclitic placement of las is thus perfectly predictable, given that the pronoun would otherwise be CP-initial. In contrast, the corresponding unit in (26) – i.e. los uencieron – shares its subject with the preceding clause and hence should be analysed as exhibiting subject ellipsis, analogously to the structure used in the English translation ‘(and) defeated them’. The governing & must therefore be analysed as conjoining two partial clauses rather than two complete clauses, with the common subject los de carthago being structurally outside the conjoined unit. In this arrangement, & does not select CP as its complement but rather some sub-CP unit. What this unit is becomes clearer if the relevant elements in the syntactic derivation of (26) are made explicit. Note first of all that the word order is VSO, which, according to the standard analysis of that pattern (see end of Sect. 2.3.2), is derived by the finite verb moving out of the verb phrase and across the subject to reach its surface position at the head of TP, the tense phrase or basic clause. In respect of (26), this implies a structure along the lines of (27) below, where maltroxieron is a phonologically null copy or ‘trace’ of maltroxieron at its original merge site inside VP.

(27) & [CP a las uezes otrossi [TP maltroxieron los de carthago [VP maltroxieron a los Romanos] & [VP los uencieron]]]

As can be seen, what is coordinated by the second & is VP rather than CP. Accordingly, proclitic los is not CP-initial and there is no incompatibility with generalization (12). In the data for the period 1250–1289, proclisis has a rate of 2.1% (260/12,421) in main clauses in which the word order would otherwise be ‘coordinator–verb–clitic’. This is low but unexpected if main clause coordinators are assumed to correlate 100% with enclisis, as is sometimes suggested. In light of the discussion just above, the majority of the unexpected cases of proclisis can be attributed to the fact that the relevant constraint, viz. generalization (12), references CP boundaries, whereas coordinate clauses need not have the status of CP.

86 

I. E. Mackenzie

3.2.3.3  Circumstantial Adverbials A locus of greater variation is comprised by the post-circumstantial adverbial context, where the relevant proportions in the period 1250–1289, based on a selection of adverbials which commonly occur in sentence-­ initial position,3 are 50.9% proclisis (259/509) versus 49.1% enclisis (250/509). Here again, though, syntactic ambiguity seems likely to account for at least some of the variation, given that a single lexical item may be capable of being projected either as an initial circumstantial element or as a VP-internal adjunct, which may or may not subsequently be moved to the edge of CP.  Given generalization (12), the first case is expected to correlate with enclisis and the second, if the adverbial is raised to the CP edge, with proclisis. The distinction can be illustrated using the adverbial despues, which is compatible with both types of clitic placement, as (28) (=  (29) from Chap. 2) and (29) below demonstrate: (28) Despues troxieron le otras mugeres Egipçianas otrossi mas dize Josepho que non quiso mamar njnguna njn llegar se a ellas. (General Estoria I, fol. 136v) ‘Afterwards they brought him other Egyptian women as well, but Joseph says that he refused to be breastfed by any of them or to go near them.’

(29) Otrossi a Sarra llamaron primera mientre este otro nombre Saray. & despues le dixieron Sarra. (General estoria I, fol. 42r) ‘Also, Sarah was first of all called by the other name Sarai and they called her Sarah after that.’

Despues as it occurs in (28) is peripheral to the main predication, providing a context for it but not being part of the actual assertion. Given this fact, and given that peripheral or circumstantial elements of this kind fail to trigger a V2 effect (cf. examples (40) to (42) in Chap. 2), it reasonable to analyse the adverbial as being external to the core CP. In contrast, despues in (29) should be analysed as being part of the predicate of its containing clause, modulo the fact that it has undergone fronting to a preverbal position. In other words, the sequence & despues le dixieron

  Clitic Linearization: A Tale of Successful and Failed Changes 

87

Sarra in (29) is equivalent to & dixieron le Sarra despues, except that despues is fronted in the former but remains in situ in the latter (with corresponding enclisis of le). In situ occurrences of despues are common enough in the Old Spanish manuscripts, one illustration being (30) below, where the adverbial is obviously part of the predicate focus within its relative clause.

(30) E que auran los que y fueren mugeres escosas. non destas que son agora en este mundo; mas dotras que uernan despues. (Estoria de España I, fol. 166r) ‘And that those who go there will have virgin women, not of the type that are in the world now, but of another type that will come later.’

Assuming that the fronting operation posited for despues in (29) is an instance of Ā-movement, the adverbial in that type of use is CP-internal; hence the preverbal placement of le is not at variance with generalization (12). The alternation in clitic linearization across (28) and (29) can thus be regarded as a function of the structural locus of the adverbial despues: external to the CP containing the finite verb and clitic in (28) but internal to it in (29). The structural alternation just illustrated can be assumed to apply to other adverbials which behave like despues as regards their effect on clitic placement, implying that at least some of the proclisis encountered in this context is explicable in terms of the principle captured in generalization (12). On the other hand, given examples such as the earlier (23) and (24), which attest to the presence already in the thirteenth century of a competing inherently proclitic weak pronoun, it would be rash to assume that all of the proclisis encountered in the post-adverbial context can be reconciled with (12).4

3.2.3.4  Preverbal Subjects A third important finite context in which enclisis alternates with proclisis is the post-subject environment, which is extensively discussed by Wanner (1991: 342–345) and Granberg (1999) among others. Two illustrative examples are given below, the first showing enclisis and the second proclisis.

88 

I. E. Mackenzie



(31) Ell astrolabio puede se fazer de todas aquellas cosas de que se faz ell alcora. (Libros del saber de astronomía, fol. 66r) ‘The astrolabe can be made from any of those things from which the sphere is constructed.’



(32) ca esta melezina les faze pro. pora muchas malabtias. que les acaeçe. en los logares que non pueden fallar las otras melezinas que las guareçen. (Libro de las animalias, fol. 72r) ‘For this medicine does them good in the case of many ailments which affect them in situations in which the other appropriate medicines cannot be obtained.’

In the dataset for the period 1250–1289, the proportions of enclisis and proclisis in this context are 68.0% (930/1368) and 32.0% (438/1368) respectively. At first sight, the clear preponderance in favour of enclisis appears to be at variance with generalization (12), given that preverbal subjects are usually assumed to be CP-internal, the Chomskyan structural Subject Position traditionally being identified with the left edge of TP. However, this assessment overlooks the fact that Old Spanish (like modern Spanish) was a pro-drop language, with pronominal subjects tending to receive a null phonological spell-out. This state of affairs leaves open the possibility that a preverbal noun phrase which appears to be the subject is in reality a left-dislocated topic, the true subject being a silent pronoun which referentially resumes the noun phrase in question. This case is revealed transparently when the left-dislocated topic and the silent resumptive pronoun fail to agree in terms of their grammatical features, as in (33) below, where los latinos is a third-person noun phrase but llamamos must be inferred to have a first-person subject.

(33) & dize en este lugar la glosa que aquel aruol çino. era al que los latinos llamamos lentisco. (General estoria IV, fol. 70r) ‘And the gloss says in this place that that schínos tree was the one that we, the Romance speakers, call a lentiscus.’

There is no reason to assume that the syntactic structure manifested overtly in (33) cannot also exist covertly; that is, in cases in which the relevant preverbal noun phrase has the same person and number features as the finite verb form and hence looks exactly like a genuine preverbal subject.

  Clitic Linearization: A Tale of Successful and Failed Changes 

89

Indeed, there is a school of thought which assumes that SVO order in prodrop languages generally involves left dislocation (see Alexiadou and Anagnostopoulou 1998: 501–511, together with references cited there). The latter position is probably too categorical. However, the assumption that some apparent subjects are in reality left-dislocated topics accounts neatly for the occurrence of enclisis in examples like (31), given that such an analysis implies that the ‘verb–clitic’ string is CP-initial. Proclitic examples like (32) should then be analysed as diagnosing the alternative structure, in which the preverbal noun phrase is the true subject and hence is internal to the CP unit.5 The proposed structural ambiguity might also help to explain an initially surprising statistical trend in the data, which is that enclisis occurs at a higher rate when the preverbal “subject” is a strong pronoun (yo, tu, ella, uos, ellos etc.) than when it is a full noun phrase, the corresponding enclitic rates in the 1250–1289 data being 75.2% (492/654) and 61.3% (438/714) respectively. The strong subject pronouns in Spanish always have definite reference, implying that as a class they are more suitable for being used as dislocated topics than are noun phrases, which may be indefinite in their reference. Under the proposed analysis, the higher rate of enclisis with strong pronouns is a by-product of the fact that enclisis requires the apparent subject to be a dislocated topic. Proclisis does of course grow in the post-subject context across the course of the Middle Ages (see Fig. 3.3 below), and it seems unlikely that this attests to a reduction in the rate of occurrence of the left-dislocated type of “subject”. It is more plausible to view this diachronic trend as further evidence of growth in the use of the competing, inherently proclitic weak pronoun alluded to in connection with examples (23) and (24). By the same token, it would be sensible to assume that not all of the variation in the post-subject context can be attributed to syntactic ambiguity, even in the earliest data.

3.2.4 U  nconstrained Variation in Post-prepositional Infinitival Clauses Turning now to post-prepositional infinitival clauses, variation in clitic placement in this context does not appear to follow any overarching structural principle in the early data (cf. Wanner 1991, 2007). In the

90 

I. E. Mackenzie

system overall, there were two main ways in which placement could be regulated, viz. either through obligatory proclisis, as in embedded finite clauses, or through the Tobler–Mussafia Law, reinterpreted here as generalization (12). The first principle obviously does not apply, given that both enclisis and proclisis are possible with post-prepositional infinitives, throughout the Middle Ages. As regards generalization (12), this too appears irrelevant for this context, given that the position of the CP boundary does not have any appreciable effect on the placement of the clitic. For example, a control clause governed by por/pora can be assumed to have the status of CP; hence enclisis is expected under (12). This is the case in the earlier example (4), but not in (34) below, where proclitic lo comes immediately after a CP boundary, in violation of (12).

(34) quanto tomo por [CP lo soltar] pechelo en siete doblos a aquel a quien fiziera el danno el preso. (Fuero Juzgo, fol. 66r) ‘Whatever he took for releasing him, he must pay it fourteen times over to whoever the prisoner harmed.’

The somewhat ‘unruly’ character (to use Wanner’s 2007 term) of the data in this context is further underlined by the fact that there is variation even with negation (which in finite contexts has a 100% correlation with proclisis), as is illustrated by (35) and (36) below.

(35) & es faz de fermosura. & de entendimiento. & de mansedumbre con assessegamiento. & de non se ensannar. (Judizios de las estrellas, fol. 5v) ‘And it is a face of beauty, of understanding, of gentleness and equanimity, of not becoming enraged.’



(36) yerra el omne contra dios en non conoscer le ni saber le guardar. (Libro de las leyes, fol. 2r) ‘Man sins against God by not knowing him or knowing how to respect him.’

Proclisis is actually strongly preferred with negation in the thirteenth century, the rate being 82.4% (42/51) according to the data in Mackenzie 2017 (p.  142). However, it never becomes categorical, the peak value being 90.5% (19/21) in the first three decades of the fifteenth century.

  Clitic Linearization: A Tale of Successful and Failed Changes 

91

3.3 Quantitative Evolution of the System Figure 3.1, providing a visualization of the numerical data in Table A.2 in Appendix 1, shows the advance of proclisis in those clause-types in which there was competition between enclisis and proclisis, viz. main clauses, post-prepositional infinitival clauses, imperative clauses and gerundial clauses.6 The imperative context, defined here narrowly as direct commands, is actually a sub-context within main clauses (cf. Bouzouita 2008a); however, it is also shown separately, given the manner in which proclisis is eventually lost in this environment.

3.3.1 T  he Common Trend in Main Clauses and Post-­ prepositional Infinitivals Although the endpoint of the systemic reorganization depicted in Fig. 3.1 falls slightly outside the time window shown, it is clear from the shape of the various curves, together with what is known about Spanish in the modern era, that the advance of proclisis is successful in one context but fails in the other three. These failures are examined in detail in Sect. 3.3.3 below. For the moment, it is important to observe that proclisis evolves in a very similar fashion in two of the contexts, viz. main clauses and post-prepositional infinitivals, up until the 1480 data point (which averages the data for the period 1470–1489). After this time, the corresponding curves part company definitively. The visual impression of a common evolution in these two contexts is confirmed by Table 3.1, which gives regression estimates of the growth rate of proclisis in each of these two environments (based on the numerical data in Table A.2 in Appendix 1). The growth rates are expressed as odds ratios (ORs), which refer to the factor by which the odds of the structure under examination increase per unit of time (see Sect. 1.2.2). As can be seen, proclisis had a very similar decadal growth rate in each of the two contexts. Given Kroch’s Constant Rate Effect (see Sect. 1.2.1), this finding implies that the late medieval upswing in proclisis in post-prepositional infinitival clauses and the advance of proclisis in main clauses were contextually specific manifestations of the same underlying

92 

100%

80%

I. E. Mackenzie

Main clauses Post-prepositional infinitivals Imperatives Gerundials

60%

40%

20%

0% 1260

1290

1320

1350

1380 1410

1440

1470

1500

1530

1560

1590

Fig. 3.1  Evolving rates of proclisis (1250–1609) (excluding negative contexts)

change. Indeed, the regression trendlines for the two processes are almost identical, as Fig. 3.2 shows. Thus although proclisis in the infinitival context undergoes a striking retreat at the end of the Middle Ages, its original advance in this context appears to be an essentially normal context-specific expression of a systemic trend. This calls into question the view, advanced by Wanner (2006, 2007, 2008), that the historically temporary shift towards proclisis in the infinitival context was a peripheral event, taking place among ‘configurations that as yet lacked linguistic organization in the pre-­ modern period’ (2007: 193). The latter perspective no doubt stems, in part at least, from an undue focus on the ultimate decline of proclisis in the infinitival context. However, to reiterate the point just made, prior to the time at which main clauses part company with infinitival ones, proclisis appears to exhibit a common advance in both contexts. Some scholars, particularly with respect to the history of French, have assumed a diachronically staggered collapse of the Tobler–Mussafia system, in which discrete periods manifest specific degrees of relaxation of the original constraints (see e.g. Hirschbühler and Labelle 2001). In contrast,

  Clitic Linearization: A Tale of Successful and Failed Changes 

93

the data reported here point to a historically seamless trend, with proclisis growing at a steady rate, analogously to when a new lexical item or a novel pronunciation undergoes sociolinguistic diffusion. The innovative item in the present case is plausibly identified as the inherently proclitic weak pronoun alluded to in Sect. 3.2.3.1. Gradual sociolinguistic diffusion of this variant of the weak pronoun, which competed with the conservative Tobler–Mussafia clitic from at least the thirteenth century, is the change model which best matches the pattern of steady quantitative growth evinced by the trends in Table 3.1 and Fig. 3.2. Table 3.1  Rate of advance of proclisis in main clauses and post-prepositional infinitival clauses Curve

Odds ratio (per decade)

Main clauses (1250–1609) Post-prepositional infinitivals (1250–1489)

1.131 1.145

100% Post-prepositional infinitivals 80%

Main clauses

60%

40%

20%

0% 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 3.2  Regression trendlines (parameter values: slope 0.1355, intercept −1.5416 (infinitivals); slope 0.1227, intercept −1.4589 (main clauses)) for the advance of proclisis in main clauses (1250–1609) and post-prepositional infinitival clauses (1250–1489)

94 

I. E. Mackenzie

3.3.2 D  econstructing the Main Clause Context: The Granularity Restriction The previous section alluded to Kroch’s (1989) hypothesis of a Constant Rate Effect (CRE), according to which a common rate of replacement is expected across all the contexts in which a syntactic change occurs. Fundamental to this concept is the notion of linguistic context; indeed, in their recent remodelling of the phenomenon, Kauhanen and Walkden (2017) contend that the ‘CRE occurs because of context-specific production biases which serve either to promote or to hinder an underlying change in progress’ (ibid., Sect. 1.2). A tacit assumption in both approaches is that linguistic contexts are in a sense independently determined, whereas in reality the analyst is free (within limits) to decide on how the totality of the tokens of a syntactic variable should be categorized. Here, for example, it has been taken for granted that main clauses constitute a context for the evolution of clitic linearization, and probably few people familiar with the history of Spanish would find this objectionable. On the other hand, given the discussion in Sect. 3.2.3, this macro-­ context could be further segmented into finer-grained micro-contexts, defined by what immediately precedes the finite verb and its clitic pronoun.7 Figure 3.3 below shows the quantitative evolution of proclisis in four of the most salient such micro-contexts (the corresponding numerical data can be found in Table A.3 in Appendix 1).8 As can be seen in the figure, although the lines converge as they approach the saturation level of 100%, they have quite widely dispersed values at the first data point, which references the period 1250–1269. The post-pause and post-coordinator curves occupy one extreme, with initial values of 1.5% and 1.9% respectively, while the post-adverbial context occupies the other, with an initial value of 50.3%, implying that the adoption of proclisis was already well advanced in this context. The post-subject curve has an initial value of 30.8%, which is a little above the midpoint between these extremes. Given that proclisis appears to reach the saturation level at roughly the same time in each context – i.e. the beginning of the seventeenth century or shortly thereafter – the dispersion in the ‘baseline’ values entails that the curves have differing degrees of steepness. This is reflected in the divergent growth rates shown

95

  Clitic Linearization: A Tale of Successful and Failed Changes 

100%

Post-adverbial Post-subject Post-pause

80%

Post-coordinator

60%

40%

20%

0% 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 3.3  Evolving rates of proclisis in main-clausal micro-contexts (1250–1609) Table 3.2  Rate of advance of proclisis in main-clausal micro-contexts (1250–1609) Micro-context

Odds ratio (per decade)

Post-adverbial Post-subject Post-pause Post-coordinator

1.094 1.184 1.233 1.242

in Table 3.2 (which have been estimated by regression analysis from the numerical data in Table A.3). As the table shows, proclisis grows in the post-subject context at almost twice the rate at which it grows in the post-adverbial context, while in the post-pause and post-coordinator contexts the growth is faster still. It thus turns out that the common growth rate that is observable at the level of the broadly defined macro-contexts disappears if the tokens are reapportioned to finer-grained micro-contexts. This suggests that there may be a granularity restriction on the CRE, in the sense that in order for the rate of change to be constant across a set of contexts, the latter cannot be too narrowly defined.

96 

I. E. Mackenzie

One possible reason for this is that narrowly defined micro-contexts may be susceptible to interference from the superordinate macro-context. In the classic CRE model, the change in each context corresponds to a distinct logistic curve (see Sect. 1.2.2), each one with its own intercept (reflecting the different baseline values) and each one keeping its geometric distance from all the others. In this way, although the change will be more advanced in some contexts than in others, the growth rate is able to be the same in each context.9 In contrast, the micro-contextual curves in Fig. 3.3 come together rather suddenly in the sixteenth century despite having widely dispersed values at the first data point. As was noted above, this inevitably means that the growth rates corresponding to the various micro-contexts cannot be identical. Thus it is the tendency towards convergence, particularly at the end of the change period, which prevents the set of micro-contexts from manifesting a CRE. This pattern of late convergence may indicate that as the macrocontext begins to approach the saturation level, micro-contexts which are lagging behind start to look anomalous, causing speakers to increase the rate of change in these contexts in order to maintain regularity in the macro-context overall. In the present case, for example, the rate of proclisis in the macro-context of main clauses stood at 71.7% by the beginning of the sixteenth century, while the rate in the post-coordinator micro-context was 17.2%. It does not seem implausible to imagine that speakers adjusted their usage in the post-coordinator context as a consequence of their sense that proclisis generally in main clauses was by far the dominant option. That some contexts lag behind others as a linguistic change takes place is a well-known fact and it is built into the idea underlying the CRE, both in its classic form and in the more recent operationalization proposed by Kauhanen and Walkden (2017). Thus although change may be significantly more advanced in some contexts than in others at any given time, synchronic disparities in frequency do not appear to matter at the macro-level, in the sense that the principle of a common rate of change seems to be unaffected. The finding here suggests that the same may not be true of micro-contexts. Obviously, this finding would need to be replicated in additional case studies in order to establish whether the granularity restriction was a general

  Clitic Linearization: A Tale of Successful and Failed Changes 

97

phenomenon as opposed to a chance occurrence in the specific case of clitic linearization in Spanish. However, given that the notion of ‘context’ is a fairly fluid one, with macro-contexts always being capable of subdivision into more narrowly defined micro-­ contexts, it seems unlikely that differences in baseline values should not over time result fairly commonly in analogous macro–micro mismatches to the one highlighted here.

3.3.3 The Failed Changes 3.3.3.1  Inherent or Accidental Failures? As can be seen in Fig. 3.1 given earlier, the successful shift towards proclisis in main clauses is flanked by three failed changes. That which takes place in the context of post-prepositional infinitivals is clearly the most spectacular, but the smaller failures in the imperative and gerundial contexts are not without interest. The fact that the failed changes occurred in conjunction with a successful change is significant, as it enables a direct test to be conducted of Postma’s (2010, 2017) central hypothesis concerning the relationship between failed and successful changes. ­ According to Postma, when a generally successful change S fails in a specific context, the bell curve F described by the failure will be roughly proportional to the (first) derivative of S.10 This follows from two considerations. Firstly, it is assumed that in the failing context the successful change S is inhibited by an equal but opposite counterforce, modelled as 1 − S, implying that F = S × (1 − S). Secondly, if S is a (parameterized) logistic curve of the type commonly used to model diachronic change (see Sect. 1.2.2), its derivative will, as a matter of mathematical fact, be inversely proportional to S × (1 − S). Therefore, if F = S × (1 − S), F will be directly proportional to the derivative of S. In the present case, the successful change S is the advance of proclisis in main clauses, and there are three Fs, namely the failures in the contexts of post-prepositional infinitivals, (positive) imperatives and gerundials. If the relationship between S and F is as described in the previous paragraph, we would expect F in each case to be proportional to the derivative

98 

I. E. Mackenzie

of S; that is to say, the values on the curve F should be proportional to the corresponding values on the curve described by S’s derivative. This idea is put to the test in Fig. 3.4, which compares the advance and retreat of proclisis in the three failing contexts (the unbroken lines) with a selection of bell curves (dotted lines) that are all proportional to the derivative of S as computed using the relevant parameter values given beneath Fig. 3.2. in Sect. 3.3.1 above. The second lowest of these dotted curves is the one that is specifically equal to S × (1 − S), the proportionality coefficient in this particular case being 8.15. As can be seen, the curve S × (1 − S) is not a good match for any of the failed changes, and neither are any of the other dotted curves shown. Furthermore, it is clear that the mismatch is not due to a failure to find an appropriate coefficient to apply to the underlying derivative. The dotted bell curves have a completely different geometry from the curves for the failed changes: they peak too soon and their flanks, particularly the left-hand ones, are noticeably less steep than the flanks of the unbroken curves, implying that the fluctuations in proclisis were in reality rather

100% Post-prepositional infinitivals 80%

60%

Imperatives

40% S (1 S) 20%

0% 1260

Gerundials

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 3.4  Observed failures (unbroken lines) compared to S  ×  (1  −  S) and other curves proportional to S’s derivative

  Clitic Linearization: A Tale of Successful and Failed Changes 

99

more abrupt than would have been the case had they evolved in accordance with the derivative-based model. Given this rather poor fit, it seems safe to conclude that the failed changes involving proclisis are not ­proportional to the derivative of the relevant successful change – proclisis in main clauses – even by approximation. The core idea behind the derivative-based model is that a failed change is doomed to fail from the start, the growth trend S being immediately undermined, in the relevant context, by its converse (1 − S). The inapplicability of this model to the reversals suffered by proclisis thus implies that, whatever factor or factors were responsible for the failures, they were not present from the outset. More plausibly, the rise and fall of proclisis in the failing contexts are instances of ‘accidental failures’ (Postma 2010: 287), the upward trend reflecting a different process from the downward trend. The next section examines the causes of the upward trends, while Sect. 3.3.3.3 investigates the reasons for the subsequent decline of proclisis in these failing contexts.

3.3.3.2  The Upward Curves In light of the discussion in Sect. 3.3.1, the cause of the initial increase in the post-prepositional infinitival context should already be fairly clear; that is to say, it is a context-specific manifestation of the system-wide diffusion of an inherently proclitic weak pronoun. The key evidence for this is the long period (1250–1489) during which proclisis in post-­ prepositional infinitivals evolves at almost exactly the same rate as it does in main clauses, the latter being the principal locus of the change. Superficially, the imperative context does not seem to belong to this change nexus, given that the growth of proclisis here (i.e. until the decline sets in after 1540) is comparatively slow, the odds ratio being 1.081 (the OR for main clauses overall is 1.131).11 Recall, however, that the imperative context is actually a micro-context within the macro-context of main clauses. Given the granularity restriction discussed in Sect. 3.3.2, there is no expectation that the rate of increase in proclisis in a micro-context should match the rate in a superordinate macro-context. Thus the different rates of growth between the imperative and the main clause environments should not be regarded as being a decisive diagnostic.

100 

I. E. Mackenzie

Indeed, there is a good reason why proclisis might be expected to increase more slowly in the specific context of imperatives than it did in main clauses generally; this is that imperative verb forms have a relatively high rate of clause-initial placement. It is true that late medieval Spanish and early modern Spanish offer examples in which a normally post-verbal element is fronted to the left of an imperative verb form. However, such structures are relatively infrequent and they also appear to be quite literary, being noticeably commoner in poetry than in prose. For example, the Sumario de la medicina, a late fifteenth century verse text (not included in the main corpus), is peppered with examples like the one below, in which the direct object el cibo (together with a dative clitic) precedes the imperative verb form dad:

(37) y con venda ligera se ligue y el cibo le dad a escaseza (Sumario de la medicina, fol. 5v) ‘and it should be covered with a light bandage, then give it food sparingly’

Notwithstanding this type of example, associated particularly with literary registers, pre-seventeenth century (positive) imperative verb forms are not radically different from their modern counterparts in terms of their tendency to surface in either clause-initial position or at the beginning of the intonation phrase. The advance of proclisis in the imperative context thus presupposes the emergence of structures such as that shown in the mid-sixteenth century example (38) below, where the placement of me would, in earlier times, have represented a strong violation of the Tobler– Mussafia principle.

(38) No tengo más que deciros; ahora, debaxo de quien sois y de lo que de vos confío, me responded vuestro parescer. (Cervantes de Salazar, Crónica de la Nueva España, book 4, chapter LXX) ‘I have nothing else to say to you. Now, setting aside who you are and what I expect from you, tell your view on the matter.’

It is therefore not surprising that, as can be seen in Fig. 3.5, the curve for proclisis in the imperative context tracks the curves for proclisis in the post-pause and post-coordinator main clausal micro-contexts quite

  Clitic Linearization: A Tale of Successful and Failed Changes 

101

100% Imperatives

80%

Post-pause 60%

Post-coordinator

40%

20%

0% 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 3.5  Imperative, post-pause and post-coordinator micro-contexts

closely, modulo the mini-peak in the latter at the 1440 data point (and modulo the fact that proclisis in the imperative context ultimately fails). The post-pause and post-coordinator micro-contexts are both environments in which the proclitic weak pronoun is in an initial position: initial within the intonation group or initial within the relevant coordinate structure, which is commonly (though not always) a clause. The factors militating against proclisis in these two environments are thus analogous to those which apply in the case of (positive) imperatives, making it p ­ lausible to suppose that the advance of proclisis across all three contexts reflects a single underlying sub-trend within the overall advance of proclisis in main clauses. Viewed in this light, the increase in proclisis in the imperative context – the first phase in this particular failed change – turns out to be one further context-specific manifestation of the system-wide diffusion of an inherently proclitic weak pronoun. However, this is obscured by the micro–macro relationship between the imperative environment and the broader main clause context, which enables the rate of advance in the former to be non-identical with the rate of advance in the latter.

102 

I. E. Mackenzie

It would be natural to assume that the emergence of proclisis in the (positive) gerundial context formed part of the same trend. However, the gerundial case is arguably somewhat distinct from what is observed with imperatives and post-prepositional infinitives. Unlike the other environments in which proclisis advanced, (positive) gerundials are not a variation context in the early part of the time window defined by the dataset, enclisis being categorical until the early decades of the 1400s. On the face of it, this is not incompatible with the Constant Rate Effect model, given that the CRE allows for a given change to have different actuation times in the different contexts which are affected. However, the assumption would then be that the rate of increase in the distinct contexts was approximately the same. That is clearly not the case here: in the principal change nexus involving main clauses and post-prepositional infinitivals, the odds of proclisis increase by about 13% or 14% per decade (the odds ratios being 1.131 and 1.145 according to Table 3.1), whereas in the gerundial context, in the brief period during which proclisis increased, the odds increased by almost 70% per decade, the OR being 1.692. Thus, despite its low peak frequency (9.5% in 1440), the growth rate for gerundial proclisis is something of an outlier, being approximately five times greater than the corresponding rate in the principal locus of the proclitic advance. In addition, rather than attesting to the spread of an inherently proclitic weak pronoun, capable of occurring in positions forbidden to the Tobler–Mussafia clitic, the emergence of proclisis in the gerundial context primarily reflects a trend towards the fronting of usually postverbal items to a position to the left of the gerund, triggering pronominal ­proclisis by analogy with the situation in finite clauses (see Sect. 2.3.1). Fifteenth century examples like (39) below are typical in this regard, the object poca onrra undergoing Ā-movement to the left periphery of the gerundial clause and the dative pronoun les being proclitic as a corollary, exactly as would happen in analogous finite structures.

(39) El que Ama El quatro mandamjento non guarda A su padre & madre por esta rrazon desonrrando mal trayendo & poca onrra les catando. (Corbacho, fol. 23r) ‘He who loves does not observe the fourth commandment, for this reason showing dishonour to his mother and father, bringing harm and having little respect for them.’

  Clitic Linearization: A Tale of Successful and Failed Changes 

103

The relevance of this type of example is underlined by the fact that gerundial proclisis peaks at the 1440 data point, which is an exact or close match for all but one of the peaks in the trend towards increased constituent fronting (in finite clauses) discussed in Sect. 2.5. According to the data adduced there, the peak for prepositional phrases, objects and nonfinite verb forms was 1440, that for predicative adjectives and nouns was 1460 and that for adverbs was 1500. It thus seems likely that the transient advance of proclisis in the gerundial context reflects analogical extension to this environment of the vogue for V2 structures which occurred towards the end of the Middle Ages, and hence is not a genuine symptom of the system-wide shift towards proclisis. This assumption immediately accounts for the phenomenon’s isolation, both in time and rate of advance, from the principal manifestations of this shift. We thus arrive at a scenario in which, of the three failed changes, two derive their initial impetus from a system-wide process – and hence could in theory have ended up being successful – while one appears to be the by-product of an unrelated stylistic trend occurring elsewhere in the language.

3.3.3.3  The Downward Curves Turning now to the reversals of proclisis in the relevant contexts, it is clear that post-prepositional infinitives constitute the most intriguing locus of failure, given the very high peak rate in this environment as well as the longevity of the upward trend. Wanner (2006: 149) attributes the loss of proclisis in this context to an abrupt ‘change of guard among who writes, edits and prints texts’, occurring at some unspecified time in the sixteenth century. However, that suggestion is predicated on the assumption that the loss of proclisis is sudden and dramatic, taking place within a timeframe of about fifty years (2006: 148). In fact, as can be seen in both Figs. 3.1 and 3.4 given earlier, the decline is rather more gradual than that; it begins in the first half of the fifteenth century and does not reach completion until the beginning of the seventeenth century. Moreover, the ten data points corresponding to the decline (i.e. 1420 through to 1600) produce good results under logistic regression, the chi square statistic for overall model fit being 409.39, which delivers a p-value of 0.0000 (0.05 being the custom-

104 

I. E. Mackenzie

ary upper limit for statistical significance). Both of these facts point to a normal pattern of syntactic evolution, which obviates the need to invoke a deus ex machina of the type proposed by Wanner. Given that proclisis came to be lost in all infinitival contexts, it can be surmised, grosso modo, that infinitival proclisis was at some point analysed as being ungrammatical. In diachronic syntax, changes of this kind are usually attributed to learners, who in turn are assumed to respond to determinate cues in the primary linguistic data (PLD) from which they infer the grammar of their language (see Roberts 2007: 122–140). The fundamental cue for infinitival proclisis is the string ‘clitic–VINF’ (where ‘VINF’ means ‘infinitival verb form’), given that if this string is not present in reasonable quantities in the PLD it seems impossible that any learner should posit that infinitival proclisis is a structure that can be generated by the grammar. Now the two principal contexts in which this string occurred in Old Spanish were (i) post-prepositional infinitival clauses and (ii) ‘VFIN–clitic–VINF’ structures, the latter instantiated most o­ bviously by cases of restructuring and clause union, as in the earlier example (5), repeated below, or causatives such as (40).

(5) pero sil uyan acorrer con aquellas cosas que son pora esto; pueden le estorcer de muerte. (Lapidario, fol. 20r) ‘But if they attend to him with those products that are suitable for this purpose, they can save him from death.’



(40) & manda lo apedrear ante que dañe muchos con sus costunbres (Mostrador y enseñador de los turbados, fol. 115v) ‘And it [the law] orders him to be stoned to death before he can do significant harm with his practices’

As was implicit in the discussion in Sect. 3.2.1, the string ‘clitic–VINF’ has a different structural analysis depending on whether it is in the post-­ prepositional frame or the ‘VFIN–clitic–VINF’ one. In the first context it is always proclitic on VINF, but in the second it is always enclitic on VFIN. In early acquisition, however, this fine structural distinction is likely to be irrelevant, the string ‘clitic–VINF’ being a potential cue for infinitival proclisis regardless of its locus. Among adult speakers, in contrast, the distinction clearly was relevant. One obvious reflection of this is the fact that

  Clitic Linearization: A Tale of Successful and Failed Changes 

105

the string ‘clitic–VINF’ evolved differently in each of the two contexts, as can be seen from Fig. 3.6 (the corresponding numerical data can be found in Table A.4 in Appendix 1). The reason for this divergence is fairly straightforward. In the post-­ prepositional case, the weak pronoun is proclitic and hence benefits from the general advance of proclisis. In the ‘VFIN–clitic–VINF’ case, the weak pronoun is enclitic and hence its frequency declines as proclisis advances. It is interesting to note that the two lines in Fig. 3.6 start from a similar place and, after a long period of being very wide apart, end up converging. This suggests that, despite the differing structural analyses attaching to the string ‘clitic–VINF’ in the two contexts, one context nevertheless had an effect on the other. As regards the directionality of this interference, it has to be assumed that the ‘VFIN–clitic–VINF’ context exercised an effect on the post-prepositional one, given that the string ‘clitic–VINF’ in the first context has a consistent (downward) trajectory while in the latter it has an initial (upward) trajectory from which it is subsequently diverted. At first glance, this seems paradoxical, given that the diversion of ‘clitic–VINF’ in the post-

100%

80%

Post-prepositional context Verb + clitic + infinitive

60%

40%

20%

0% 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

Fig. 3.6  Differential evolution of ‘clitic–VINF’ according to context

1560

1590

106 

I. E. Mackenzie

prepositional context away from its initial trend occurs at a time when this string was more frequent in that context than it was in the sequence ‘VFIN– clitic–VINF’, the latter, like all instances of pronominal enclisis, suffering a progressive decline across the Middle Ages. However, abstracting away from the specific position of the clitic within its context, it turns out that VFIN + VINF complexes are, overall, a more productive environment for clitics than are post-prepositional infinitival clauses. In the dataset collected here, for example, VFIN + VINF tokens involving a clitic outnumber equivalent post-prepositional infinitival tokens by 35.7% (33,408/24,615).12 Given the minority status of post-­prepositional infinitival clauses in this regard, the clitic linearization pattern associated with VFIN  +  VINF complexes can be assumed to have been relatively salient in the PLD from which learners inferred their knowledge of the grammar. Against that backdrop, the decline of the specific string ‘VFIN–clitic–VINF’ becomes potentially very significant. This process, by leaving the minority post-prepositional infinitival context as the only productive environment for the string ‘clitic– VINF’, must over time have created a situation in which the majority of tokens in the PLD in which a clitic construed with an infinitive were tokens in which the clitic did not occur to the latter’s immediate left. Probably this situation was already in place in the first half of the fifteenth century, given that by 1400 the rate of ‘VFIN–clitic–VINF’ was below 5%, meaning that clitics in VFIN  +  VINF complexes were almost always linearized as either ‘clitic–VFIN–VINF’ or as ‘VFIN–VINF–clitic’. Infant learners of Spanish in the first few decades of the fifteenth century would thus have been confronted with conflicting cues in the PLD, ‘clitic–VINF’ being solidly represented in the post-prepositional context but radically underrepresented among VFIN + VINF complexes. The latter, however, given their numerical superiority as an environment for clitics, must have represented the more salient of the two cues. It is thus plausible to suppose that learners analysed the near absence of the string ‘clitic– VINF’ among VFIN + VINF complexes as evidence that infinitival proclisis was ungrammatical, an inference which would, over time, result in the loss of such proclisis in the post-prepositional context. Intriguingly, then, the failure of proclisis in post-prepositional infinitival clauses appears to be indirectly the result of the advance of proclisis elsewhere in the system, specifically in finite clauses. For it is this latter process which results in the elimination of the ‘medial’ clitic position in ‘VFIN–clitic–VINF’

  Clitic Linearization: A Tale of Successful and Failed Changes 

107

structures, given that this medial occurrence was in fact an enclitic occurrence (see Sect. 3.2.1). From that perspective, the very trend of which the advance of proclisis in post-prepositional infinitivals was a manifestation ended up denuding the PLD of a cue that was essential to the survival of such proclisis. The key ingredient in this process is the clitic climbing structure, where increasing proclisis equates to a decline in the string ‘clitic–VINF’, the allimportant learner cue for infinitival proclisis. It is thus interesting to note that French, a language that now has categorical proclisis with infinitives, is not a clitic climbing language, whereas, conversely, Italian prohibits proclisis with infinitives but retains clitic climbing. The latter case is particularly suggestive, given that, according to Wanner (2006: 130–131), Italian underwent a transient swing towards proclisis in the post-prepositional infinitival context that is at least partially analogous to what happened in Spanish. If there is a pattern in this regard, with languages which retain clitic climbing but have lost enclisis in finite clauses tending to disallow proclisis in ­infinitival ones, it may well be due to the mechanism described above. This in turn would imply that, in the majority of such cases, one might expect to find in the historical record some evidence of a transient advance of proclisis in the specific context of post-prepositional infinitivals. Such an investigation lies, unfortunately, outside the scope of this book. The foregoing model for the decline of infinitival proclisis is not directly applicable to the other two contexts in which proclisis failed. However, as regards the imperative context, it is possible that the disappearance of infinitival proclisis paved the way for a secondary reanalysis whereby proclisis was interpreted as being associated exclusively with tensed verb forms. As is apparent in Fig. 3.4 (see also Table A.2 in Appendix 1), proclisis in (positive) imperative clauses declines after 1540, by which time proclisis with post-prepositional infinitives had fallen to approximately 25%, the string ‘clitic–VINF’ in VFIN  +  VINF complexes had all but disappeared (the rate being 1.1%) and gerundial proclisis was oscillating between 0% and just over 2%. In striking contrast, finite proclisis was close to being categorical at this time. The primary linguistic data of the early to mid- sixteenth century thus presented learners with a substantially degraded cue for nonfinite proclisis, allied to a very strong cue for finite proclisis. It would be unsurprising, in such a situation, if learners began to interpret the data in terms of a dichotomous governing principle, whereby clitic linearization was a function purely of verbal morphology. Within such a binary restructuring

108 

I. E. Mackenzie

of the system, it is easy to see how imperative verb forms could have ended up being subsumed with nonfinite forms, given (i) their ambiguous status as finite but not tensed and (ii) the fact that proclisis never enjoyed a majority frequency with such forms. As regards gerundial clauses, the decline of proclisis here correlates closely with the mid-fifteenth century downturn in finite constituent fronting. This is a direct consequence of the fact, noted at the end of Sect. 3.3.3.2, that the peak for gerundial proclisis is a near or exact match for all but one of the peaks for constituent fronting. Therefore, just as the rise of gerundial proclisis can be seen as an oblique reflex of the late medieval vogue for V2 structures, so its disappearance can be regarded as the indirect expression of a waning enthusiasm for that syntactic pattern.

3.4 Interpolation 3.4.1 Overview and Quantitative Presence/Evolution The clitic-related final topic that needs to be addressed is so-called interpolation, whereby a (preverbal) weak pronoun is not linearly adjacent to its host verb. This is usually discussed in connection with finite subordinate clauses, but it could in principle occur in all clause-types in which proclisis was possible, as (41) to (44) illustrate. Main clause:

(41) ca lo non deuen todos razonar de so uno de ambas las partes. (Fuero Juzgo, fol. 17r) ‘For not all of them should plead as one person on each side.’

Finite subordinate clause:

(42) Onde cuenta maestre pedro que fueron doze aquellos a quien le el dio [. . .] mas que non fueron mas de quatro los que regnaron. (General estoria V, fol. 96r) ‘For this reason, master Pedro recounts that there were twelve people among whom he divided it up . . . but only four actually ruled.’

  Clitic Linearization: A Tale of Successful and Failed Changes 

109

Infinitival clause:

(43) & touieron lo por mal de ombres agenos & aquien ellos cogieran en lo suyo de se assi apartar et se les alçar en su tierra (General estoria I, fol. 128v) ‘And they objected to outsiders, who they had welcomed into their society, separating themselves in this way and isolating themselves in their country.’

Gerundial clause:

(44) Eso mismo seria si el conprador consintiesse que fiziessen alguna cosa sagrada delo que conpro plaziendole o lo no contradiziendo. (Siete partidas, fol. 297r) ‘This would be the case if the buyer agreed to his purchase being put to a sacred use, by either expressing satisfaction or not objecting to it.’

The discontinuous linearization expressed by examples of this kind has given rise to a number of quite far-reaching theoretical claims about the syntax of the medieval weak pronoun. For example, Rivero (1986) and Fontana (1993) propose that it had phrasal rather than ‘head’ status while Rini (1990) contends that it was not fully grammaticalized until relatively late. From a somewhat different perspective, Poole (2013) argues that interpolation is the reflex of a Germanic-style V2 constraint. We should, however, be wary of assuming that the syntax manifested in ­interpolation structures has implications that are relevant to Old Spanish clitics generally. At the phenomenon’s height in the thirteenth century, interpolation accounts for just 4.4% of all tokens of weak pronouns in finite and infinitival clauses (2735/61,526) and 7.1% of preverbal tokens (2735/38,410), proportions which are indicative of quite a peripheral placement option. Moreover, while the principal diachronic process affecting weak pronouns in Old Spanish is the advance of proclisis, interpolation declines consistently, implying that the phenomenon was subject to a fundamentally separate change dynamic. This divergence is captured visually in Fig. 3.7, which compares the evolving rates of interpolation in finite and infinitival clauses with those of proclisis (the underlying numerical data are given in Table A.5 in Appendix 1).

110 

I. E. Mackenzie

100%

80%

60% Finite proclisis

40%

Infinitival proclisis Infinitival interpolation

20%

0% 1260

Finite interpolation

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 3.7  Evolving rates of finite and infinitival interpolation compared to proclisis

As can be seen, interpolation follows a quite different diachronic trajectory from proclisis, despite the fact that its contexts of occurrence are a subset of the contexts of occurrence of proclisis, both placement options being in opposition to enclisis. The fact that manifestly different change processes applied to the two forms of non-enclitic placement suggests that the weak pronoun surfacing in interpolation structures was not simply a variant of the proclitic weak pronoun but actually a separate item altogether. This would of course explain the otherwise somewhat mysterious co-existence of a general pattern in which the weak pronoun had to be linearly adjacent to its verb and a statistically infrequent pattern in which the weak pronoun could be separated from it. In the next section, I pursue this train of thought further, proposing that rather than being an output of the core grammar on a par with proclisis and enclisis, interpolation may well have been an instance of what certain authors refer to as ‘grammatical viruses’ (Sobin 1997; Lasnik and Sobin 2000; Sundquist 2011). This proposal is essentially in line with Wanner’s (1991: 352) suggestion that ‘[i]nterpolation is a superimposed phenomenon on top of the basic syntactic mechanisms responsible for clitic regulation’.

  Clitic Linearization: A Tale of Successful and Failed Changes 

111

3.4.2 Interpolation as a Grammatical Virus The grammar of a standardized language, as Spanish was rapidly on the way to becoming in the thirteenth century, is typically heterogeneous. In addition to the maximally general structural rules which are internalized through normal, unconscious acquisition, standard grammars almost invariably contain highly specific conventions which are linked to particular types of discourse and which are acquired outside of the basic acquisition process, for example in pedagogical contexts or through cultural exposure. Virus Theory captures this duality by positing the existence of peripheral rules or ‘viruses’ that augment the output of the basic or core grammar. In their discussion of English whom, which they analyse as a viral output, Lasnik and Sobin (2000: 366) state that ‘the characteristics which set viruses apart include lexical specificity, directionality, underextension, overextension, and delayed acquisition’. To this list of signature properties can presumably be added optionality, which has long been assumed to be incompatible with minimalist assumptions about core syntax. Indeed one of the aspects of English whom highlighted by Lasnik and Sobin (ibid. p. 346) is the fact that augmenting who with the accusative suffix -m is entirely optional, something which sets whom apart from other accusative pronouns such as him (which cannot be replaced by he). Looking at interpolation in Old Spanish, it is quite striking how many of the viral signatures it displays. Most obviously, it is a syntactic operation which is entirely optional, in the sense that whenever interpolation occurs strict proclisis would also be possible. Thus for every example of interpolation such as (41) to (44) given above, it is always possible to find a structurally analogous example in which the pronoun is adjacent to the verb rather than separated from it, as (45) to (48) illustrate:

(45) Ca non le puede njngun omne uer. (General estoria I, fol. 137v) ‘For no man can see him.’



(46) Y enuio much enporidat un so sieruo que dizien carthon que era omne bueno e sesudo. y en qui ella se fiaua mucho (Estoria de España I, fol. 25r) ‘And she secretly dispatched one of her servants, named Carthon, who was a decent and able man, and in whom she placed great trust.’

112 

I. E. Mackenzie



(47) touo lo por grand marauilla el. & quantos auie y. de assi se yr toda una yente tan grand como aquella. (General estoria IV, fol. 7r) ‘He and all those who were present marvelled that the entirety of such a large population should go away in this manner.’



(48) Si algun omne recibe sieruo agenno fuydo. non lo sabiendo. [. . .] (Fuero Juzgo, fol. 77r) ‘If a man takes in an escaped serf without realizing it . . .’

Several authors have argued that interpolation in Old Spanish has a pragmatic correlate, in the sense that the item placed between the pronoun and the verb is either a focus (Battlori et al. 1995) or a topic (Poole 2013). However, any such correlate could not be viewed as being the trigger for the interpolation, given that interpolation is by no means obligatory when the verb is preceded by a focal or topical constituent; indeed, proclisis represents the normal linearization pattern in such circumstances, as is discussed in detail in Chap. 2. In any case, as is pointed out by Poole (2013: 87–88), when the interpolated element is the negator no/non, which represents by far the commonest type of example, it is difficult to analyse interpolation as anything other than ‘formal movement’; that is, as movement ‘which has no pragmatic or semantic consequences’ (see also Castillo Lluch 1997: 420). This is strikingly demonstrated in cases of expletive or pleonastic negation, as in (49) below (cited by Poole 2013: 88), where the form non is an empty marker, devoid of semantic content (see Sect. 6.3.3).

(49) [. . .] que podrian quebrantar las arcas & los çilleros; & tomar lo que quisiessen; & despues negar que lo non tomaron. (Ordenamiento de Alcalá, fol. 26v) ‘. . . they could break into the storehouses and cellars and take what they wanted and then later deny that they took it’

For Poole (ibid. p. 88–89), semantically and pragmatically vacuous interpolation is the reflex of a Germanic-style V2 (‘verb-second’) constraint, the interpolated item serving as the ‘first-position’ element. However, an account along those lines is problematic given the finding in Chap. 2 that Old Spanish was not a V2 language and that, to the extent that V2 struc-

  Clitic Linearization: A Tale of Successful and Failed Changes 

113

tures occurred, they actually correlated with proclisis when weak pronouns were involved. We are therefore left with the conclusion that interpolation in Old Spanish was simply optional. A second virus-like property exhibited by interpolation is lexical specificity. Based on her detailed corpus study of eight high and late medieval texts, Castillo Lluch (1996: 307) finds that interpolation was only really productive with the negator non and subject pronouns: ‘En el resto de casos, la interpolación es tan limitada estadísticamente que difícilmente podemos considerarla como una práctica corriente’.13 She also finds that many of the examples not involving negation or subject pronouns are actually fixed legal or notarial formulas, such as delo asi fazer ‘in doing so’, moneda que se agora usa ‘valid currency’ and que vos esta mj carta mostrare ‘who shows you this charter of mine’. This state of affairs contrasts with the situation concerning proclisis and enclisis, which are not lexically restricted in any way at all. An additional viral signature alluded to by Lasnik and Sobin is ‘directionality’, whereby the output of the virus is subject to arbitrary restrictions based on linear order. For example, the normative/viral rule that licences nominative I in English compound subjects such as Bill and I, disallows I-first formulations such as I and Bill, despite the general commutativity of coordination. In light of the existence of this type of restriction, it is significant that interpolation is only possible when the weak pronoun occurs to the left of its anchoring verb and never when it occurs to the latter’s right. This gap in the data cannot be due to any limitation of interpolation to embedded contexts, given that, contrary to what is often stated, the phenomenon did occur in main clauses. In other words, given examples like (50) and (51) below, the absence of formulations of the type shown in (52) is, from the structural point of view, arbitrary.

(50) ca los el auje conbidado que fuesen sus huespedes ese dia (Libro del caballero Zifar, fol. 94r) ‘For he had invited them to be his guests that day.’



(51) ca sabedes lo uos & fuestes encrubencia conuinient a muchos & a muchas que fizieron so uos lo que quisieron. (General estoria II, fol. 125v) ‘For you know this and you provided cover to many men and women who realized their desires beneath you.’

114 



I. E. Mackenzie

(52) *ca auje el los conbidado / *ca sabedes uos lo

The foregoing directional limitation implies, in addition, that interpolation exhibits underextension, which is Lasnik and Sobin’s third viral signature. The idea in this regard is that, unlike normal syntactic operations, the viral rule does not generate all of the structures that one would expect. In this case, the ungenerable structure is the one shown in (52). Lasnik and Sobin also allude to overextension, ‘the unexpected application of a rule to a structure that it “ideally” should not affect’ (ibid. p. 367). For example, the English ‘…and I’ rule, whose intended effect is to prevent pronouns in compound subjects from having accusative case, is often extended to compound objects, resulting in hypercorrections like They’ve approached Lesley and I. Identifying hypercorrections in the Spanish language of centuries ago is obviously a rather speculative enterprise and any remarks made here about possible instances of overextended interpolation should be interpreted in that light. With that caveat in place, example (53) below (cited by Bouzouita (2008a) and Fontana (1993) among others) is conceivably an instance of overextension, in the sense that the interpolation rule is arguably being applied to a structure that should have been immune to the rule.

(53) dixo le yo dare A esta vjllana los tornos (Corbacho, fol. 100r) ‘She said, “I will give this pauper the run around.”’

Note that le construes here with dare and not with dixo. The interpolation of the subject pronoun yo between the clitic and its verb thus causes the clitic to be discourse-initial, in the sense that it is the first element in the quoted utterance. This placement option may have been normatively unacceptable in the language of the time (the manuscript is from 1466), given that some form of embargo on discourse-initial weak pronouns appears to have existed until at least the Siglo de Oro (Lapesa 1981: 407). Turning to the final viral signature, late acquisition, at this historical distance it is impossible to know with any certainty at what phase in the learning cycle interpolation was internalized. However, according to one traditional view (Chenery 1905: 50; Lapesa 1981: 241),

  Clitic Linearization: A Tale of Successful and Failed Changes 

115

interpolation may have been borrowed into Spanish court usage from Galician, suggesting that the structure was characteristic of literary registers, which in turn implies that it was acquired late, as part of literacy rather than core language acquisition. The notion that interpolation was a courtly usage modelled on Galician literature has been called into question by Castillo Lluch (1997: 418) although, more recently, Matute Martínez (2013: 169) has argued that the phenomenon was at the very least subject to stylistic and/or dialectal constraints. This would of course be consistent with Castillo Lluch’s own finding (1996: 307) that interpolation was relatively common in fixed legal formulas. The evidence is clearly well short of being conclusive, but there are at least some grounds for thinking that interpolation was not acquired in childhood as part of core syntax. Overall, then, interpolation has or potentially has all of the signature properties of grammatical viruses. This state of affairs, coupled with the fact that its historical evolution appears to be unrelated to the main diachronic trend observable in the clitic system, viz. the advance of proclisis, suggests that we should be cautious about using interpolation as a basis for major generalizations about clitics or their placement in Old Spanish.

3.5 Conclusion The present chapter has examined the evolution of clitic linearization across a time window which opens at the end of the High Middle Ages and closes at the outset of the modern period. In the early part of this timeframe, the principal constraint on clitic placement was a rule, applying in finite contexts only, which forbad the occurrence of a clitic in a position immediately after a CP boundary. At least some of the apparent variation in clitic placement can be regarded as being in reality a reflex of this rule, given that the distribution of CP boundaries is not necessarily expressed overtly in surface strings of words. On the other hand, even in the thirteenth century one finds examples of proclisis which appear to be

116 

I. E. Mackenzie

completely unconstrained in terms of what, if anything, precedes the weak pronoun. The quantitative diachronic data point to a gradual diffusion of this ‘inherently proclitic’ weak pronoun, finite main clauses, together with post-prepositional infinitival clauses, constituting the principal locus in which this evolutionary process is manifested. However, by indirectly denuding the primary linguistic data of an important cue for infinitival proclisis, the same process ultimately triggers a reanalysis whereby infinitives cease to be able to host proclitic weak pronouns. This process in turn results in a dearth of tokens of proclisis with nonfinite verb forms, which, allied to a near absence of enclisis with finite verb forms, triggers a further reanalysis, whereby the choice between proclisis and enclisis becomes a function essentially of verbal morphology. In the dichotomous system which is instituted by this change, the tenseless but finite (positive) imperative forms fall together with the nonfinite ones, becoming incompatible with proclisis despite having initially participated in the general advance of that placement option. A third ‘failed change’ can be identified in gerundial contexts, although this appears to be peripheral to the events just described, as it plausibly owes its occurrence to the transient vogue for constituent fronting which was a feature of late medieval Spanish. The near identical rates of change in the principal nexus comprised by main clauses and post-prepositional infinitivals (up until 1480) can be regarded as further empirical confirmation of Kroch’s Constant Rate Effect. At the same time, it has been shown that a rate of change that is common across broadly defined contexts may not carry over to contexts which are more narrowly defined, a phenomenon labelled here as the ‘granularity restriction’. Whether this is a general property of quantitative diachronic change or a chance feature of the particular case examined here is a matter for further research. The finding does, however, point to the need to interrogate more systematically the notion of ‘context’ insofar as it is used as a prism through which to analyse language change. As regards the failed shifts towards proclisis with post-prepositional infinitives, positive imperatives and gerunds, their existence in close proximity to the successful shift towards proclisis in tensed finite clauses can be regarded as supporting Postma’s (2010, 2017) general proposition that suc-

  Clitic Linearization: A Tale of Successful and Failed Changes 

117

cessful changes will normally be flanked by related unsuccessful changes. However, the geometry of the curves described by the three failed changes considered here cannot be modelled in terms of the derivative of a logistic function, even by approximation, contrary to what is predicted by Postma’s model of inherent failures. The failures should thus be regarded as accidental failures and their existence should be treated as evidence that failures of the accidental type may be the commoner variety. From an even more general perspective, the landscape defined by the evolution of clitic placement in Old Spanish, with its interplay between successful and failed changes, is a striking reminder that language change is a more complex and less unidirectional affair than is commonly supposed.

Notes 1. At this point I abstract away from so-called interpolation, the phenomenon whereby the weak pronoun could be separated from its anchoring verb. The topic is discussed in Sect. 3.4 below. 2. I assume here the model of ‘structured coordination’ proposed in Chomsky (2013: 46). 3. The items in question are as follows (shown using their modern spelling): ahí, ahora, allí, antes, aquí, así, aun, con eso, con esto, después, entonces, luego, por ende, por eso, por esto, siempre, sobre eso, sobre esto. 4. For detailed descriptive coverage of variation in this context, see Granberg (1988) and Castillo Lluch (1996). 5. Given that left-dislocated topics by definition are not focal, an analysis along the lines suggested here is potentially compatible with Granberg’s (1999) proposal that proclisis is found after emphatic subjects and enclisis is found in the absence of such emphasis (assuming the notion of emphasis can be reconstructed in terms of focus). 6. Negated instances are excluded, given the absence of variation under negation in all but one of the contexts shown. 7. Imperatives could also be treated as a micro-context within the superordinate main clause context. However, matters are complicated in this regard by the fact that the shift towards proclisis with (positive) impera-

118 

I. E. Mackenzie

tives was ultimately reversed. The imperative curve is discussed in more detail in Sect. 3.3.3 below. 8. The post-pause micro-context refers to contexts in which the verb and its clitic come immediately after an orthographic stop such as a semi-colon, a colon or a full stop (the latter item typically delimiting breath groups rather than marking sentence divisions). 9. As part of this, the change is expected to approach completion (100%) at different times in different contexts. 10. In Postma (2017), this relationship is said to hold ‘by approximation only’. 11. Bouzouita (2008a) finds rates of proclisis among imperatives that are more in line with finite main clauses generally. This can be explained, in part at least, by the fact that Bouzouita employs a wider definition of imperatives than that used here. For example, she includes optatives, whereas the data here refer only to commands. She also includes negative instances, which are excluded here on the grounds that they were not a variation context in Old Spanish. 12. These data include negative tokens. 13. ‘In the remaining cases the statistical frequency is so low that it is difficult to argue that we are dealing with a productive process.’

References Alexiadou, Artemis, and Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory 16: 491–539. Battlori, Montserrat, Carlos Sánchez, and Avel·lina Suñer. 1995. The incidence of interpolation on the word order of Romance languages. Catalan Working Papers in Linguistics 4 (2): 185–209. Bouzouita, Miriam. 2008a. At the syntax-pragmatics interface: Clitics in the history of Spanish. In Language in flux: Dialogue coordination, language variation, change and evolution, ed. Robin Cooper and Ruth Kempson, 221–263. London: College Publications. Castillo Lluch, Mónica. 1996. La posición del pronombre átono en la prosa hispánica medieval. Doctoral dissertation, Universidad Autónoma de Madrid.

  Clitic Linearization: A Tale of Successful and Failed Changes 

119

———. 1997. La interpolación en español antiguo. In Actas del IV Congreso Internacional de Historia de la Lengua Española, ed. Claudio García Turza et al., 409–422. La Rioja: Universidad de La Rioja/Servicio de Publicaciones. Chenery, W.H. 1905. Object pronouns in dependent clauses: A study in Old Spanish word-order. Publications of the Modern Language Association of America 20: 1–151. Chomsky, Noam. 2013. Problems of projection. Lingua 130: 33–49. Fontana, Josep. 1993. Phrase structure and the syntax of clitics in the history of Spanish. Doctoral dissertation, University of Pennsylvania. Granberg, Robert. 1988. Object pronoun position in medieval and early modern Spanish. Doctoral dissertation, University of California, Los Angeles. ———. 1999. Clitic position in thirteenth-century Spanish: Sentences with preverbal subject. La Corónica 27 (2): 89–113. Hirschbühler, Paul, and Marie Labelle. 2001. Evolving Tobler–Mussafia effects in the placements of French clitics. In New approaches to old problems: Issues in Romance historical linguistics, ed. Steven Dworkin and Dieter Wanner, 165–182. Amsterdam: John Benjamins. Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Kuchenbrandt, Imme. 2016. Prosody and object clitic placement: A comparison of Old and Modern Spanish. Lingua 181: 81–98. Lapesa, Rafael. 1981. Historia de la lengua española, 9th ed. Madrid: Gredos. Lasnik, Howard, and Nicholas Sobin. 2000. The who/whom puzzle: On the preservation of an archaic feature. Natural Language and Linguistic Theory 18: 343–371. Mackenzie, David. 1997. A manual of manuscript transcription for the dictionary of the Old Spanish Language. (5th edition, revised and expanded by Ray Harris-Northall). Madison: Hispanic Seminary of Medieval Studies. Mackenzie, Ian. 2017. The rise and fall of proclisis in Old Spanish postprepositional infinitival clauses: A quantitative approach. Bulletin of Hispanic Studies 94 (2): 127–146. Matute Martínez, Cristina. 2013. Hacia una caracterización dialectal de la interpolación en el castellano de la Edad Media. In Actas del XXVI congreso inter-

120 

I. E. Mackenzie

nacional de lingüística y filología romances, 2010. Universitat de València, ed. Casanova Herrero and Césareo Calvo Rigual, 155–164. Berlin: Walter de Gruyter. Mussafia, Adolfo. 1886. Una particolarità sintattica della lingua italiana dei primi secoli. In Miscellanea di filologia e linguistica, dedicata alla memoria di Napoleone Caix e Ugo Angelo Canello, ed. G.I. Ascoli et al., 255–261. Florence: Le Monnier. Poole, Geoffrey. 2013. Interpolation, verb-second and the low left periphery in Old Spanish. Iberia: International Journal of Theoretical Linguistics 5 (1): 69–98. Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. ———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Rini, Joel. 1990. Dating the grammaticalization of the Spanish clitic pronoun. Zeitschrift für romanische Philologie 106: 354–370. Rivero, María Luisa. 1986. Parameters in the typology of clitics in Romance and Old Spanish. Language 64: 774–807. Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Sobin, Nicholas. 1997. Agreement, default rules, and grammatical viruses. Linguistic Inquiry 28: 318–343. Sundquist, John D. 2011. Negative movement in the history of Norwegian: The evolution of a grammatical virus. In Grammatical change: Origins, nature, outcomes, ed. Diane Jonas et al., 293–312. Oxford: Oxford University Press. Tobler, Adolf. 1875. Review of Jules Le Coultre, De l’ordre des mots dans Crestien de Troyes. Göttingische Gelehrte Anzeigen 34: 1057–1082. Wanner, Dieter. 1991. The Tobler–Mussafia law in Old Spanish. In Current studies in Spanish linguistics, ed. Héctor Campos and Fernando Martínez Gil, 313–378. Washington, DC: Georgetown University Press. ———. 2006. The power of analogy: An essay on historical linguistics. The Hague: Mouton de Gruyter.

  Clitic Linearization: A Tale of Successful and Failed Changes 

121

———. 2007. Wild variation, random patterns and uncertain data. In Historical linguistics 2007. Selected papers from the 18th international conference on historical linguistics, ed. Monique Dufresne et al., 185–194. Amsterdam: John Benjamin. ———. 2008. Pronombres átonos de objeto con infinitivo o la pertinencia de la periferia. In Actas del VII congreso internacional de historia de la lengua española, ed. Concepción Company and José G.  Moreno de Alba, 197–224. Madrid: Arco Libros.

4 DP Structure: From Multiple Determiners to Just One

4.1 Introduction Over the last two decades or so, syntacticians have converged on the notion that the fundamental element inside nominal projections such as subjects and objects is not the noun itself but the determiner (D). From that perspective, a subject or an object belongs to the syntactic category of DP (‘determiner phrase’) rather than NP, as was previously assumed. In the newer model, the NP is analysed as being the complement of D, from which it inherits its referential capability. Given this framework, determiner-related syntax has acquired a special importance, with historical data from Romance providing important clues as to the pathways involved in the evolution of DP structure. One such pathway relates to the emergence of constraints on the occurrence of subjects and objects that lack any overt D-element. As is well known, these occurred freely in Latin but have acquired a rather more constrained distribution in the modern Romance languages. As part of this change, what in Latin would have been determinerless DPs have come to be headed by definite or indefinite articles. The latter items did not exist in Latin but evolved in Romance through the © The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_4

123

124 

I. E. Mackenzie

r­ eanalysis of erstwhile demonstratives such as ille or the numeral ūnus. A number of researchers have focused on the issue of whether there is any residue of this process in Old Spanish. For example, Batllori and Roca (2000) propose that the forms el, la, los and las are lexically ambiguous in early Spanish literature, as between a definite article and a demonstrative.1 Somewhat analogously, Company (1991: 411, n. 13) has argued that un, una, unos and unas as used in the medieval language are forms of a numeral rather than an article. In contrast, the situation reported in the present chapter is consistent with the assumption that the modern patterns of usage of both the definite and indefinite articles, together with the modern constraints on the distribution of D-less subjects and objects, were largely in place by at least the second half of the thirteenth century and had probably been in existence for quite some time. According to the analysis developed here, the principal axis of differentiation between the medieval and modern systems relates to the manner in which the DP is built up. Whereas modern Spanish allows for the insertion or merge of only one D-element, forcing any competing determiner-­like constituent to occupy a post-nominal adjectival position, as in un amigo tuyo ‘a friend of yours’ or el chico aquel ‘that boy’, the medieval language allowed multiple D-elements to be stacked in prenominal position, as in un su hermano ‘a brother of his’ or la otra mi carta ‘my other charter’. The present chapter treats such structures as evidence of the availability in Old Spanish of DP recursion, a determiner-related analogue of the clause-level process known as CP recursion (de Haan and Weerman 1986). In the diachronic dimension, DP recursion was stable for much of the late Middle Ages, but it declined from the early 1400s. This decline is characterized by a Constant Rate Effect (Kroch 1989), confirming that the individual structures involved were manifestations of a single underlying pattern. Notwithstanding the common rate of change, disparities in synchronic use provide a nice illustration of the phenomenon of ‘time separation between contexts’ (Kauhanen and Walkden 2017: 1.3), whereby accumulated change is unequal across the various contexts.

  DP Structure: From Multiple Determiners to Just One 

125

4.2 Overview of the Modern D-System The DP or determiner phrase is the canonical syntactic category of subjects and objects (including the object of a preposition), which together are known as arguments. The assumption underlying this approach is captured by Longobardi’s (1994: 620) dictum that ‘[a] “nominal expression” is an argument only if it is introduced by a category D’. For example, the singular count noun médico ‘doctor’ is banned from occurring without a determiner (D) if it is a subject or an object, but not if it is the predicative complement of a copula, as (1) and (2) illustrate. (1) El comité a nombrado a *médico/un médico como gerente general. ‘The committee has a appointed a doctor as chief executive.’ (2) Pedro es médico. ‘Pedro is a doctor.’

The possibility for bare plural count nouns and mass nouns to appear as arguments, as in (3), is customarily assumed to reflect the presence of a phonologically null D, which satisfies the same syntactic condition as an overt determiner such as el or un. (3) Comimos melocotones/queso. ‘We ate peaches/cheese.’

As noted by Longobardi (ibid. p. 618), the null determiner (in languages like Italian and, by extension, Spanish) has a default existential interpretation, making it roughly equivalent to ‘some’ or ‘an unspecified quantity of ’. This is reflected in the well-known generalization that bare nouns in Spanish – and fairly generally (but not universally) in Romance – never have a generic or kind-denoting interpretation, unlike their counterparts in English. An additional aspect of null D is that it is subject to what Longobardi (1994: 617) originally called a ‘lexical government requirement’. This latter concept does not translate easily into the terminology of more recent syntactic theory, but the import is essentially that the DP headed by null D must be the complement of a lexical head, such as a verb or a

126 

I. E. Mackenzie

preposition. Example (3) above illustrates the case in which where the lexical government requirement on null D is satisfied by a verb, while (4) below illustrates the case in which a preposition performs that role (‘e’ stands for ‘(phonologically) empty’): (4) Fregué el suelo con [DP [D e] lejía]. ‘I mopped the floor with bleach.’

The most striking manifestation of the lexical government requirement is the so-called Naked Noun Constraint (NNC), formulated by Suñer as in (5) below: (5) An unmodified common noun in preverbal position cannot be the surface subject of a sentence under conditions of normal stress and intonation. (Suñer 1982: 209)

The NNC captures the fact that formulations such as (6) and (7) below are ungrammatical: (6) *Alumnos se quejaron. ‘Students complained.’ (7) *Humo salía de la chimenea. ‘Smoke was coming out of the chimney.’

The subject DPs in (6) and (7) do not occur as the complements of any lexical item. Therefore, if they are headed by null D, the latter will not be “lexically governed” in Longobardi’s sense, and the ungrammaticality of the containing sentences follows accordingly. Moreover, as is shown by the ungrammaticality of examples like (8), the lexical government requirement applies more generally than is suggested by Suñer’s NNC, the scope of which is limited by the reference to the preverbal position: (8) *Me gustan [DP [D e] manzanas]. ‘I like apples.’

Although manzanas comes after the verb, it is not the latter’s object but its subject. Hence, from a structural point of view, manzanas is not the

  DP Structure: From Multiple Determiners to Just One 

127

complement of a lexical head and the containing sentence (8) is ungrammatical for the same reason that (6) and (7) are; viz., a null D which is “lexically ungoverned” (see also Contreras 1986: 26).2 The foregoing cases should be distinguished from the case in which the DP complement of a verb is fronted to a preverbal position, as in (9), where the copy of manzanas in strikethrough font shows the initial merge site or ‘base position’ of this item.

(9) Manzanas hay manzanas. ‘There are apples/Apples we have.’

Although manzanas in its preverbal surface position is not lexically governed in Longobardi’s sense, in its base position it is, given that this item is initially merged as the complement of hay. According to Longobardi (ibid. p. 616, note 10), in this type of case the lexical government requirement is satisfied by ‘reconstructing’ the raised DP in its base position. More generally, Ā-movement (see Sect. 2.3.1) can be assumed not to interfere with the licensing of null D, as (10) and (11) confirm:

(10) Pan tenemos pan. ‘Bread we have.’



(11) Excusas no quiero escuchar excusas. ‘I don’t want to hear any excuses.’

Both the lexical government requirement and the existential interpretation attaching to null D are commonly nullified in Spanish (and other Romance languages) through the insertion of an expletive or pleonastic definite article into what would otherwise by an empty D position, as in (12):

(12) [DP [D Los] lobos] cazan en manadas. ‘Wolves hunt in packs.’

The presence of phonological material in the D position obviates the violation of the lexical government requirement that would arise if the preverbal subject were the bare noun lobos. Moreover, the comparison of (12) with its English translation shows that the generic reading in this

128 

I. E. Mackenzie

type of case arises independently of the presence of the definite article, implying that the latter item, in the Spanish example, is semantically vacuous (cf. Longobardi 1994: 650; also, Vergnaud and Zubizarreta 1992). Turning now to the kinds of overt item that can occupy the D position in Spanish, these include the definite and indefinite articles, demonstratives, possessives and indefinite quantifiers such as algún ‘some’, mucho ‘much/many’ and poco ‘little/few’.3 Notwithstanding this multiplicity, a characteristic feature of the modern DP is that an array of stacked D-elements is not permitted. This constraint is reflected most obviously in the placement of demonstratives and possessives, which varies as a function of whether the D position is occupied by some other item. A selection of allowable permutations is given in (13) to (16):

(13) [DP [D la] casa aquella] ‘that house’



(14) [DP [D el] proyecto suyo] ‘his/her project’



(15) [DP [D esa] idea tuya] ‘that idea of yours’



(16) [DP [D tu] teoría aquella] ‘that theory of yours’

Superficially, the kind of formulation shown in (17) below appears to involve stacked determiners.

(17) los muchos/pocos libros que ha escrito ‘the many/few books he has written’

However, in this type of example, the indefinite quantifiers muchos and pocos do not have the meaning which they exhibit when used as determiners. One of the characteristics of indefinite determiners is that the DPs in which they occur may be inserted as the complement of existential hay, as is shown in (18):

  DP Structure: From Multiple Determiners to Just One 



129

(18) Hay muchos/pocos/varios/otros libros (en esta casa). ‘There are many/few/several/other books (in this house).’

Insertion after hay is unavailable, however, for the type of DP shown in (17):

(19) *Hay los muchos/pocos libros (que ha escrito). ‘There are the many/few books (which he has written).’

The ungrammaticality in (19) stems from the so-called definiteness effect, whereby definite DPs are excluded from existential constructions in a wide variety of languages. Given their sensitivity to the definiteness effect, DPs such as los muchos/pocos libros – more generally, those of the form ‘definite article + quantifier + NP’ – should be analysed as being definite. Accordingly, the quantifier in such DPs must be analysed as not being in the D position, assuming that DPs inherit their [+/−] value for definiteness from their D head. It follows that linear sequences such as los muchos + NP or los pocos + NP do not involve stacked determiners.4 Summing up the points made in this section, the principle features of the modern Spanish DP are as set out in (20):

(20) (i) D can be phonologically null but, in that case, the containing DP is interpreted existentially and must occur as the complement of a lexical head; (ii) the definite article in some instances is expletive, in the sense that it does not in itself carry any meaning and instead its purpose is to provide phonological material to obviate the lexical government requirement attaching to null D; (iii) various categories of item can occupy D, but stacked or iterated Ds are disallowed.

4.3 C  omparison Between the D-Systems of Old Spanish and Modern Spanish As with the grammar generally, the syntax of the DP evolves over time, meaning that single generalizations made in respect of the ‘Old Spanish DP’, with no temporal qualification, are somewhat misleading. The dis-

130 

I. E. Mackenzie

cussion below takes the first two data points in the corpus (covering the period 1250–1289) as a benchmark, with which developments in the later Middle Ages can be compared.

4.3.1 Null D and the Expletive Article In the Spanish of the latter half of the thirteenth century, both the interpretation of null D and the constraints on its distribution are already largely the same as in modern Spanish. Null D therefore carries a default existential interpretation and is restricted (in principle) to DPs which are complements of a lexical head such as a verb or a preposition. Illustrations are given in (21) to (23) below, the relevant DP and its “governing” lexical head being shown in bold.

(21) por que es costumbre de todos los que fazen libros de render primera mient grado e gracias a dios. (Libro de las animalias, fol. 2v) ‘because it is the practice of all those who write books to offer first of all gratitude and thanks to God.’



(22) otorgo dios alos omnes. que despues del diluuio que comiessen carne. (General estoria I, fol. 14v) ‘God stipulated to men, that after the flood they should eat meat.’



(23) & mando a todos los buenos omnes a los caualleros sennalada mientre & a quantos eran guisados pora ello que andudiessen con armas & con cauallos. (General estoria IV, fol. 9v) ‘And he ordered that all good men, and particularly men of the equestrian class and those who were equipped to do so, should present themselves with arms and with horses.’

While the pattern in (21) to (23) represents the basic system in place in the latter half of the thirteenth century, a small but non-negligible quantity of examples can be found in which the conditional complementizer si/ssi appears to be a licensor for null D. Three examples are given below.

  DP Structure: From Multiple Determiners to Just One 

131



(24) Et si portero fuere a fazer esta entrega que el derecho de la porteria que lo peche aquel que fizo la pendra. (Alfonsine diploma, 6 October 1252, Seville; Archivo municipal de Alcalá de Henares, Leg. 9000/1) ‘And if an official messenger executes the restitution, the cost must be borne by the person who gave the security.’



(25) Pero sobre todo. por que si diluuio uiniesse que non alcançasse a ellos. (General estoria I, fol. 18r) ‘But above all, so that if a flood came it would not reach them.’



(26) e aun te digo mas. que si rayos cayeren del cielo siempre te tembras que ferran entj por la falsedat quem fezist. (Estoria de España I, fol. 28r) ‘And I say this to you as well, that if bolts of lightning fall from the sky you will always fear that they will fall on you because of the way you deceived me.’

In the specified period, the rate of occurrence of this structure – expressed as a proportion of all existential preverbal subject DPs in conditional clauses – is 13.6% (32/235). While this figure points towards a reasonable degree of vitality, it also demonstrates that, even in this context, lexically ungoverned null D was very much a minority option. The fact that conditional si/ssi could license null D at all may be related to the interaction between the conditional operator and the scope of the quantifier to which null D corresponds. As Longobardi notes (1994: 618), this quantifier, in addition to being existential and unspecified for number, takes ‘the narrowest possible scope’. It might therefore be conjectured that the lexical government requirement on null D reflects, at a more fundamental level, a need to exclude quantifiers with narrow scope from occupying the preverbal field, which is usually assumed to be associated with wide quantifier scope (cf. the Mapping Hypothesis of Diesing 1992). Now in conditional clauses, the conditional operator ‘if . . . then’ (or ‘→’ in logical notation) takes scope over any clause-internal quantifier. In such cases, then, even occupancy of the preverbal subject position does not enable the quantifier to have wide scope. It may be this circumstance which provides determinerless DPs with immunity from the lexical government requirement.

132 

I. E. Mackenzie

As regards the expletive use of the article, the Spanish of the latter half of the thirteenth century again largely resembles its modern counterpart, with both the plural and the singular forms being capable of supplying semantically empty phonological material that enables the relevant noun to escape the lexical government requirement associated with null D, together with the latter’s existential interpretation. Examples (27) to (29) illustrate this possibility with plural, singular mass and singular count nouns respectively. (27) Las cabras duran alli .xi. annos. en otra tierra biuen mas. (General estoria I, fol. 256r) ‘Goats last for eleven years there; in another land they live longer.’ (28) El cobre es mas fuerte metal que ell oro nila plata. (Libros del saber de astronomía, fol. 25v) ‘Copper is a stronger metal than gold or silver.’ (29) La yegua trae el parto .xj. meses. & all otro luego de adelant. se emprenna si marido le dan. (General estoria I, fol. 253r) ‘Mares give birth after eleven months and will conceive one month later if given a mate.’

In light of examples such as these, it can be seen that, of the three core features of the modern D-system listed earlier in (20), two were already largely in place by the latter half of the thirteenth century. That is to say, DPs headed by null D were interpreted existentially and, in principal, they had to occur as the complement of a lexical head, while the definite article could be used expletively to obviate the lexical government requirement attaching to null D. Now Batllori and Roca (1998, 2000), Bartra-Kaufmann (2009), Company (1991) and Ishikawa (1997) highlight examples from earlier texts, primarily the Poema de mio Cid, in which the occurrence of bare nouns appears to be somewhat less constrained than was the case in the time period just considered. However, a good proportion of these examples involve articleless singular nouns inside prepositional phrases, a type of occurrence which may ultimately, following Longobardi (1994: 612), be assimilable to the predicative category. Predicative nominals by definition

  DP Structure: From Multiple Determiners to Just One 

133

are non-arguments and hence do not require a D-element, null or otherwise. Other examples involve nouns, like çielo ‘heaven’, that could in principle be treated as proper nouns. DPs consisting in a proper noun are conventionally analysed as not requiring the insertion of a D-element; instead the proper noun is assumed to undergo N-movement itself to the D position.5 With these exclusions, the number of examples from Spanish texts composed before 1250 that unambiguously point towards a different D-system from the post-1250 one becomes very small indeed. Arguably the most suggestive of these examples are (30) and (31) below, the latter being all the more striking on account of the bare subject’s definite reference. (30) Moros le Reçiben por la senna ganar. (Poema de mio Cid, fol. 15v) ‘He is engaged by Moors, who try to capture the ensign.’

(31) E quando alçaua moyses sus manos vençia israel. Manos de moysen eran pesadas prisieron pyedra & pusieron de yuso. e estido sobrella (Fazienda de ultramar, fol. 17v) ‘And when Moses raised his hands, Israel prevailed. The hands of Moses were tired; they took a stone and put it under him and he sat upon it.’

While it is conceivable that the steady state in the D-system that is observable from 1250 onwards was of relatively recent origin, one has to be cautious when basing major claims about usage on isolated examples drawn from the Poema de mio Cid, which is, after all, a verse text. Caution may also be appropriate in respect of example (31), even though it is from a prose text. The extract shown forms part of a direct translation of Exodus 17: 10–15, which also incorporates two iterations of the explanatory comment shown as (32) below, no doubt capturing what the author takes to be the allegorical significance of (31).

(32) Jn hoc loco signyficat/signyficant manos del sacerdot. Qvando las tiene alçadas al altar. (Fazienda de ultramar, fol. 17v) ‘Here it means “hands of the priest” when he has them raised towards the altar.’

As can be seen, the explanatory comment omits the definite article in the phrase manos del sacerdot, which does not result in ungrammaticality

134 

I. E. Mackenzie

because the phrase in this context is not an argument (e.g. a subject or an object) but a metalinguistic gloss. Given that the manos de moysen mentioned in (31) are a symbolic instance of the manos del sacerdot referenced in the gloss, it is possible that the determinerless syntax of the former, in principle ungrammatical, reflects interference from the determinerless syntax of the latter, which is grammatical. Indisputable prose examples which depart from the post-1250 system in respect of null D and the expletive definite article are thus not forthcoming in the literary period. Indeed, as Bartra-Kaufmann (2009: 2–3) implies, it would be surprising if literary texts of the twelfth or early thirteenth century did attest to a situation in which determinerless DPs could occur freely, given that use of the definite article is already mandatory in eleventh century legal documents such as the Fuero de Palenzuela and the Fuero de Sepúlveda. It seems likely, therefore, that the modern-looking features of the D-system that are apparent in the 1250–1289 period, viz. restricted null D and free availability of the expletive article, were not recent innovations at that time.6

4.3.2 Stacked Determiners 4.3.2.1  The Core Data Turning now to the third of the fundamental properties of the modern D-system listed at the end of Sect. 4.2, viz. the prohibition on stacked determiners, it is immediately apparent that in this regard the Spanish of the 1250–1289 period differs significantly from its modern counterpart. For the DP-related syntax of that period, and in fact for some time afterwards, is characterized by the possibility of iterations of two or even three D-elements to the left of the noun. This is illustrated by the possessive DPs in (33) to (36) below, which represent the commonest (though not the only) instance of this pattern.

(33) vn omne demando por un su hermano que guerreaua conel Rey. (Judizios de las estrellas, fol. 90r) ‘A man asked about his brother who was fighting in the king’s service.’

  DP Structure: From Multiple Determiners to Just One 

135



(34) mas en uez de lagremas ayna cadran y gotas de la mi sangre. (Estoria de España I, fol. 29v) ‘But instead of tears, drops of my blood will soon fall there.’



(35) & fallaron le muerto en su lecho cerca aquella su mugier. (General estoria I, fol. 93v) ‘And they found him dead in his bed next to his wife.’



(36) El segundo iuego iogara el Rey prieto con ell otro su peon prieto que esta en la quarta casa daquel mismo cauallo blanco. (Libro de ajedrez, dados y tablas, fol. 59v) ‘The black king will make his second move with his other black pawn which is situated on the white knight’s fourth square.’

Possessive DPs in modern Italian such as la mia casa ‘my house’ and il suo cavallo ‘his/her horse’ are, superficially at least, analogous to some of the structures above. In the Italian case, the possessive element is analysed as being an adjective or adjectival phrase rather than a determiner. However, such an analysis would be inappropriate for Old Spanish, given that the possessive element on its own was capable of satisfying the lexical government requirement alluded to in Sect. 4.2, implying that it was a genuine D-element rather than simply an adjective. This possibility is illustrated in (37) below, where su uirtut occupies the “lexically ungoverned” preverbal subject position:

(37) Et su uirtut es que el qui la touiere consigo non sera ferido de ningun bestiglo malo. ninle nuzra su pozon. (Lapidario, fol. 112v) ‘And its virtue is that anyone who carries it will not be wounded by any terrible beast, nor will its poison harm him.’

This DP-initial occurrence of the prenominal possessive was actually commoner than the embedded type of occurrence illustrated in (33) to (36), the proportions of the two types in the 1250–1289 data being 81.3% and 18.7% respectively (31,344 DP-initial cases versus 7209 embedded cases). The key point, however, is that the concurrent productivity of both patterns rules out an analysis along the lines of modern Italian.7

136 

I. E. Mackenzie

Conversely, the possibility for the medieval (prenominal) possessive to be preceded by one or more other determiners, as in (33) to (36), militates against analysing this item as a phrasal category that is assigned genitive case by the associated noun. That approach is sometimes assumed for Germanic prenominal possessives, analysed as pronouns, and Ishikawa (1995: 209–210) adopts a variant of it for Old Spanish. However, the analysis in question relies on the distributional parallel which exists in Germanic between the prenominal possessives on the one hand and full genitive phrases such as John’s on the other (cf. Giorgi and Longobardi 1991). This parallel is not applicable to medieval Spanish, given that genitive phrases like John’s or the president’s are (i) invariably DP-initial and (ii) mutually exclusive with all other determiners, whereas the prenominal possessive in medieval Spanish exhibits neither of these properties.

4.3.2.2  DP Recursion and Double Articulation Assuming, then, that the medieval Spanish prenominal possessive is neither an adjective, like its modern Italian counterpart, nor a species of genitive phrase, like its counterpart in modern Germanic (according to some analyses), it can plausibly be inferred that it is a legitimate D-element. The highlighted DPs in (33) to (36) should thus be regarded as being recursive; that is to say, they are built up through successive merge of a D-element into the (left) edge of the developing structure. Under these, maximally simple assumptions, a phrase like un su hermano has the structure shown in (38) below.8

(38) [DP un [DP su [NP hermano]]]

Given the analytical framework implied by (38), a parallel can be drawn between the stacked determiner pattern illustrated in (33) to (36) and CP recursion structures such as the clausal complement of pregunto in (39) below, which involves two C-elements, viz. que and si:

(39) & por que non sabie lo que auien fecho por las otras cibdades & los otros logares; quel pregunto [CP que [CP si querie aun que fiziesse y mas.]] (General estoria IV, fol. 190r)

  DP Structure: From Multiple Determiners to Just One 

137

‘And because he did not know what had been done in respect of the other cities and places, he asked her whether she still wanted him to do more there.’

In both this type of case and the one involving stacked determiners, a recursive or iterative structure is built up by merging a head (D or C) with a structure that is already headed by an element belonging to the same syntactic category. As with the D-elements in the recursive DPs shown in (33) to (36), the linearly adjacent complementizers in (39) are non-identical. Indeed, sequences of adjacent identical complementizers appear to have been ungrammatical in Old Spanish, a left-dislocated nominal such as a subject always being inserted between the two iterations of the complementizer, as with dios in (40) below:

(40) & que asman otrossi [CP que dios [CP que non a cuedado delas cosas deste mundo.]] (General estoria IV, fol. 72r) ‘And who also claim that, God, he does not take care of things in this world.’

We thus have a situation in which the linear adjacency of two complementizers was blocked whenever the complementizers were identical but not, as (39) shows, when the complementizers were non-identical. This suggests that a spellout rule or surface filter banning the linear sequence of identical complementizers was operative. Stated in these terms, however, the filter is too specific; for Old Spanish appears to have manifested a DP-related analogue of the same constraint. This latter point can be inferred from examples like (41) and (42) below, which, descriptively at least, are instances of double articulation (Plank 2003; Alexiadou 2014) or the repetition of the definite article within a single DP:

(41) et parte lo que se ayuntare desso. sobre [DP la cuerda [DP la mas luenga]]. (Cánones de Albateni, fol. 17v) ‘And divide the output of this operation by the [value of the] longest chord.’

138 



I. E. Mackenzie

(42) & sennala enel logar del cabo de [DP la sombra [DP la que fue ante de mediel dia]]. A. (Cánones de Albateni, fol. 11r) ‘And mark an “A” at the tip of the pre-midday shadow.’

In both examples, the embedded DP is interpreted as being coreferential with the noun that immediately precedes it, viz. cuerda in (41) and sombra in (42). This suggests that the embedded DPs contain a silent pronominal element, which refers back to the preceding noun and also functions as the nominal element that is modified by the DP-final modifying expression, viz. mas luenga in (41) and que fue ante de mediel dia in (42). Under this analysis, the nouns cuerda and sombra in these examples are analogous to dios in (40), modulo the difference that (40) involves resumption/coreference across a CP boundary whereas (41) and (42) involve resumption across a DP boundary. Furthermore, just as the left-­ dislocation of dios in (40) has the effect of preventing an illicit sequence of two identical complementizers, so the positioning of cuerda and sombra in (41) and (42) prevents an illicit sequence of two identical determiners. Plausibly, then, both types of structure reflect the operation of the same constraint, viz. a prohibition on linear sequences of identical heads. The highlighted DPs in (41) and (42) can thus be regarded as a being a special case of the recursive DP structure represented in (38), viz. one in which a “dislocated” nominal intervenes between otherwise adjacent instances of D.  Assuming, for simplicity, that the unit created by the insertion of this nominal has the label ‘NP’, the structures can be shown more explicitly as in (43) and (44), where ‘e’ stands for the silent pronominal element which resumes the dislocated nominal.

(43) [DP la [NP cuerda [DP la [NP e mas lengua]]]]



(44) [DP la [NP sombra [DP la [NP e que fue ante de mediel dia]]]]

Under this analysis, the initially mysterious existence of double articulation in Old Spanish can be explained as a by-product of the general availability in the medieval language of DP recursion, the most transparent instance of this phenomenon being the stacked pattern of determiners illustrated by the earlier examples (33) to (36). Viewing the phenomenon

  DP Structure: From Multiple Determiners to Just One 

139

in this light also obviates the need for the rather complex derivations based on remnant small clause movement proposed by Kayne (2008) and Bartra-Kaufmann (2013).

4.3.2.3  The Ordering of D-Elements Returning to the structures in (33) to (36), it can be further noted that not any ordering of determiners was possible: formulations analogous to the DPs shown in bold in those examples but with the determiners re-­ ordered, viz. *su un hermano, *mi la sangre, *su aquella mugier and *otro el su peon, appear to have been ungrammatical. This can be related to the fact that it was from the outermost D that the DP overall inherited its value for the definiteness feature. For example, the indefiniteness of un su hermano is determined by the indefiniteness of un, which is the outermost D-element in the structure.9 In the same way, the [+ definite] value of la mi sangre is inherited from [+ definite] la, also in the outermost D position. Given that both indefinite un su hermano and definite la mi sangre contain an internal possessive determiner, this latter item must be assumed to have been unmarked for definiteness when it occupied an inner D position. On the other hand, in view of the fact that bare possessive DPs like mi uerguença ‘my shame’, su coraçon ‘his heart’ etc. are [+ definite], it can be inferred that the possessive was [+ definite] whenever it occupied the outermost D position. Structures like *su un hermano thus involve a definiteness conflict, between [+ definite] su and [− definite] un, and it is to this circumstance that their ungrammaticality can be attributed. The distribution of otro appears to have paralleled that of the prenominal possessive. In an internal position it was unmarked for definiteness, appearing equally in definite DPs like ell otro su peon prieto in (36) and in indefinite ones such as algun otro hospital in (45) below. However, when it occupied the outermost D position, as in otro sancto logar, also in (45), it was [− definite].

(45) [. . .] ay algunos que se prometen a iherusalem pora seruir enel tiemplo. & otrossi a algun otro ospital. o a otro sancto logar. (General estoria I, fol. 277v)

140 

I. E. Mackenzie

‘There are some who promise themselves to Jerusalem in order to serve in the temple or to some other hospital or to another holy place.’

Accordingly, a formulation like *otro el su peon was ungrammatical for the same reason that *su un hermano was ungrammatical. That is to say, it involves a definiteness conflict between [− definite] otro and [+ definite] el.10 As regards formulations like *mi la sangre and *su aquella mugier, there is no definiteness conflict, given that in both cases the two occurrences of D are [+ definite], the possessive by virtue of occupying the outermost D position and the definite article and the demonstrative by virtue of being inherently [+ definite]. However, this type of case can be brought within the scope of a refined version of the spellout rule introduced in Sect. 4.3.2.2. According to Chomsky (2013), the presence of shared salient or ‘prominent’ semantic features can block certain syntactic operations. In the same spirit, the spellout rule of Sect. 4.3.2.2 could be re-expressed in more abstract terms as a ban on linearly adjacent heads (belonging to the same syntactic category) that share a salient semantic feature, such as [+ declarative] in the case of que or [+ definite] in the case of the definite article. Revised in this way, the filter also excludes formulations like *mi la sangre and *su aquella mugier, given that these involve a linear sequence of two [+ definite] D-elements. Their ungrammaticality can thus be assimilated to that of sequences like *la la cuerda mas lengua, a failed version of the DP la cuerda la mas lengua contained in the earlier example (41). Indeed, the spellout violation is, in both cases, remedied by exactly the same strategy. This is shown in (46) and (47), where the nouns logar and usos, by separating a DP-initial possessive determiner and a following definite article or demonstrative, achieve the same effect as the noun cuerda in the DP la cuerda la mas lengua.

(46) Desi mueue ell orizon enclinado de [DP su logar [DP el primero]] .lx. grados. et faz assi como es ante dicho. (Libros del saber de astronomía, fol. 125r) ‘Afterwards move the tilted horizon from its initial position by sixty degrees and do what was previously specified.’

  DP Structure: From Multiple Determiners to Just One 



141

(47) & nos guardaremos [DP nuestros usos [DP aquellos que auemos de nuestros padres antigos]]. (Estoria de España I, fol. 170v) ‘And we will keep to our customs which we inherited from our ancestors.’

On the face of it, therefore, two constraints appear to govern the ordering of D-elements within the recursive medieval DP: (i) a prohibition on definiteness conflicts and (ii) a spellout rule preventing like-featured heads from being linearly adjacent. However, the existence of examples like (48) below calls for one final refinement of the analysis, whereby the two constraints turn out, in fact, to be expressions of a single more abstract rule.

(48) dixo que no era nada uencer por palaura. mas que troxiessen alli delante [DP un toro [DP el mas fuerte que auer pudiessen]]. (Estoria de España I, fol. 116r) ‘He said it was nothing to win by words and that they should bring forth a bull of the strongest kind that they could find.’

As can be seen, there is a definiteness conflict between [− definite] un and [+ definite] el in (48), although the containing DP un toro el mas fuerte que auer pudiessen is fully grammatical. However, the two conflicting determiners are separated by the word toro, which is analogous to the “dislocated” type of nominal that occurs in the double articulation structures examined in Sect. 4.3.2.2. As was just outlined, the occurrence of such an item prevents a violation of a spellout rule prohibiting like-­ featured heads from being linearly adjacent. The fact that the same strategy obviates the definiteness conflict in (48) implies that the spellout rule in question and the prohibition on definiteness conflicts are manifestations of the same filter. In view of the prohibited structures examined so far, as well as the grammaticality of examples like (48), what this filter appears to block is the linear adjacency of heads (belonging to the same syntactic category) whenever they are valued for the same (prominent) semantic feature, regardless of whether these values are identical or conflicting. Ungrammatical sequences like *el el or *su el can be seen as illustrating the first case and sequences of the type *un el or *su un can be seen as illustrating the second one. As regards the grammaticality of sequences

142 

I. E. Mackenzie

like el su or un su, this would follow from the fact that only the first D-element in each sequence carries a value for the definiteness feature, given that the possessive determiner is unmarked for definiteness unless it occupies the outermost D position (see the discussion at the beginning of this section).

4.3.3 The Indefinite Partitive Determiner A further point of divergence between the modern system and the system that can be reconstructed for the second half of the thirteenth century is the existence in the latter  – albeit at a low level of frequency  – of an indefinite determiner constructed from the preposition de followed by an expletive occurrence of the definite article: del, de los, de la, de las. This determiner is illustrated in (49) and (50) below.

(49) Toma delos huessos delas oliuas & del alfoztec. (Picatrix, fol. 32v) ‘Take some olive stones and some pistachio’



(50) e despues tomen del uerde e del uermeion e de la sal tanto delo uno como de lo al. e muelanlo e mezclenlo con del uinagre. e despues remogen del algodon en ello. (Libro de las animalias, fol. 180v) ‘And afterwards take some green barley and some cinnabar and some salt, in equal portions, and grind it and mix it with vinegar. Then soak some cotton wool in it.’

As can be seen, the determiner in question corresponds to an existential quantifier and selects either a singular mass NP or a plural count NP as its complement. It is thus analogous to the modern French or Italian partitive article when the latter is used with indefinite meaning. Carlier and Lamiroy (2014) analyse the evolution of partitive constructions in Romance in terms of a ‘grammaticalization chain’. According to their proposal, Old Spanish did not go beyond Stage III of this chain, the stage at which the partitive construction has not yet developed an indefinite meaning but instead ‘presupposes extraction from a contextu-

  DP Structure: From Multiple Determiners to Just One 

143

ally (deictically or anaphorically) defined partition set’ (ibid. p.  502). While this seems likely to be true for the vast majority of examples, (49) and (50) attest to the existence of a small number of cases in which there is no partition set as such, merely the open-ended universe of discourse over which a regular indefinite determiner quantifies. Accordingly, cases like (49) and (50) can be treated as reflecting reanalysis of the preposition de as a component of a morphologically complex D-element. This type of occurrence of de should therefore be distinguished from the superficially similar partitive prepositional use illustrated by examples (51) and (52) below.

(51) Et si destempraren della en el agua. & metieren y huuas passas; torna las frescas como ante eran. (Lapidario, fol. 86r) ‘And if it is dissolved in water and dry grapes are soaked in the solution, it makes them fresh again, as they were before.’



(52) Ca fallamos que de muchas carnes solien comer los iudios ante deste mandado. de que non comieron despues. (General estoria I, fol. 237r) ‘For we find that the Jews ate many different kinds of meat before this commandment, which they ceased to eat afterwards.’

Here, the preposition de is external to the DP, which either does not require an overt D-element, ella in (51) being a pronoun, or it already has one, viz. muchas in (52). This differs from what is observed in (49) and (50), where in each case the unit formed by de and the definite article actually is the D-element inside the DP. The de of (51) and (52) can thus be regarded as an independent preposition expressing a partitive relationship between its DP complement and the governing verb, viz. destempraren or comer. While this use of the preposition de survives residually in modern Spanish formulations such as de esta agua no beberé ‘I won’t drink any of that water’ (Lapesa 2000: 80) or tener de todo ‘to have all kinds of things’, the partitive determiner illustrated in (49) and (50) is now completely extinct. From a functional perspective, the Old Spanish indefinite partitive determiner was in competition with both null D and the plural forms of the indefinite article derived from ūnus, i.e. unos and unas. The availabil-

144 

I. E. Mackenzie

ity of the latter forms probably exercised an inhibitory effect on learners’ propensity to reanalyse the sequence de + los/las as a determiner, given that examples of the medieval partitive determiner with plural count nouns are much rarer than are examples with singular mass nouns. This is one way in which the Old Spanish partitive determiner differed from its equivalent in modern French or Italian, where the determiner is productive in both the singular and the plural. A second difference relates to the possibility for associated pronominals to be spelled out using a bespoke partitive form, usually a reflex of Latin inde ‘thence’. In modern French, for example, pronominal objects corresponding to indefinite partitive DPs are obligatorily realized by the partitive clitic en, as is illustrated by the clitic left dislocated structure in (53):

(53) Des erreurs, j’en ai commis. ‘Errors, I’ve made some.’

In contrast, ende, the equivalent element in Old Spanish, almost never occurred as an indefinite object. One of the very small number of examples where it does is given in (54) below. (This example comes, incidentally, from the very text which accounts for the majority of the thirteenth-century examples of the indefinite partitive determiner.)

(54) e la carne de los ratos pequennos e delos mures. es les muy buena. mayor mient sigelos dieren con sos cueros. e si les ende dieren cada dia; fazelles a grand pro. (Libro de las animalias, fol. 29r) ‘And the flesh of small rats and mice is good for them, especially if they are given with their skin. And if you give them some each day it will benefit them greatly.’

Elsewhere, however, ende invariably references a prepositional relationship of some sort, as exemplified in (55) and (56):

(55) E encabo tanto los aquexo la grand sed. que poro estauan se cayen muertos. tan bien omnes como bestias. & los que ende fincauan. estauan todos por perder se. (Estoria de España I, fol. 48r)

  DP Structure: From Multiple Determiners to Just One 

145

‘In the end they were so afflicted by thirst that wherever they were they fell dead, men and beasts alike, and those of them who remained were at the end of their strength.’

(56) que quiso una pieça auer de los sos sabios que gelo espusiessen & a paladinassen. & ouo ende setaenta. (General estoria IV, fol. 246v) ‘He summoned a quantity of his wise men to explain it and clarify it, and there were seventy of them.’

Considering both the near absence in Old Spanish of a bespoke indefinite object pronoun corresponding to French en or Italian ne and the low productivity of the plural form of the medieval Spanish partitive determiner, it is clear that the latter item, even in the thirteenth century, had a peripheral role in the grammar of the period. The emergence and subsequent disappearance of this determiner should be regarded as a further instance of a failed change, although the very reduced numbers of tokens mean that quantitative analysis is not feasible in this particular case. As regards the reason why the Spanish structure did not become permanent, in contrast to what happened in French and Italian, one possibility would be that the accusative case-marking preposition a was implicated, given Stark’s (2008: 55–56) allusion to the existence of an inverse correlation in the Romance languages between the availability of differential object marking (DOM) and the development of an indefinite partitive determiner. However, an explanation along these lines is called into question by the fact that DOM in Spanish was already well established by the time the indefinite partitive emerged. Indeed, the very text which contains the bulk of the thirteenth-century examples of the indefinite partitive also has many examples of accusative a, as (57) illustrates.

(57) E quando ouieren fecho so fornicio. metan a las fembras en sus traellas. (Libro de las animalias, fol. 193v) ‘And when they have mated, use the females to separate them.’

A more likely explanation for the ephemeral nature of the Old Spanish indefinite partitive lies in the existence of the well-established plural forms of the indefinite article, viz. unos and unas. As was mentioned above, these were functionally equivalent to the plural forms of the

146 

I. E. Mackenzie

indefinite partitive delos and delas, the productivity of which appears to have been inhibited accordingly. While the singular indefinite partitive used with mass and abstract nouns may well have had the potential to become established on a permanent basis, the extreme dearth of plural tokens in the primary linguistic data probably deprived learners of the necessary cues for postulating the existence of a fully-fledged partitive system.

4.4 Quantitative Evolution of the DP 4.4.1 Loss of DP Recursion The principal change which has occurred in the Spanish DP system since the High Middle Ages is the loss of DP recursion. In terms of surface expression, this change is manifested partly as a progressive loss of structures involving multiple determiners, and partly as what Lapesa (2000: 434) called ‘una tendencia a repartir los actualizadores de modo que el sustantivo quede en el centro del sintagma como núcleo suyo’.11 The paradigm instance of the first process consists in possessive DPs like la mi tierra ‘(the) my land’ being outcompeted by the corresponding articleless forms. Rosemeyer and Enrique-Arias (2016: 328) situate this development within the context of a gradual diffusion of the articleless possessive into usage contexts previously associated with other possessive constructions, a trend which in their view caused the form involving the definite article to become increasingly restricted to emphatic contexts (cf. Eberenz 2000: 265–319). As regards the process referenced by the foregoing Lapesa quotation, this primarily involves indefinite DPs like una su hermana ‘a sister of hers’ or algun su amigo ‘some friend of his’ being outcompeted by functionally equivalent structures which make use of the postnominal ‘strong’ possessive, as in una hermana suya and algún amigo suyo. The same development also affected demonstrative possessive DPs, in the sense that modern formulations like aquel hijo suyo ‘that son of his’ represent an obvious functional continuation of aquel su fijo. However, the availability in modern

  DP Structure: From Multiple Determiners to Just One 

147

Fig. 4.1  Evolving rates of occurrence of three recursive DP structures

Spanish of strings like esta nuestra ciudad ‘this, our city’ or aquella mi propuesta ‘this, my proposal’, which presumably involve a pronominal DP and a lexical one in apposition, raises the possibility that recursive demonstrative DPs were reanalysed as appositive structures, following the pattern illustrated in (57):

(57) [DP esta [DP nuestra [NP ciudad]]] > [DP esta] [DP nuestra [NP ciudad]]

Given that the appositive structure is formally identical to the recursive one, in any given instance it is difficult to know with certainty whether the token instantiates the recursive structure characteristic of the High Middle Ages or the appositive structure which was capable of being generated by the grammar even after DP recursion was lost. Figure 4.1 shows the evolving rates of occurrence of the three commonest instantiations of the recursive DP (the corresponding numerical data can be found in Table A.6 in Appendix 1).

Example la mi sangre un su hermano aquella su mugier

Recursive DP

Definite article + possessive + NP Indefinite D + possessive + NP Demonstrative + possessive + NP

Possessive + NP Indefinite D + N + strong possessive Demonstrative + N + strong possessive

Competing form

Table 4.1  Recursive possessive DPs and their competitors in Medieval Spanish mi sangre un hermano suyo aquella mugier suya

Example

148  I. E. Mackenzie

  DP Structure: From Multiple Determiners to Just One 

149

Table 4.2  Rate of change in use of the recursive possessive DPs (1250–1419) Structure

Odds ratio (per decade)

P-value

Definite article + possessive + NP Indefinite D + possessive + NP Demonstrative + possessive + NP

0.978 0.987 0.922

0.223 0.668 0.009

For each of the three types of DP, the rate of occurrence is relativized to the frequency of the relevant competing structure. For this purpose, I assume the axes of competition shown in Table 4.1. Given the foregoing, the percentage rates in Fig.  4.1 should not be understood as reflecting the absolute frequencies of the structures in question. Indeed, the recursive structure with the lowest rate when measured against its competitor, viz. ‘definite article + possessive + NP’, has, for most of the time period shown, by far the highest absolute frequency of the three types of recursive DP. This discrepancy is due to the fact that the functional space hosting competition between ‘definite article + possessive + NP’ and the corresponding articleless structures was much larger than was the space defined by the competition between, say, ‘un + possessive + NP’ and ‘un + N + strong possessive’. However, in terms of assessing the vitality of a structure, it is the frequency within the relevant competition space which counts and not the absolute frequency (cf. the Principle of Accountability in Tagliamonte 2012).12 As can be seen from Fig. 4.1, the recursive system is fairly stable until the early decades of the fifteenth century. Visually, this is indicated by the absence of any clear upward or downward trend in the three lines during that period. This visual impression is largely confirmed by regression analysis (based on the numerical data contained in Table A.6), the results of which are summarized in Table 4.2. All three odds ratios shown in the table have a value which is lower than 1, implying that any trend was a downward one. However, the odds ratios for the definite and indefinite possessive DPs are very close to 1, the latter number implying no change at all. For example, the OR of 0.987 for the indefinite possessive DPs indicates that the odds (or relative probability) of the structure declined by 1.3% per decade, which is a fairly negligible rate. As a consequence, the p-values for these structures point towards quite a high probability that the slight downward movement

150 

I. E. Mackenzie

Table 4.3  Rate of change in use of the recursive possessive DPs (1410–1609) Structure

Odds ratio (per Percentage decade) equivalent decline

Definite article + possessive + NP 0.828 Indefinite D + possessive + NP 0.832 Demonstrative + possessive + NP 0.985

17.2 16.8 1.5

P-value 0.000 0.000 0.429

observed is due to chance rather than reflecting a genuine trend or ‘effect’. For indefinite DPs, the p-value indicates that there is a 66.8% probability that the movement is due to chance, while for the definite DPs, the p-value points to a 22.3% probability of this being the case. Generally speaking, a probability of 5% or higher that the findings could be due to chance is regarded as indicating that the apparent trend is not ‘statistically significant’. At 0.922, the OR for the demonstrative structures is a little more suggestive, implying a decline of 7.8% per decade. Moreover, the corresponding p-value is 0.009, which is well inside the conventional 5% significance limit. However, both this sub-5% p-value and the relatively high rate of decline are no doubt a by-product of the sudden and somewhat isolated dip in the demonstrative rate that is visible in Fig. 4.1 at the 1380 data point, which will have had an effect on the regression procedure. Accordingly, and given the general stability of the system as evidenced by the lines (Fig. 4.1) and ORs (Table 4.2) for the definite and indefinite structures, it is probably safer to assume that DP recursion generally, including the case in which the outermost D was a demonstrative, underwent little change across the thirteenth and fourteenth centuries in terms of frequency of usage. However, from the earliest decades of the fifteenth century a major restructuring appears to take place. From that time onwards, as can be seen in Fig. 4.1, both the indefinite and definite types of recursive DP are in a clear process of decline (modulo a short-lived upswing in the definite category at the 1440 data point), while the demonstrative type maintains a broadly horizontal trajectory. The relevant odds ratios and p-values are given in Table 4.3. In contrast to what we see in the thirteenth and fourteenth centuries, the ORs shown in Table  4.3 for the definite and indefinite structures point to robust rates of decline, the equivalent decay in percentage terms

  DP Structure: From Multiple Determiners to Just One 

151

being 17.2% and 16.8% respectively per decade, with associated p-values at zero when estimated to three decimal places. It is also worth noting that these rates of decline are very similar to each other, suggesting, given Kroch’s Constant Rate Effect (see Sect. 1.2.1), that a single underlying trend is at work across both contexts. This is to be expected if, as is assumed here, the two types of structure are different instances of a single type of syntax, characterized here as DP recursion. As will be recalled from the discussion in Sect. 3.3.2, while Kroch’s Constant Rate Effect implies a common rate of change across the different contexts in which a change is manifested, it also envisages that the synchronic percentage rates may be non-identical across these contexts. This appears to be the situation in the present case, at least as regards the definite and indefinite structures. For while the rates of change in these two contexts are approximately equal (as has just been observed), it can be seen from Fig. 4.1 that the corresponding synchronic percentage rates of usage are always very different, the indefinite structure exhibiting a higher value than the definite structure at all the data points in the figure. This highlights, once again, the fact that it is the rate of change which is important in identifying connections between changes rather than the productivity at any given moment of the specific structures involved. An interesting side effect of this phenomenon in the present case is that, although the rate of change is constant across the definite and indefinite contexts, the large difference in the synchronic percentage rates results in a situation in the early to mid- 1500s in which the definite variant of the recursive structure has close to zero frequency while the indefinite variant is still productive. The indefinite variant of the recursive possessive DP would thus have still been in use in the early colonial period, which is consonant with the fact that possessive DPs of the form un mi amigo survive in certain Latin American dialects even today (see e.g. Elsig 2017). As regards the OR shown in Table 4.3 for the demonstrative structure, this is noticeably different from that which is reported for the definite and indefinite ones. At 0.985, it implies only a very modest decline in the odds – 1.5% per decade – an effect which could easily be due to random variation in the data. Indeed, the corresponding p-value of 0.429 implies that the demonstrative structure was not subject to any statistically significant trend over the relevant period. Now it is plausible to suppose that

152 

I. E. Mackenzie

this divergence vis-à-vis the definites and indefinites is due to the fact that the demonstrative recursive DP was capable of being reanalysed as an appositive structure, as illustrated in (57) given earlier. Loss of DP recursion would necessarily lead to a decline in strings like un/el/algun su amigo, whereas strings like esta/esa/aquella su tierra, reanalysed as involving apposition rather than recursion, could survive this change. It is thus not unreasonable to imagine that the demonstrative recursive DP underwent an analogous decline to the definite and indefinite ones, but that tokens of this structure were replaced by superficially identical appositive tokens approximately as fast as the recursive tokens were lost. From that perspective, the impression given by Fig.  4.1 that demonstrative possessives were immune to the process undergone by their definite and indefinite counterparts is somewhat misleading. It is probably safer to assume a general loss of DP recursion, masked in the case of demonstratives by the possibility of generating the same linear string as the recursive structure but with appositive rather than recursive syntax.

30% 25% 20% 15% 10% 5% 0% 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

Fig. 4.2  Late medieval increase in ‘definite article + possessive + NP’

1560

1590

  DP Structure: From Multiple Determiners to Just One 

153

4.4.2 T  ransient Increase in ‘Definite Article + Possessive + NP’ Despite the general picture concerning the recursive DP of stability followed by decline, a number of authors (e.g. Clavería Nadal 1992; Batllori 2010) have suggested there was an increase in the use of the form involving the definite article in the later Middle Ages. An effect of this type appears to be apparent in the data collected here. The relevant increase is obscured to an extent in Fig. 4.1 on account of the vertical axis showing the full percentage scale (needed because of the high percentage values for the indefinite and demonstrative structures). However, it can be seen clearly in Fig. 4.2, where the vertical axis has a maximum value of 30%. As can be seen, the rate of occurrence of the structure ‘definite article + possessive + NP’ enjoys a late surge towards the end of the 1300s, peaking at the 1440 data point with a value of 25.2% (see Table A.6 in Appendix 1) before subsequently declining. Although the rate dips in the early 1400s, the increase appears to begin at the 1380 data point (which averages the data for the period 1370–1389), suggesting that 1370–1449 should be identified as the period of increase. During this period, the decadal OR for the structure is 1.103, which equates to a decade-on-­ decade growth of 10.3% in the odds or relative probability.13 The exactly equivalent rate of decline would involve an OR of 0.907 or, in percentage terms, a decrease of 9.3% per decade.14 This is actually rather different from the observed rate of decrease from 1440 onwards, the OR in this case being 0.755, equating to a decade-on-decade contraction in the odds of 24.5%. Given this difference, the bell curve described by the transient increase in ‘definite article + possessive + NP’ should not be deemed to be symmetrical, suggesting that this particular ‘failed change’ (see Sect. 1.2.3) is again of the accidental variety rather than the inherent one. In accidental failures, the decline and the increase represent separate events. In the present case, the decline can be linked to the general abandonment of the recursive type of DP, which was an emblematic feature of the medieval syntax. In the context of this general trend, the fact that the definite variant of the recursive structure rallies towards the end of the Middle Ages is somewhat anomalous. Probably this latter event should be

154 

I. E. Mackenzie

seen as stylistically rather than structurally motivated, meaning it would be analogous to the transient increase in V2 structures discussed in Sect. 2.5. This assessment is consonant with Lapesa’s claim (2000: 434) that forms like los mis ojos ‘my eyes’ came to represent the ‘uso casi general del estilo noble’;15 with Rosemeyer and Enrique-Arias (2016: 328) observation that in fifteenth century biblical translations, as compared with thirteenth century ones, the use of this structure ‘continues, or even increases, in a number of contexts that can be described as emphatic’16 and with Clavería Nadal’s finding (1992: 357) that in the fourteenth and fifteenth centuries ‘es característico de ciertas expresiones y fórmulas de los documentos que salen de la Cancillería’.17 From this perspective, the late medieval increase in the use of the article before prenominal possessives reflects a transient fashion, albeit one which has sufficient quantitative density, in written registers at least, to temporarily impinge on the system’s overall direction of travel.

4.5 Conclusion Like the clausal left periphery considered in Chap. 2 and clitic linearization discussed in Chap. 3, DP structure is an area of the grammar in which Old Spanish distinguishes itself in important ways from the modern language. On the other hand, the syntax which characterizes the medieval DP, as manifested from the outset of the literary period, is not in any appreciable sense more Latin-like than its modern counterpart. In the earliest prose manuscripts, recognizably modern constraints on the distribution of the null determiner are already in place, as is the availability of an expletive definite article. It is true that conditional clauses in thirteenth century texts can license determinerless subjects in a way that would not be possible in the modern language; however, such clauses are a subclass of so-called polarity contexts and these are known to license null D in structures in which it would normally be ungrammatical (cf. Roberts and Roussou 2003: 151). What in fact distinguishes the medieval DP is its frequently recursive syntax, which in terms of surface output is reflected in the presence of two or sometimes three D-elements stacked to the left of the nominal

  DP Structure: From Multiple Determiners to Just One 

155

nucleus, configurations which clearly owe their existence to Romance innovation rather than any Latin inheritance. A striking by-product of the availability of DP recursion was the existence in the language of the period of double articulation structures such as la cuerda la mas luenga ‘the longest chord’. This latter pattern can be likened to CP recursion with subject left dislocation. Both configurations evince a similar strategy for preventing two identical heads from being linearly adjacent, suggesting the existence of a surface filter prohibiting the occurrence of such strings. The eventual loss of DP recursion provides a further illustration of Kroch’s Constant Rate Effect, as well as a reminder that a common rate of change does not imply a uniform rate of synchronic usage. Thus while forms like el su amigo and un su amigo decline at approximately the same rate, the former is moribund by the mid-1500s whereas the latter retains sufficient vitality to take root in the emergent colonial dialects, surviving in isolated pockets even today. In addition, as with other changes in the history of Spanish, the loss of DP recursion is accompanied by a transient quantitative shift against the tide of history, as possessive DPs introduced by the definite article enjoy a brief upsurge in popularity at the end of the Middle Ages. This event is reminiscent of the late medieval increase in V2 structures discussed in Chap. 2, both processes no doubt reflecting a vogue for archaic syntactic patterns which had become imbued with expressive or stylistic value. The discussion in the chapter also highlights the transient existence in Old Spanish of an indefinite partitive determiner, which was analogous in key respects to the indefinite partitive determiners in modern French and Italian. The failure of the Spanish structure to become permanent is likely to be due to competition from unos and unas, given the existence of a number-related asymmetry in the occurrence of the indefinite partitive, plural examples being much rarer than singular ones.

156 

I. E. Mackenzie

Notes 1. In fact, according to these authors (ibid. p. 252), grammar competition between el, la etc. qua article and qua demonstrative continued until the seventeenth century. On the other hand, in a later paper, Batllori (2003) concludes that in the thirteenth century novel Calila y Dimna, the older, etymological grammar had only residual vitality. 2. The kind of postverbal subject illustrated by (8) should not be confused with the postverbal ‘subject’ of an unaccusative or presentational verb, which is usually analysed as being thematically an object. In (i) below, for example, salir is unaccusative/presentational and its postverbal ‘subject’ agua can accordingly be analysed as being in object position. By the same token it is “lexically governed” and hence there is no violation of the lexical government requirement. (i)

Salía agua de la canilla. ‘Water was leaking from the gutter.’

3. See Batllori and Roca (2000) for an alternative view, according to which the only true D-element is the definite article. 4. I assume that in modern formulations like esta tu casa ‘this your house’, the demonstrative is a pronoun rather than a determiner (cf. Cornilescu 1992). The relation between esta and tu casa is thus one of apposition. As will become apparent in Sect. 4.3.2, this analysis does not extend (at least not generally) to superficially analogous Old Spanish examples like aquella su mugier ‘his wife’. 5. The same approach may also be appropriate for very late medieval examples such as (i) below, from an incunabulum printed in 1489: (i) ca amores an tanto poder que hazen hazer las cosas como les plaze (Historia de la linda Melosina, fol. 23v) ‘for love has such power that it makes people do things as it wishes’ Although nominally plural, the subject amores is potentially interpretable as a proper name for the personified phenomenon of love. Assuming N-to-D movement, no lexical government requirement would obtain and the phrase as a whole would escape the existential interpretation attaching to null D, exactly as is the case.

  DP Structure: From Multiple Determiners to Just One 

157

6. In terms of the expletive article, the Spanish of the 1250–1289 period is arguably more ‘advanced’ than its modern counterpart. For we occasionally find semantically empty occurrences of the definite article in contexts in which null D would suffice; that is, in complement position and with an existential interpretation: (i)

e den a las menores segund su guisa. o usen en dar les las moscas e las abeias. (Libro de las animalias, fol. 75v) ‘and the smaller ones should be given however much they want, or they should be given flies and bees’

The structure illustrated here foreshadows the indefinite plural definites of modern Italian discussed by Zamparelli (2002). 7. The bare possessive structure exemplified in (37) appears to have been semantically equivalent to the specific structure illustrated in (34), i.e. ‘el + possessive + NP’, both types of DP receiving a [+ definite] interpretation. Various proposals have been advanced as to the existence of pragmatic factors which may have governed the alternation between the two. For example, Eberenz (2000: 265–319) and Lapesa (2000: 433) link use of ‘el  +  possessive  +  NP’ to expressivity or stylistic tendencies, while Company (2006) associates it with contexts in which either the possessor or the possessed item had a high degree of cognitive prominence. As Pountain (2001: 287) observes, however, the putatively marked or emphatic value of the structure with the article ‘is not always self-evident’. 8. For an alternative, more complex approach, in the cartographic framework, see Batllori (2010). 9. On the indefiniteness of this structure, see also Company (2006: 96, 99). 10. Examples like (i) below should not be seen as exceptions to the prohibition on definiteness conflicts assumed in the text, despite the co-­ occurrence of el and un. (i)

& cata ala estrella con el un oio çerrando ell otro. fata quela ueas delos dos forados delas axatabas. (Libros del saber de astronomía, fol. 74r) ‘And look at the star with one eye while keeping the other closed, until you see it in both pinnule apertures.’

158 

I. E. Mackenzie

The DP in bold in this example can be seen as a special instance of the now archaic construction discussed by Bosque and Picallo (2012) and Lapesa (2000: 488–514) in which a numeric partitive phrase is introduced by the definite article but retains its indefinite value, as in tomaron a los dos de los soldados ‘they took two of the soldiers’ (Pedro de Aguado, Historia de Santa Marta y Nuevo Reino de Granada). In (i) above, the partitive meaning is implicit rather than explicit, relying on the tacit knowledge that the eye in question comes from the specific pair of eyes belonging to each individual reader. Modulo this difference, however, the two structures are fundamentally the same. For Bosque and Picallo, the apparent definite article in these partitive structures is actually a pronoun rather than a D-element; hence there is no definiteness conflict or, in their terms (ibid. p.  140), no violation of the ‘Anti-uniqueness Condition’. 11. ‘A tendency to separate the modifiers so that the noun is left in the centre of the structure as its nucleus.’ 12. It is interesting to note that the evolving rate captured in Fig. 4.1 for the competition between ‘definite article + possessive + NP’ and the corresponding articleless form is broadly in line with Clavería Nadal’s (1992: 353) findings. On the basis of notarial documents, she estimates the occurrence rate of the structure involving the definite article to be 18.2% in the fourteenth century and 22.4% in the fifteenth. I assume that the corresponding rate for the thirteenth century is intended to be 14% rather than the 86% shown, given that she describes the rate as increasing between the thirteenth and fourteenth centuries. On the other hand, Rosemeyer and Enrique-Arias (2016) give much higher estimates. According to their Table  3 (ibid. p.  318), ‘definite article  +  possessive + NP’ in direct competition with the bare possessive has a rate of 48.0% (599/1247) in the thirteenth century and 36.9% (1148/3111) in the fifteenth. It seems likely that the discrepancy vis-à-vis the present findings (and those of Clavería Nadal) can be attributed either to the fact that Rosemeyer and Enrique-Arias base their own findings exclusively on biblical translations or to their choice of methodology, which relies on seeing how possessive expressions in the Latin Vulgate are translated in the corresponding Spanish texts. 13. As usual, the OR is estimated by logistic regression, applied to the relevant data in Table A.6 in Appendix 1. 14. The value 0.907 is the reciprocal of the OR associated with the increase, i.e. 1/1.103 (for an explanation, see Sect. 1.2.3).

  DP Structure: From Multiple Determiners to Just One 

159

15. ‘Almost generalized usage in elevated styles.’ 16. Note that Rosemeyer and Enrique-Arias’s comment should be seen in the context of their finding of an overall decline in the rate of the structure in the fifteenth century translations as compared with the thirteenth century ones. See note 12 above. 17. ‘It is characteristic of certain expressions and formulas that come out of the Chancellery.’

References Alexiadou, Artemis. 2014. Multiple determiners and the structure of DPs. Amsterdam: John Benjamins. Bartra-Kaufmann, Ana. 2009. Some remarks about the grammaticalization process of the DP functional domain in Old Romance. In Proceedings of the IV Nereus international workshop on definiteness and DP structure in Romance languages, Arbeitspapier, ed. María Teresa Espinal et  al., vol. 124, 1–26. Konstanz: University of Konstanz. ———. 2013. Indirect evidence of D development in Romance: ‘Internal’ D. Paper given at the Workshop European Dialect Syntax VII, 13–15 June, University of Konstanz. Batllori, Montserrat. 2003. La sintaxis de los determinantes en el Calila e Dimna. In Actas del IV congreso de lingüística general (Cádiz, 3–6 de abril de 2000), vol. II, ed. María Dolores Muñoz Núñez et al., 191–203. Cádiz: Servicio de Publicaciones de la Universidad de Cádiz. ———. 2010. La periferia izquierda del sintagma nominal: artículo ante posesivo en español medieval. In Actes du XXVe congrès international de linguistique et de philologie romanes, ed. Maria Iliescu et al., 419–430. Berlin: Walter de Gruyter. Batllori, Montserrat, and Francesc Roca. 1998. Los determinantes y los cuantificadores en el español medieval y moderno. Paper delivered at the 28th symposium of the Sociedad Española de Lingüística (Relaciones entre Semántica y Sintaxis), Madrid, 14–18 December. ———. 2000. The value of definite determiners from Old Spanish to Modern Spanish. In Diachronic syntax: Models and mechanisms, ed. Susan Pintzuk et al., 241–254. Oxford: Oxford University Press.

160 

I. E. Mackenzie

Bosque, Ignacio, and M. Carme Picallo. 2012. Articles as partitives. In Functional heads: The cartography of syntactic structures, ed. Laura Brugé et  al., vol. 7, 138–149. Oxford: Oxford University Press. Carlier, Anne, and Béatrice Lamiroy. 2014. The grammaticalization of the prepositional partitive in Romance. In Partitive cases and related categories, ed. Silvia Luraghi and Tuomas Huumo, 477–522. Berlin: Walter de Gruyter. Chomsky, Noam. 2013. Problems of projection. Lingua 130: 33–49. Clavería Nadal, Gloria. 1992. La construcción artículo+posesivo en los siglos XIV y XV.  In Actas del II Congreso Internacional de Historia de la Lengua Española, vol. 1, 347–359. Madrid: Pabellón de España. Company, Concepción. 1991. La extensión del artículo en el español medieval. Romance Philology 44 (4): 402–424. ———. 2006. Artículo + posesivo + sustantivo y estructuras afines. In Sintaxis histórica de la lengua española, vol. 2, tomo 1, segunda parte: la frase nominal, ed. Concepción Company, 759–880. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. Contreras, Heles. 1986. Spanish bare NPs and the ECP. In Generative studies in Spanish syntax, ed. Ivonne Bordelois, Heles Contreras, and Karen Zagona, 25–49. Dordrecht/Riverton: Foris. Cornilescu, Alexandra. 1992. Remarks on the determiner system of Rumanian: The demonstratives AL and CEL. Probus 4: 189–260. de Haan, Germen J., and Fred P. Weerman. 1986. Finiteness and verb fronting in Frisian. In Verb second phenomena in Germanic languages, ed. Hubert Haider and Martin Prinzhorn, 77–110. Dordrecht: Foris. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Eberenz, Rolf. 2000. El español en el otoño de la Edad Media. Sobre el artículo y los pronombres. Madrid: Gredos. Elsig, Martin. 2017. New insights into an old form: A variationist analysis of the pleonastic possessive in Guatemalan Spanish. Language Variation and Change 29 (2): 157–186. Giorgi, Alessandra, and Giuseppe Longobardi. 1991. The syntax of noun phrases. Cambridge: Cambridge University Press. Ishikawa, Masataka. 1995. On categorial evolution: A case study in Spanish possessives. In Historical linguistics 1993: Selected papers from the 11th international conference on historical linguistics, ed. Henning Andersen, 205–216. Amsterdam: John Benjamins. ———. 1997. A note on reference and definite articles in Old Spanish. Word 48 (1): 61–68.

  DP Structure: From Multiple Determiners to Just One 

161

Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1. Kayne, Richard S. 2008. Some preliminary comparative remarks on French and Italian definite articles. In Foundational issues in linguistic theory, ed. Robert Freidin et al., 291–322. Cambridge, MA: MIT Press. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Lapesa, Rafael. 2000. Estudios de morfosintaxis histórica del español. Madrid: Gredos. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Plank, Frans. 2003. Double articulation. In Noun phrase structure in the languages of Europe, ed. Frans Plank, 337–395. Berlin: Mouton de Gruyter. Pountain, Christopher J. 2001. A history of the Spanish language through texts. London: Routledge. Roberts, Ian, and Anna Roussou. 2003. Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rosemeyer, Malte, and Andrés Enrique Arias. 2016. A match made in heaven: Using parallel corpora and multinomial logistic regression to analyze the expression of possession in Old Spanish. Language Variation and Change 28 (3): 307–334. Stark, Elisabeth. 2008. Typological correlations in nominal determination in Romance. In Essays on nominal determination: From morphology to discourse management, ed. Henrik Høeg Müller and Alex Klinge, 45–61. Amsterdam: John Benjamins. Suñer, Margarita. 1982. The syntax and semantics of Spanish presentational sentence-­types. Washington, DC: Georgetown University Press. Tagliamonte, Sali A. 2012. Variationist sociolinguistics: Change, observation, interpretation. Oxford: Wiley-Blackwell. Vergnaud, Jean Roger, and Maria Luisa Zubizarreta. 1992. The definite determiner and constructions in French and in English. Linguistic Inquiry 23: 595–652. Zamparelli, Roberto. 2002. Definite and bare kind-denoting noun phrases. In Romance languages and linguistic theory 2000: Selected papers from ‘Going Romance’ 2000, ed. Claire Beyssade et  al., 305–343. Amsterdam: John Benjamins.

5 The wh-System: Free Relatives, Double Articulation and Free Choice

5.1 Introduction Many languages contain a set of morphologically distinctive elements that are used in interrogative and relative clauses, and often also in exclamatives. Examples from within the Romance family are French qui, Italian quale and Spanish quien. While no single morphological description suffices to characterize such elements cross-linguistically, they are usually referred to in discussions on syntax as wh-words or wh-­expressions, in view of the fact that the relevant English items who, which, what etc. begin with the letters wh-. The term ‘wh-system’ refers to the inventory of such expressions in a given language, together with the syntactic rules governing their use. The focus of the present chapter, as the title indicates, is the wh-system of Old Spanish. However, rather than looking exhaustively at the individual lexical items involved in wh-constructions, the focus is on the general character of the system and the principal changes it underwent over the course of the timeframe covered by this book. At first glance the system at the beginning of this period looks rather different from its modern counterpart, the relativizing item qual in © The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_5

163

164 

I. E. Mackenzie

­ articular being involved in structures that seem unfamiliar when viewed p from a modern standpoint. On further inspection, however, the wh-­ related syntax of Old Spanish turns out to governed by essentially the same constraints as those which apply in the modern language. Indeed, in terms of the system overall, the major changes reduce to (i) the creation of the compound series of relative pronouns el que, el cual etc. and (ii) the creation of a series of so-called free choice items (FCIs) based on the suffix -quier(a). The first of these changes can be linked to the availability in medieval Spanish of DP recursion, an operation consisting in the iteration of D-elements within DP (see Sect. 4.3.2.2). From that perspective, the emergence of these relative pronouns is a by-product of the very same syntax which delivered compound possessive structures (un su hermano ‘a bother of his’, el mi amigo ‘my friend’ etc.) as well as double articulation constructions such as la cuerda la mas luenga ‘the longest chord’. As regards the emergence of FCIs like cualquier(a), quienquiera etc., this represents an important qualitative shift in the evolving grammar. To begin with, Old Spanish lacked genuine FCIs, deploying in their place free or ‘headless’ relative structures such as qual marido escogiese ‘whichever husband she would choose’ or qui tanxiere omne muerto ‘whoever touches a corpse’. However, the possibility for the complementizer que to be used in such structures led to the wh-element, augmented by the particle quier, being reanalysed as a quantifier of the free choice type. The development of the free choice series thus eliminated from the language an important class of free relatives, whose abundance in the early manuscripts is an important feature of the medieval landscape.

5.2 Overview of the wh-System 5.2.1 The Basic Framework A characteristic feature of wh-constructions such as relative and interrogative clauses in languages like Spanish is that the relevant wh-pronoun or wh-phrase occurs at the beginning of the clause, to the left of the finite

 The wh-System: Free Relatives, Double Articulation… 

165

verb, analogously to the fronted topical and focal items (discourse-given objects, predicative adjectives, adverbs of manner etc.) which were discussed in Chap. 2. It will be recalled that these latter items are conventionally analysed as being moved from their base position, for example as the verb’s complement, to a position in the (left) edge of CP, the full clause (see Sect. 2.3.1.1). Exactly the same analysis is assumed to apply to wh-structures, as is shown in (1) and (2) below (the item in strikethrough font in each case indicating the initial merge site of the fronted wh-phrase): (1) ¿[CP Cuántas unidades vendieron cuántas unidades el año pasado]? ‘How many units did they sell last year?’ (2) Quiero concocer a [DP los colegas [CP con quienes vamos a trabajar con quienes]]. ‘I want to meet the colleagues with whom we are going to work.’

The CP units highlighted in the above examples illustrate a direct question and a relative clause respectively. The syntax of direct questions is fairly transparent. As regards relative clauses, these come in several varieties, the type shown in (2) being arguably the most familiar one. The hallmark of this type of relative clause is that the relative pronoun has an antecedent, in this case los colegas, which is external to the relative clause. Together, the antecedent and the relative clause constitute a DP (see Sect. 4.2) that functions as an argument within the structure of the matrix clause. In the case of (2), the DP in question is los colegas con quienes vamos a trabajar, which is the object of the verb conocer. Due to the presence of the antecedent, this type of relative clause is commonly referred to as a headed relative clause, although syntactically speaking the antecedent is not actually a head at all. Despite this terminological infelicity, I will use this label to refer to this type of structure. The embedded CP in (2) can be further characterized as a restrictive relative clause, in the sense that it restricts or defines the reference of the antecedent: the latter refers not to all colleagues, but just those with whom the speaker is going to work. Example (3) below illustrates an appositive or non-restrictive relative clause; that is, one which does not

166 

I. E. Mackenzie

limit the reference of the antecedent but, rather, serves as a parenthetical comment on it and, as a consequence, is often separated from it by an intonational break (or a comma in written language). (3) Ayer concocí a Pedro y María, [CP con quienes vamos a trabajar con quienes]. ‘Yesterday I met Pedro and María, with whom we will be working.’

5.2.2 F ree Relatives in Modern Spanish and Old Spanish A further category of relative clause, which is particularly relevant to Spanish, is that of free relatives. Unlike headed relatives, these lack an overt antecedent and instead have referential import themselves, occurring often as the subject or object of their matrix clause. An example is given by the bracketed unit in (4) below, which functions as the direct object of the gerundial verb form recordando. As can be seen, there is no item in the matrix clause that could be said to function as the antecedent of the relative pronoun quienes. (4) El papa se despidió de España recordando a [quienes no tienen trabajo]. ‘The Pope marked his departure from Spain by recalling those who do not have a job.’

An important point to note about (4) is that the accusative preposition a which immediately precedes quienes belongs to the syntax of the matrix clause, rather than to that of the relative clause. That is to say, the preposition in question does not construe with quienes, which is actually a subject, but with the free relative as a whole; the latter unit, in this arrangement, being structurally equivalent to a regular accusative DP such as los desempleados ‘the unemployed’ in the sequence recordando a los desempleados ‘remembering the unemployed.’ The constructional relations manifested in (4) contrast with those which arise in (5) below.

 The wh-System: Free Relatives, Double Articulation… 

167

(5) Queremos tener la misma relación con el nuevo presidente, que es [a quien han elegido los ciudadanos]. ‘We want to have the same relationship with the new president, who is the person that the people have chosen.’

Here the accusative preposition a cannot be associated with the matrix verb, given that the latter is intransitive. Instead, the presence of this preposition is a by-product of the fact that quien is the object of elegido inside the free relative. By the same token, the preposition belongs to the internal syntax of the free relative and not to the syntax of the matrix clause. That the preposition is located at the beginning of the free relative in (5) follows from the fact that Spanish, unlike English, disallows preposition stranding. Accordingly, when quien undergoes wh-movement from object position to the left edge of the relative clause, the preposition is “pied-piped” with it. Pied-piping of the preposition (if one is present) is a routine phenomenon in regular headed relative clauses (i.e. those that have an overt antecedent) and also in questions, both direct and indirect. However, it may look superficially perplexing in the context of free relatives, owing to the kinds of syntactic frame in which these structures can occur. It should also be noted that the pattern illustrated by (5) applies generally in Spanish, and not just in the case where the preposition is the special accusative-marking ‘personal a’, which some analysts might treat as not actually being a preposition (cf. Rivero 1984: 96–97; Jaeggli 1980). This can be seen from (6) below, where con is unambiguously a preposition. (6) Las organizaciones ciudadanas son [con quienes tenemos el mayor contacto]. ‘Citizens’ organizations are who we have our greatest contact with.’

An interesting consequence of the prohibition on stranding in Spanish is that it entails that sequences of two adjacent prepositions will arise if the kind of pattern illustrated by (4) coincides with the kind illustrated by (5) and (6); that is, if prepositions are simultaneously required by the grammatical role of the free relative within the matrix clause and by the

168 

I. E. Mackenzie

grammatical role of the wh-pronoun within the free relative. This scenario is illustrated in (7) below, where, at the level of the matrix clause, de expresses the possessive relationship between the referent of the bracketed free relative and the noun piel while, within the free relative, a indicates that quien is the indirect object of dirigimos. (7) Pongámonos en la piel de [a quien dirigimos el plan de negocio]. ‘Let’s put ourselves in the skin of the person to whom we are addressing the business plan.’

As (8) below indicates, exactly the same phenomenon can arise with indirect questions, which cross-linguistically tend to overlap with free relatives (cf. van Riemsdijk 2006: 356–357). (8) Depende de [con quién hables]. ‘It depends on who you talk to.’

The difference between the above type of example and a structure involving a free relative resides in the interpretation rather than in the syntax. That is to say, while free relatives typically refer to individuals (which may or may not be specific), the bracketed structure in (8) has a propositional type of meaning: the contingency referred to by the matrix subject it does not depend directly on a person but on who this person is. Although Spanish allows the occurrence of two adjacent prepositions in cases like (7), there appears to be a surface filter which blocks a sequence of two identical prepositions.1 Alarcos Llorach (1994: 336) provides a nice example of the phenomenon, which is shown below as (9). (9) Obsequió con diatribas feroces a quienes él consideraba intelectuales indiferentes. (Francisco Ayala, Recuerdos y Olvidos, Madrid 1988) ‘He treated those he considered to be intellectually mediocre to ferocious diatribes.’

Here, the accusative preposition a is required by both the free relative as whole, given that it occurs as the object of transitive obsequió, and by the wh-pronoun quienes, which is the object of consideraba within the free relative. However, rather than two adjacent instances of a, which would

 The wh-System: Free Relatives, Double Articulation… 

169

be analogous to what can be observed in (7), only one occurrence of the preposition is actually spelled-out phonologically. Note, in addition, that the same filter appears to be operative in structures involving indirect questions. Thus (10a) is ungrammatical and its content would either be expressed using a different structure or one instance of the relevant preposition would be deleted, as in (10b).

(10a) *Depende de [de quién estés hablando]. ‘It depends on who you are talking about.’



(10b) Depende de quién estés hablando.

Examples like (9), in which a preposition required by the internal syntax of the free relative does not conflict with the syntax of the matrix context, are instances of what have been called matching free relatives. Conversely, examples such as (5), (6) and (7), where the preposition required by the internal syntax does not mesh naturally with the matrix syntax, involve non-matching free relatives. The latter are fact rather useful as diagnostics for understanding the structure of free relatives. In particular, they reveal quite transparently that a fronted wh-phrase in a free relative does not participate directly in the syntax of the matrix clause. That is to say, a quien cannot be analysed as the complement of either the preposition de in (7) or of the copular verb form es in (5); nor does it make sense to regard con quienes as the complement of son in (6): in all three cases, the containing matrix syntax requires a prepositionless constituent. This can only be the free relative as a whole, within which the fronted wh-phrase is insulated from the matrix context. This finding suggests that neither the Head Hypothesis of Bresnan and Grimshaw (1978) nor the Shared Structure Hypothesis discussed by van Riemsdijk (2006) are appropriate for Spanish free relatives. The first-mentioned approach treats the fronted wh-phrase as being in effect the antecedent of the relative structure while the second approach assumes that the wh-phrase is “shared” between the containing matrix structure and the embedded relative structure. Under both analyses, the fronted wh-phrase must satisfy whatever syntactic requirements are imposed by the preceding material in the matrix clause. As regards (5), (6) and (7), this means that the

170 

I. E. Mackenzie

wh-­phrase cannot be a PP, given that neither the copula (when used as an identity verb) nor prepositions select PP complements. Self-evidently, neither a quien nor con quienes meets this requirement. In light of these points, and adapting an idea from Fassi Fehri (1980), I assume that free relatives can be analysed as a special, nominal instance of CP, an approach which preserves the structural parallel with regular headed relative clauses, as discussed in Sect. 5.2.1. From this perspective, the relevant aspects of the structure of (5) and (6), for example, can be analysed as in (11) and (12) below.

(11) Las organizaciones ciudadanas son [CP con quienes tenemos el mayor contacto con quienes].



(12) Pongámonos en la piel [PP de [CP a quien dirigimos el plan de negocio a quien]].

According to this analysis, what construes with the matrix syntax is not the wh-phrase but the bracketed CP, which has a nominal or referential interpretation, rather than the usual propositional one.2 Within each of these embedded CPs, the wh-PP (con quienes or a quien) has undergone routine Ā-movement from its base position to the CP edge (see Sect. 2.3.1). For the matching type of case illustrated by (9), a rule of preposition deletion can be assumed, as in (13), where the a of fronted a quienes receives a phonologically null spellout.

(13) Obsequió con diatribas feroces a [CP a quienes él consideraba a quienes intelectuales indiferentes].

The foregoing CP-based analysis of free relatives is applicable to both modern Spanish and Old Spanish, given that the same prohibition on preposition stranding existed in the medieval language and gave rise to the same effects as those illustrated above. The thirteenth century examples (14) to (17) below contain non-matching free relatives, which are indicated by brackets. Within each of these free relatives, the wh-PP (shown in bold) has been fronted from a postverbal position, analogously to a quien and con quienes in the modern examples considered earlier (although note that qual in (14) and (17) occurs as a determiner rather than a pronoun).

 The wh-System: Free Relatives, Double Articulation… 

171



(14) E [CP en qual mienbro fuere el infortunador]; enel iudga la feedat & el dannamiento. (Judizios de las estrellas, fol. 68r) ‘And whichever limb the malefic is in, judge in it the ugliness and the imperfection.’



(15) Ca tengo yo que el Rey que en paz esta. si a [CP con qui contienda] que bien le este; quanto fuelga. tanto pierde de so tiempo. (General estoria IV, fol. 198r) ‘For I believe that the King who is at peace, if he has someone with whom he can fight, this is good for him. As he idles, so he wastes his time.’



(16) Mas estas costellationes de que en este libro fablamos. & son [CP de lo que obrauan las yentes que nombramos antes destas]; son mucho apoderadas constellationes. (Libro de las cruzes, fol. 5r) ‘But these constellations about which we speak in this book, and (which) are what the people mentioned earlier worked with, are very powerful constellations.’



(17) que por [CP de quales quier malas costumbres que ell omne sea]. que si dellas se parte & se echa a buenas costumbres & dellas usa & en ellas acaba. que derecho es de seer puesto en el çielo.3 (General estoria I, fol. 272v) ‘. . . that whatever bad habits a man might be prone to, if he departs from them and adopts good habits and partakes of them and does not abandon them, then he is eligible to be put in heaven.’

As regards matching free relatives, preposition deletion applies exactly as in modern Spanish. This is illustrated in (18) below where both casassen ‘get married’ and se pagassen ‘were satisfied’ can be analysed as selecting for a PP complement introduced by con. Deletion of one instance of the latter preposition (indicated below by strikethrough font) follows according to the filter mentioned earlier.

(18) et mando luego dar pregon por la uilla. quelas mugieres que dell auenimiento de pharaon & de Moysen fincaran bibdas & todas las otras que maridos ouiessen mester. que casassen con [CP con quien se pagassen]. (General estoria I, fol. 189r)

172 

I. E. Mackenzie

‘And he later ordered that it should be proclaimed in the town that women who had been widowed by the events concerning the Pharaoh and Moses and all the others who needed a husband, should marry whoever they were happy with.’

Additional confirmation of the structural parallel in this area of the grammar between the medieval and modern varieties of the language comes from the fact that, exactly as in modern Spanish, indirect questions in thirteenth century Spanish pattern with free relatives. The bracketed CPs in (19) to (21) can be thought of as non-matching indirect questions (but note that (21) also exhibits structurally redundant “doubling” of the preposition por), while (22) has a matching indirect question, with concomitant preposition deletion.

(19) & por que podrie seer dubda [CP a quales dellos deuen demandar el diezmo al que uende o al que compra]. touo por bien santa eglesia de lo mostrar. (Libro de las leyes, fol. 112r) ‘And because there could be doubt as to which of these they should demand the tithe from, the seller or the buyer, the Holy Church decide to clarify this.’



(20) E otrossi pleyto de las eglesias. [CP de qual obispado o arcidianado deuen seer]. & de los obispados [CP a qual prouincia pertenescen]. (Libro de las leyes, fol. 52v) ‘And also disputes concerning which bishopric or archdeaconry churches should be under and which ecclesiastical province bishoprics belong to.



(21) ca non auie y ninguno que aso padre connosciesse. nin [CP por de quien se touiesse por heredero. poro ouiesse conseio]. (General estoria IV, fol. 179r) ‘For none of them knew their father or who they should consider themselves to be the descendant of, from whom they could obtain advice.’



(22) mas aponiendo y otras achaques por [CP por qual razon lo fazie]. fizo sus cortes & dixo delos ebreos contra los egipcianos como agora oyredes aqui. (General estoria I, fol. 132r)

 The wh-System: Free Relatives, Double Articulation… 

173

‘Instead, advancing other excuses as the reason why he was doing it, he convened his court and spoke of the Hebrews being against the Egyptians, as you will now hear.’

5.3 Changes in the wh-System Since the High Middle Ages The majority of the changes that have occurred in the wh-system are lexically specific, obvious examples being the disappearance of the pronoun qui or the loss of the predicative and adverbial uses of qual.4 In terms of the system overall, the principal developments over the course of later Middle Ages were the emergence of the compound relative pronouns el que and el qual, together with the reanalysis of certain free relative clauses as quantified DPs, specifically DPs in which the quantifying expression is a free choice item or FCI (the prototypical example of an FCI is English any in examples like Any primer will do). Each of these developments is examined in turn below, within an analytical framework that attempts to relate the relevant changes to more general aspects of the medieval syntax, rather than focusing on the structures in question as discrete units having their own specific history (for the latter type of approach, see in particular Girón Alconchel 2006; Company 2009).

5.3.1 Prefixation of the Definite Article to qual 5.3.1.1  Medieval qual as a Determiner and a Pronoun An emblematic feature of the medieval syntax is the alternation between the prenominal and the postnominal placement of qual. When prenominal, this item can be analysed as a determiner, given that ‘qual + NP’ had the distribution of a DP, being free to occur in both complement and subject positions. For example, within their respective CP units, qual entendimiento in (23) is the object of ouiere while quales aues in (24) is the subject of fueren.

174 

I. E. Mackenzie



(23) E daquestos pueden tirar daqui adelante quantos quisieren segunt [CP qual entendimiento ouiere aquel quilo quisiere fazer]. (Libro de ajedrez, dados y tablas, fol. 78v) ‘And any of these can now be tried out, depending on the understanding the person wanting to do this has.’



(24) E sea el echar de guisa que se arriedre el echador dellas. e que les mueuan las palomas trayendolas aderredor. o [CP quales aues fueren aquellas con que las sennalaren]. (Libro de las animalias, fol. 58v) ‘And in letting them [the birds of prey] go, the falconer should move away from them, and the doves should be dispersed around the place where such birds were the ones with which they were trained.’

As regards postnominal qual, García García (1992: 449) proposes that this item functions analogously to its prenominal counterpart, in the sense that in both cases the noun is modified ‘qualitatively’ by qual. However, an analysis along those lines is called into question by examples such as (25) below, where the qual-phrase is the preverbal subject of the embedded verb fueron.

(25) & ueyendo lo todos fare sennales quales nunqua fueron fechas njn uistas sobre tierra. (General estoria I, fol. 215v) ‘And in front of them all I will make signs which have never been done or seen on earth.’

An analysis in which the wh-form quales in this example is simply a postnominal modifier implies that the subject of the embedded relative is sennales quales. However, the latter sequence, if analysed as a subject DP, would have the structure shown in (26) below, in which D is phonologically null.

(26) [DP [D e] [NP sennales quales]]

As is discussed in Sects. 4.2 and 4.3.1, null D is subject to what Longobardi (1994) calls a “lexical government requirement”, entailing that structures like (26) cannot occur in the preverbal subject position,

 The wh-System: Free Relatives, Double Articulation… 

175

which is “lexically ungoverned” in Longobardi’s terms. On the other hand, if sennales is taken separately from quales and analysed as the object of fare, it does not infringe any constraint, given that null D is assumed to be generally licensed in complement position with plural and mass nouns (see Sect. 4.2). Moreover, quales, taken as a relative pronoun rather than a relative adjective, can freely be analysed as the embedded subject, given that pronouns do not require an overt D-element (Longobardi 1994: 659–660). These considerations suggest that (25) has the structure shown in (27) below, where the bracketed CP is a headed relative clause of the familiar type.

(27) & ueyendo lo todos fare sennales [CP quales nunqua fueron fechas njn uistas sobre tierra].

Here sennales is a regular antecedent, external to the relative clause, while the subject of the embedded verb fueron is quales, occurring as a pronoun. The structural role of sennales in (25)/(27) thus contrasts rather sharply with that of aues in (24), which is not an antecedent but, rather, part of the embedded subject quales aues. The foregoing analysis of postnominal qual as a pronoun can be assumed to apply generally. Example (28) below illustrates the case in which qual has been extracted from object position, while (29) illustrates extraction from the post-prepositional context (with concomitant pied-­ piping of the preposition). In both examples, strikethrough font is used to indicate the initial merge site of the item which has undergone wh-­ movement to the edge of the relative CP.

(28) Empos esto si los Romanos guardassen la fe de la postura de las pazes [CP qual ellos querien que los sos conqueridos guardassen qual a ellos] [. . .] non serien los Romanos o seruirien al sennorio de los Sampmithes que sennorearien. (General estoria IV, fol. 242r) ‘After this, if the Romans kept the faith with the peace agreement which they wanted their vanquished foes to keep in respect of themselves, they would not be the Romans or they would be serving at the pleasure of the Samnites, who would have the upper hand.’

176 



I. E. Mackenzie

(29) mas por tal que en los nuestros dias non auenga el tiempo [CP de qual dixo el apostolo san paulo de qual. tiempo sera que los omnes non querran buena doctrina mas querran beuir segundo sus ueluntades]. (Fuero Juzgo, fol. 89r) ‘However, so that in our days there does not come the time about which Saint Paul the Apostle said, “there will come a time when men spurn good doctrine and seek instead to live according to their desires.”’

It is worth noting in passing that pronominal qual was quite often doubled by a resumptive clitic, as in (30) below, where the -l of quil is an apocopated form of the third-person object pronoun le.

(30) & diz que este linnage destas serpientes an un entendimiento & una cobdicia [CP qual non a otra serpiente dotra natura [CP quil aya]]. (General estoria IV, fol. 18v) ‘And he says that this type of snake has an understanding and a tenacity which there is no other type of snake that has it.’

In this and similar examples, where one relative clause is contained within another, the use of the resumptive pronoun is analogous to that found in marginal English examples such as (31):

(31) ?It’s a book [CP that I don’t know anyone [CP who has read it]].

In both (30) and (31) the use of a resumptive pronoun follows from the fact that relative clauses are islands for wh-extraction, meaning that a regular wh-movement configuration, with zero realization of the moved wh-expression at the extraction site is not possible, as is shown by the ungrammaticality of (32):

(32) *It’s a book [CP that I don’t know anyone [CP who has read __ ]].

Adopting the approach of Pesetsky (1998), the resumptive pronoun located at the extraction site in each of (30) and (31) can be seen as remedying the problem illustrated in (32), by providing a minimal spellout of the fronted wh-pronoun. The fact that medieval Spanish qual could be

 The wh-System: Free Relatives, Double Articulation… 

177

involved in structures of this type can be regarded as additional, indirect evidence for the pronominal status of this element in its post-nominal occurrence.

5.3.1.2  El qual and ‘el qual + NP’ as Recursive DPs The previous section has highlighted the fact that medieval qual occurred either as a determiner or as a pronoun, the first use correlating with prenominal position and the second with post-nominal position. However, despite this difference in grammatical classification, both ‘qual  +  NP’ and pronominal qual should be analysed as being DPs headed by qual. This approach is transparent for ‘qual + NP’ and, as regards pronominal qual, it follows from Longobardi’s assumption (1994: 637) that pronouns are ‘base-generated in D’. Thus what distinguishes the two cases is the fact that qual takes an overt complement when it occurs as a determiner but has a null complement when it occurs as a pronoun, as shown in (33) and (34) below (‘e’ in the latter indicating a phonologically empty position):

(33) [DP [D qual] [NP entendimiento]]



(34) [DP [D qual] [NP e]]

Now the prefixation of the definite article to qual affected both of the above structures, as (35) and (36) illustrate:

(35) Sepas que el Sol entra en los compeçamentos de los signos segund pareçe por las tablas en .xxiij. dias de los meses latinos. & esto es su entramento en compeçamento de las formas. Por las quales formas los signos recyben sus nombres. (Libro de las cruzes, fol. 79r) ‘Know that the sun is involved in the formation of the signs in the way stated in the tables for twenty-three days of the Latin months, and this is also its involvement in the creation of the forms, by which forms the signs acquire their names.’

178 



I. E. Mackenzie

(36) & aun sobresto aya aquella pena la qual dioron los sacerdotes en su decreto el tercer anno de nuestro Regno por tal pecado. (Fuero Juzgo, fol. 35r) ‘And in addition he must receive that punishment which the priests stipulated for this sin in their decree in the third year of our reign.’

Given that both ‘qual + NP’ and pronominal qual are DPs in the own right, prefixation of the definite article as in the above examples implies DP recursion, as shown in (37) and (38) below.

(37) [DP [D las] [DP [D quales] [NP formas]]]



(38) [DP [D la] [DP [D qual] [NP e]]]

These structures are in fact a wh-analogue to the possessive recursive DPs, la mi casa ‘my house’, un su hermano ‘a brother of his’ etc., which are discussed in Chap. 4. Indeed, it is plausible to suppose that it was the availability of DP recursion in the medieval language which enabled the structures in (35) and (36) to come into existence. That is to say, in the absence of a parametric or microparametric setting enabling the stacking of two or more adjacent D-elements in a single DP, it is difficult to see how structures like (35) and (36) could have arisen (but see Pons Rodríguez 2007; Barra Jover 2007, 2008 for the proposal that they reflect the influence of written Latin). The foregoing recursion-based analysis of the genesis of el qual and ‘el qual + NP’ may also help to explain why prefixation of the article was limited to relative qual and did not include interrogative qual. A suggestive datum in this regard is the alternation which existed in interrogative contexts between ‘qual + NP’, as in (39), and ‘definite article + NP + qual’, as in (40) and (41).

(39) & por los guardar deste yerro; touo por bien sancta eglesia de mostrar quales mugieres pudiessen con ellos morar (Libro de las leyes, fol. 45r) ‘And to keep them from this error, the holy church thought it right to clarify which women could live with them.’

 The wh-System: Free Relatives, Double Articulation… 

179



(40) & uio por todas las tierras. & mesuro los logares quales serien mas guisados & meiores pora fazer y çilleros al Rey. (General estoria I, fol. 100v) ‘And he scoured the land and pondered which places would be most appropriate and best for building grain stores for the king.’



(41) E despues daquesto quando la sacaren a caça conuiene que uea el caçador antes la caça qual es. ca mas uale que la ecchen ante a la mayor caça que a la menor; (Libro de las animalias, fol. 62v) ‘And after this, when it is taken hunting, the huntsman should first of all check what is being hunted, for it is best to set [the bird] after larger game before smaller animals.’

The NPs in (40) and (41) which intervene between the definite article and the relevant form of qual are left-dislocated constituents, in the sense that they are referentially resumed from inside the qual-clause (the locus of this resumption plausibly being qual’s null complement, shown in (34) as ‘[NP e]’). Now, as was observed in Sects. 4.3.2.2 and 4.3.2.3, the separation of two D-elements by means of a dislocated nominal was a common feature of medieval Spanish, one of its effects being to obviate a spellout rule prohibiting the linear adjacency of determiners bearing a value for the same, prominent semantic feature; paradigm instances being *el el and *un el. It is not implausible to suppose that the same spellout rule lies behind the non-occurrence of interrogative DPs of the form ‘definite article + qual (+ NP)’. In more concrete terms, if the definite article is assumed to be [+ specific] and interrogative qual [− specific], then the exclusion of an interrogative DP such as *los quales logares – as opposed to attested los logares quales – would be on a par with the exclusion of DPs like *un el toro mas fuerte que auer pudiessen (as opposed to attested un toro el mas fuerte que auer pudiessen; see example (48) in Chap. 4). This approach to the failure of interrogative qual to become prefixed by the definite article presupposes that such prefixation involved DP recursion. Accordingly, the data concerning interrogative qual may be regarded as indirect evidence in favour of the proposal advanced here regarding the prefixation of the definite article to relative qual; viz. that it was a by-product of a more general medieval phenomenon, whose most salient manifestation, arguably, was the emergence of compound possessive structures such as el mi amigo.

180 

I. E. Mackenzie

5.3.1.3  Quantitative Evolution of el qual and ‘el qual + NP’ Figure 5.1 shows the quantitative evolution, across the time window covered by the corpus, of the two structures created by the prefixation of the definite article to qual. The corresponding numerical data can be found in Table A.7 in Appendix 1. The rates of occurrence represented in the graphic are actually expressed in terms of tokens per million words rather than the usual percentages. This approach has been adopted because there is no specific structure with which ‘el qual + NP’ can be said to have been in competition, the corresponding articleless construction ‘qual + NP’ being confined almost exclusively to free relatives, from which the form with the article was itself excluded.5 The open-ended nature of the linguistic competition can be illustrated using the earlier example (35), where no assumptions can be made as to what the writer would have written had he chosen not to use the formulation por las quales formas. For example, he would have achieved the same semantic and pragmatic effect if he had begun the final sentence with a non-wh resumptive phrase such as por estas formas ‘by these forms’ or por dichas formas ‘by the said forms’. Alternatively, he might simply have used a relative clause to join the penultimate and final sentences, as in the concocted example (42) below, the acceptability of whose CP-initial sequence por que appears to be vouchsafed by analogous occurrences of the same sequence in attested examples such as (43).

(42) & esto es su entramento en compeçamento de las formas [CP por que los signos recyben sus nombres]. ‘. . . of the forms by which the signs acquire their names.’



(43) una delas maneras [CP por que las pueden guardar delas enfermedades que les acaeçen]. (Libro de las animalias, fol. 196v) ‘one of the means by which you can protect them from the diseases which affect them.’

Thus it turns out that the ‘el qual + NP’ structure does not lend itself to the usual quantitative method whereby evolving proportions from a definable competition space are assigned to competing variants. Its rate of occurrence is most effectively captured using the more traditional

 The wh-System: Free Relatives, Double Articulation… 

181

Fig. 5.1  Occurrence of ‘el  +  pronominal qual/cual’ and ‘el qual/cual  +  NP’ (1250–1609)

technique of normalizing the absolute number of tokens by the volume of the discourse. To facilitate comparison, the same approach has been adopted for ‘el + pronominal qual’. Turning to Fig. 5.1 itself, it can be seen that ‘el qual + NP’ and ‘el + pronominal qual’ have analogous diachronic trajectories, increasing in frequency towards the end of the Middle Ages but declining after the 1460 data point, which averages data for the period 1450–1469. The rate of increase up to 1460 is, in fact, remarkably similar across the two structures, while the rates of decline after that data point are broadly similar to one another. Table 5.1 shows the rates of increase and decline, expressed as exponential growth ratios, estimated using the first two columns of data in Table A.7 in Appendix 1. These exponential growth ratios are analogous to the odds ratios used in previous chapters, but they measure the rate of change in the absolute quantity of tokens (per million words) rather than the rate of change in the odds. Like odds ratios, exponential growth ratios may be greater than 1, implying that the relevant quantity is increasing, or less than 1, implying that the quantity is decreasing (see Sect. 1.2.3).

182 

I. E. Mackenzie

The growth ratios shown for 1250–1469, the period of increase, imply that tokens of the two structures increased in frequency by 25.6% and 25.1% respectively per decade, findings which point rather clearly to a common rate of growth. In a sense, this is expected, given Kroch’s theory of the Constant Rate Effect (see Sect. 1.2.1); however, it is quite striking to find that the CRE is applicable even if the evolving rates of the ­structures under scrutiny are quantified using the measure of tokens per volume of discourse as opposed to the more usual percentage measure. What this suggests is that the principle of cross-contextual equivalence in the rate of change is extremely robust and is not dependent on a logistic model of language change. As regards the decline in the two structures, which begins in the middle of the fifteenth century, it can be seen in Fig. 5.1 that it develops at a much slower rate than the increase that preceded it. This is confirmed by the “growth” ratios shown in Table 5.1 for the period 1450–1609, which indicate that the quantity of tokens of ‘el + pronominal qual’ contracted at the rate of 4.0% (= 1–0.960) per decade while tokens of ‘el qual + NP’ decreased at the rate of 1.9% (= 1–0.981) per decade. Rates of decay equivalent to the growth rates cited above, viz. 25.6% and 25.1%, would be 20.4% and 20.1% per decade respectively.6 Clearly, these latter rates are much higher than what is actually observed, i.e. the rates of 4.0% and 1.9% just mentioned. The bell curves shown in Fig. 5.1 should thus be regarded as being asymmetric, implying that the advance and subsequent retreat of the two structures constitute an accidental failed change (Postma 2010); that is, a transient linguistic state whose rise and decline represent separate events (see Sect. 1.2.3). In light of the discussion in Sect. 5.3.1.2, the genesis of the two el qual structures is plausibly regarded as a further instance of DP recursion, a syntactic process that was characteristic of the medieval language (see Chap. 4). However, this does not in itself explain why tokens of these Table 5.1  Rates of increase and decline for el/la/los/las + qual Role of qual

Growth ratio (1250–1469)

P-value

Growth ratio (1450–1609)

P-value

Pronoun Determiner

1.256 1.251

0.000 0.000

0.960 0.981

0.040 0.614

 The wh-System: Free Relatives, Double Articulation… 

183

structures underwent a sustained increase from the middle of the fourteenth century. A comparison with the recursive possessive DPs considered in Chap. 4 is instructive in this regard, given that the latter structures underwent little or no significant quantitative shift until the early decades of the fifteenth century and, even then, the movement was downward rather than upward (see Sect. 4.4.1). These data suggest that the syntax of DP recursion was fundamentally stable until the end of the Middle Ages and hence cannot be invoked as an underlying causal factor in the ­quantitative movement observed here in connection with the el qual structures. In the absence, then, of any obvious syntactic driver, it is probably safest to conclude that the late medieval quantitative growth in ‘el qual + NP’ and ‘el + pronominal qual’ had a functional or stylistic cause rather than a strictly syntactic one. A reasonable hypothesis within this general approach might be that prefixation of the definite article invested the relevant DPs with greater referentiality, optimizing them for usage in contexts requiring strong discourse cohesion (cf. Pons Rodríguez 2007). In contrast to the growth in these structures, the decay phase does appear to reflect a syntactic change, viz. the system-wide loss of DP recursion which began in the early decades of the fifteenth century (see Sect. 4.4.1). Given that the 1460 data point marks the peak values for tokens of ‘el qual + NP’ and ‘el + pronominal qual’, the loss of DP recursion appears to have begun to manifest itself a few decades later in the context of the qual structures than it did in the context of the possessive DPs discussed in Chap. 4. However, as was pointed out in Sect. 3.3.2, it is normal when a change is diffused for there to be a time lag between the various contexts affected; Kauhanen and Walkden (2017: 1.3) refer to this as ‘time separation between contexts’. It can also be noted that the rates of decline for the recursive qual structures are considerably slower than the corresponding rates for recursive possessive structures of the type el mi amigo and un mi amigo, whose odds, once the decrease set in, contracted at the rate of 17.2% and 16.8% per decade respectively (see Table 4.3). This contrast is no doubt attributable in part to the different measures used to quantify the evolving synchronic frequencies of the two types of structure, the unit for the possessives being percentages/probabilities but absolute token counts (per million words) being used for the qual constructions. In addition,

184 

I. E. Mackenzie

however, it is likely that the qual structures were subject to an analogous process to that which affected those recursive possessive DPs that were introduced by a demonstrative, such as aquella su mugier ‘his wife/that wife of his’. The key point about strings of the latter type, as was discussed in Sect. 4.4.1, is that they were capable of being reanalysed as appositive structures, thereby surviving the loss of DP recursion and continuing to be generated even in the modern period (cf. modern examples such as esta nuestra ciudad ‘this, our city’ and aquella mi propuesta ‘this, my proposal’). At the level of observable quantitative trends, the possibility of generating strings of the form ‘demonstrative + possessive + NP’ using appositive rather than recursive syntax was reflected in the fact that the decline in such strings was very weak indeed, whereas the decline in unambiguously recursive strings like el mi amigo was a robust one. Given (i) that el qual, written as el cual, survives into modern Spanish as a relative pronoun and (ii) that the articleless use of qual/cual died out by the end of the sixteenth century at the latest (Lapesa 2000: 392), it is reasonable to assume that what originally were syntactically complex strings consisting in the definite article immediately followed by pronominal qual were reanalysed as tokens of a morphologically complex relative pronoun. The latter, as a single lexical item, would have been unaffected by the loss of DP recursion and tokens of it could continue to be produced even once the recursive syntax which gave rise to the structure in the first place had been lost. An equivalent process appears to have been undergone by strings involving qual in its role of determiner, given that examples such as (44) below continued to occur until at least the end of the nineteenth century:

(44) pero no se trata ahora de puntillos del carácter, de la cual dolencia todos padecemos algo (José María de Pereda, El sabor de la tierruca, Madrid 1889) ‘but this is not a question of minor feelings of injustice, a vice from which we all suffer to an extent’

The foregoing reanalysis of ‘definite article  +  qual’ and ‘definite article + qual + NP’ provides a natural explanation for the very slow rate of decline in the frequency of these strings after the 1460 watershed.

 The wh-System: Free Relatives, Double Articulation… 

185

Analogously to what was hypothesized in respect of the demonstrative possessive DPs discussed in Sect. 4.4.1, it is plausible to suppose that any reduction in tokens of the string ‘definite article + qual’ or ‘definite article + qual + NP’ which was due to the loss of DP recursion was largely offset by the generation of superficially identical tokens in which the definite article and qual formed a morphologically complex but syntactically indivisible unit. The parallel with the demonstrative possessive DPs is further supported by the p-values gauging the statistical significance of the relevant downward trends. Table 4.3 in Chap. 4 indicated that the decline in strings of the form ‘demonstrative  +  possessive  +  NP’ was statistically insignificant, the p-value generated as part of the regression analysis being 0.429. Table 5.1 in the present chapter points in similar direction, the p-value for the decline in ‘definite article + qual + NP’ being 0.614 and the corresponding figure for ‘definite article  +  pronominal qual’ being 0.040. The first figure is well outside the commonly agreed limit for statistical significance of 0.05 (or 5%), while the second is only just inside it.

5.3.2 Prefixation of the Definite Article to que In Sect. 4.3.2.2 it was observed that Old Spanish permitted double articulation structures, in which the definite article heading a DP is doubled or repeated. This phenomenon is illustrated in the example below (= (41) from Chap. 4):

(45) et parte lo que se ayuntare desso. sobre la cuerda la mas luenga. (Cánones de Albateni, fol. 17v) ‘And divide the output of this operation by the [length of the] longest chord.’

Examples such as this one are particularly striking because they are immediately reminiscent of modern French superlative formulations such as la ligne la plus longue ‘the longest line’. However, double articulation was arguably at its most productive when the modifying phrase that followed the second occurrence of the duplicated article was a relative clause rather

186 

I. E. Mackenzie

than the comparative form of an adjective. The bracketed DPs in (46) and (47) illustrate this pattern:

(46) Et cata all andamio que passa por la sennal do se taia con [DP el circulario el que es su longura dell yguador del dia]; tanto commo la ladeza de tu uilla en la parte de su ladeza. (Libros del saber de astronomía, fol. 89r) ‘And note where the meridian which passes through the mark intersects with the great circle that is described by the equator, as well as the latitude of your town within in its hemisphere.’



(47) Et la prueua desto que auemos dicho es que tornemos [DP la figura la que es enel capitulo que fue ante deste]. & que sea essa misma. (Libro del cuadrante señero, fol. 137v) ‘And the proof of what we have just said is that we rotate the figure which is in the previous chapter, and it is exactly the same.’

While Girón Alconchel (2012: 24) suggests that the sequence ‘el/la/los/las + que’ in this type of examples should be analysed as already being a compound relative pronoun, it is more plausible to assume a syntactically complex sequence involving the definite article followed by a relative pronoun, given that when the embedded verb required a preposition, this was invariably inserted between el/la/los/las and que, as in (48) below:

(48) Et pon la segunda pierna faz al centro sobre [DP la linna ascondida la sobre que cayo la otra pierna]. (Libros del saber de astronomía, fol. 164v) ‘And put the second leg facing the centre on the hidden line on which the other leg fell.’

In this regard, the medieval word order contrasts sharply with what is observed in modern Spanish, where the bracketed DP of (48) would have to be reordered as in (49) below, reflecting the fact that la que, a form of the morphologically composite pronoun el que, is syntactically indivisible.

(49) [DP la línea escondida [CP sobre la que cayó la otra pierna]]

 The wh-System: Free Relatives, Double Articulation… 

187

It is also important to recall that the bracketed DPs in (46) and (47) are paralleled by DPs such as la cuerda la mas luenga in (45), which lacks a que-clause altogether. This fact suggests that the presence of the que-­ clause was, from a broader syntactic perspective, incidental to the structure in question, corroborating the view that the linear sequences el que, la que etc. did not constitute discrete units at the time of these examples. Assuming, therefore, that the el/la/los/las element in the medieval ‘el/la/los/las + que’ sequence represented an occurrence of the definite article, and taking into account the analysis of medieval Spanish double articulation motivated in Sect. 4.3.2.2, the DPs highlighted in (46) and (47) can be analysed as in (50) and (51) below, where ‘e’ designates the (phonologically null) complement of the second article. This null complement, corresponding to what Girón Alconchel (2006, 2012) terms the ‘núcleo vacío’, is the antecedent of the relative clause.

(50) [DP el [NP circulario [DP el [NP e [CP que es su longura dell yguador del dia]]]]]



(51) [DP la [NP figura [DP la [NP e [CP que es enel capitulo que fue ante deste]]]]]

Now structures such as these, in which a nominal intervenes between two instances of a repeated definite article, are analogous to the recursive type of CP illustrated in (52) below (= (40) from Chap. 4), where the two occurrences of the duplicated complementizer que are also separated by a nominal, viz. the left-dislocated “subject” dios:

(52) & que asman otrossi [CP que dios [CP que non a cuedado delas cosas deste mundo.]] (General estoria IV, fol. 72r) ‘And who also claim that, God, he does not take care of things in this world.’

Accordingly, just as (52) involves CP recursion, so the structures in (50) and (51) can be analysed as involving DP recursion. The occurrence of examples like (46) and (47) can therefore be regarded as being a further

188 

I. E. Mackenzie

manifestation of the wider availability in medieval Spanish of DP recursion. Recursive DPs in which both of the iterated D-elements are instances of the definite article, as in the foregoing examples, are characteristic of the thirteenth and fourteenth centuries. Later on, however, one also finds examples in which the first D-element is an indefinite item such as a weak quantifier or, more commonly, the phonologically null D associated primarily with mass nouns and bare plurals, but also found occasionally with singular count nouns (particularly if they form a quasi-fixed locution with their verb: cf. Longobardi 1994: 613). These possibilities are illustrated in (53) to (55) below, where the internal structure of the bracketed DPs should be assumed to analogous to that shown in (50) and (51), modulo the difference in the initial D-element.

(53) & por este libro el adeujnaua [DP muchas cosas las que aujan de venjr] (Crónica de 1344 I, fol. 6v) ‘And with this book he predicted many things that were to happen.’



(54) sea todo feruentado en olio de coste: & añadanle [DP cera la que cumpliere] & sea fecho vnguento. (Lilio de medicina, fol. 166r) ‘And it should all be boiled in costus oil, then add to it a sufficient quantity of wax and make an ointment of it.’



(55) E otro dia de mañana vinieron los mensageros de yrlanda por la respuesta de su embaxada. & dixo el rey a oliueros que les diesse [DP respuesta la que mejor le paresciesse]. (Oliveros de Castilla, fol. 27r) ‘Later one morning the messengers from Ireland came for his ambassador’s reply. And the king told Oliveros to give them whatever reply he thought best.’

Figure 5.2 suggests that the foregoing pattern, i.e. that exemplified not just in (50) and (51) but also in (53) to (55), underwent a steady decline until the latter stages of the fifteenth century. Indeed, the exponential growth ratio for the period 1250–1469, shown in Table 5.2, equates to a decay rate of 5.7% per decade and the associated p-value of 0.005 implies that this decrease is statistically robust.7

 The wh-System: Free Relatives, Double Articulation… 

189

100 90 Observed values

Tokens per million words

80 70

Trendline to 1460

60

Trendline from 1460

50 40 30 20 10 0 1260

1290

1320

1350

1380

1410

1440

1470

1500

1530

1560

1590

Fig. 5.2  Quantitative evolution of ‘D  +  NP  +  definite article  +  relative clause’ (1250–1609)

Given that no similar, statistically reliable decline is manifested by any of the other recursive DP structures (un su hermano, las quales razones etc.) until near the end of the Middle Ages, the waning frequency of ‘determiner + NP + definite article + relative clause’ over the 1250–1469 period can be inferred to be stylistic in nature, as opposed to diagnosing a fundamental change in the syntax. A more significant datum in terms of the system overall is the apparent stabilization of the decline from the 1460 data point onwards, which is indicated by the fact that the post-1460 trendline in Fig. 5.2 is appreciably flatter than the pre-1460 one. This impression of a halt in the decline is further confirmed by the high p-value recorded in Table 5.2 in connection with the exponential growth ratio for the period 1450–1609. At 0.250, this p-value indicates that the slight downward trend detected by the regression analysis could easily be due to experimental error. Now we know from Sects. 4.4.1 and 5.3.1.3 that DP recursion in general began to be lost in the fifteenth century. Accordingly, the fact that the

190 

I. E. Mackenzie

Table 5.2  Decline and stabilization of ‘D + NP + definite article + relative clause’ (1250–1609) Period

Growth ratio (per decade)

P-value

1250–1469 1450–1609

0.943 0.978

0.005 0.250

specific recursive structure ‘determiner + NP + definite article + relative clause’ appears to stabilize from about the middle of that century is plausibly interpreted as indicating that the ‘definite article + que’ component of this construction had by then been reanalysed as a compound relative pronoun, with the NP preceding the erstwhile definite article now interpreted as the relative clause’s antecedent, as schematized in (56):

(56) Before reanalysis: [DP D [NP NP [DP def. article [NP e [CP que . . .]]]]] After reanalysis: [DP D [NP NP [CP el/la/los/las que . . .]]]

The reanalysis postulated in (56) would have enabled strings superficially identical to the medieval pattern ‘determiner + NP + definite article + relative clause’ to continue to be generated even after the loss of DP recursion. However, in their post-reanalysis guise, such strings no longer involve two D-elements in a single structure, as previously, but just one, the second D-element (i.e. the definite article) having been absorbed into the new compound relative pronoun. On the other hand, although they appear to have been quantitatively stable at the end of the Middle Ages, the strings in question have declined in the modern period, the compound relative pronoun created by the reanalysis in (56) becoming specialized as a post-prepositional element (except in non-restrictive relatives). According to Girón Alconchel (2012: 25), however, non-­ oblique occurrences are still encountered in Latin American varieties of Spanish. Like el cual, then, the compound relative pronoun el que reflects reanalysis of what originally was a recursive structure in which el (together with la, los and las) occurred as a true D-element. However, in contrast to the pattern of development exhibited by el cual, the transformation of the el que sequence into a single unit presupposes the occurrence of relatively complex, double articulation-type structures.

 The wh-System: Free Relatives, Double Articulation… 

191

In these, the presence of the que-clause was to an extent accidental, given that other modifying phrases, such as mas luenga in (45) or even a simple adjective like retornada in (57) below, were capable of discharging the same role.

(57) Et quando lo sopieres por qual carrera quier destas carreras sobredichas; sabe [DP su cuerda la retornada]. & mengua la dela cuerda retornada dell arco de mediel dia. (Cánones de Albateni, fol. 16v) ‘And when this is known by either of the above-mentioned methods, identify [the value of ] its rotated chord and subtract it from [that of ] the rotated chord of the midday arc.’

Nevertheless, both el cual and el que can be regarded as being a residue of the same, now obsolete syntax which produced formulations such as el mi amigo, un su hermano and aquella su mugier.

5.3.3 E  mergence of the Free Choice Items (FCIs) Based on -quier(a) 5.3.3.1  FCIs Versus wh-Expressions FCIs, broadly speaking, are quantifying expressions that are used analogously to English any, together with compound forms like anybody, anywhere etc. It should be noted, however, that a sharp distinction is normally drawn between ‘free choice any’ and any as a so-called polarity item (see Sect. 6.2.1), the latter use being associated with negative, conditional and interrogative contexts. The relevant analogy as regards Spanish items like cualquier(a), quienquiera, dondequiera is with free choice any and not with polarity any. FCIs like cualquier(a) differ from standard universal quantifiers such as todo ‘every/all’ in two key respects: (i) they have obligatory wide scope and (ii) they are limited in their distribution to a narrow range of contexts, most notably statements of possibility and generic assertions. The first point is illustrated by example (58) below, from Menéndez Benito (2010: 34):

192 



I. E. Mackenzie

(58) Puedes coger cualquier carta de esta baraja. ‘You can take any card from this deck.’

This sentence can be paraphrased as ‘for every card x in this deck: if x is a card, it is permitted that you take x’, the permissive component of the proposition being included in the scope of the universal operator, rather than vice versa. To include the universal operator within the scope of the permissive component – the so-called ‘narrow scope’ reading of the quantifier – a different determiner would have to be used, the obvious choice being todo, as in (59):

(59) Puedes coger todas las cartas de esta baraja. ‘You can take all the cards from this deck.’

The distributional restriction on free choice quantifiers is illustrated by the ungrammaticality of (60), again from Menéndez Benito (2010: 34):

(60) *Juan cogió cualquiera de las cartas de esta baraja. *‘Juan took any of the cards from this deck.’

The cause of the ungrammaticality here is that the pronoun cualquiera is not embedded in a possibility context like that created by puedes in (59), nor is it in a generic context, such as that illustrated in (61):

(61) El se pelea con cualquiera. ‘He quarrels with anyone.’

An additional point to note about cualquier(a), quienquiera etc., is that, like FCIs generally, they are not wh-expressions. This can be seen in an example like (62) below, which is taken from Rivero (1984: 85).

(62) Cualquiera con quien hables se convencerá. ‘Anyone to whom you talk will be convinced.’ (My translation.)

Here, cualquiera is the antecedent of the relative clause con quien hables, a syntactic role which would not be available to a relative pronoun, as the ungrammaticality of (63) demonstrates:

 The wh-System: Free Relatives, Double Articulation… 



193

(63) *Quien con quien hables se convencerá. ‘*Who(ever) to whom you talk will be convinced.’

The non-wh status of cualquier(a), quienquiera etc. is further indicated by the fact that these items cannot undergo wh-movement. For example, although (64) is perfectly grammatical, a counterpart involving wh-­ movement, as in (65), is not.

(64) La justicia tiene que ser imparcial, quienquiera que sea el acusado. ‘Justice must be impartial, whoever the accused is.’



(65) *La justicia tiene que ser imparcial, con quienquiera que esté casado el acusado. ‘Justice must be impartial, whoever the accused is married to.’

In this way it can be seen that, although the items from which they are diachronically constructed (qual, quien etc.) are wh-expressions, the -quier(a) forms themselves are not.

5.3.3.2  M  edieval Immunity to the Doubly-Filled Comp Filter Roberts and Roussou (2003: 170–171) make the observation that FCIs very often descend from free relative clauses in which the finite verb means ‘to want’. At first glance, the relevant Spanish items seem to be a good illustration of this generalization, given that the suffix -quier(a) can be etymologically linked to the verb querer ‘to want’. However it should be noted that -quier(a) does not descend directly from a finite verbal occurrence, analogous to quisieres in (66) below:

(66) El primero capitolo es pora seer omne temudo & cruo a qui quisieres. [. . .] El .xxiij. pora fazer correr qual moneda quisieres en qual uilla quisieres (Formas e imágenes, fol. 6r) ‘The first chapter is for being regarded as fearsome and cruel by anyone you want . . . The twenty-third is for circulating any currency you like in any town you like.’

194 

I. E. Mackenzie

Rather, it represents a continuation of the impersonal particle quier, which originally occurred both postnominally, as in (67), and prenominally, as in the earlier example (17), repeated below8:

(67) E quando les acaeçier qual cosa quier daquestas a alguno delos perros. tomen delas foias dela ruda e magenlas e mezclenlas con miel e con sal molida. (Libro de las animalias, fol. 200v) ‘And when any of these things happens to one of the dogs, take some rue leaves, crush them and mix them with honey and ground salt.’



(17) que por [CP de quales quier malas costumbres que ell omne sea]. que si dellas se parte & se echa a buenas costumbres & dellas usa & en ellas acaba. que derecho es de seer puesto en el çielo. (General estoria I, fol. 272v) ‘. . . that whatever bad habits a man might be prone to, if he departs from them and adopts good habits and partakes of them and does not abandon them, then he is eligible to be put in heaven.’

Importantly, the determiner or pronoun (qual, qui, que, quien etc.) which construed with the particle quier appears to still have been a wh-­ expression, in some instances at least. This is indicated by the fact that PPs containing such combinations underwent wh-movement. In (17), for example, the PP de quales quier malas costumbres can only be analysed as having been fronted from postcopular position, as is shown in (68):

(68) que por [CP de quales quier malas costumbres [C que] [TP ell omne sea de quales quier malas costumbres]] . . .

Exactly the same pattern arises in (69) to (71) below, the PPs delante qui quier, con qui quier and en que quier being the complements of fable, te tomasses and acaezca respectively:

(69) Et el omne otrossi que la touiere consigo; sera su dicho otorgado delante qui quier que fable. (Lapidario, fol. 112v) ‘And also any man who has it with him, his marriage proposal will be granted whoever he speaks to.’

 The wh-System: Free Relatives, Double Articulation… 

195



(70) ffijo alexandre o es el to uencer & o es el to fado que tu recebiste de los dioses et que siempre uençries con qui quier que te tomasses. (General estoria IV, fol. 210r) ‘Alexander my son, either by your own prowess or by fate received from the gods, you would always defeat whoever you fought against.’



(71) Et segond esto para mientes en estas figuras. & Judga por ellas en que quier que te acaezca alguna daquellas figuras. quier en alguna de las coniunctiones grandes. o en reuolution de anno. o en nascentia. o en question. (Libro de la cruzes, fol. 20v) ‘Accordingly, think about these figures, and judge by them, whatever you find the figure to be in, be it in one of the big constellations, or in the turn of the year or in birth or in question.’

Note that the occurrence of wh-movement in these examples implies that the containing structures were a form of relative clause, specifically a free relative. This in turn implies that the overt realization of the complementizer que immediately after quier in (69) to (71) and after the NP malas costumbres in (17) represents a violation of Chomsky and Lasnik’s (1977) ‘doubly-filled Comp’ filter, a constraint requiring the complementizer to be silent if an overt wh-word is also present in the same area of clause structure.9 However, such violations were not uncommon in Old Spanish, particularly with qual, as examples such as (72) and (73) illustrate.

(72) & por esta carrera podras iudgar en cadauna figura dellas quando te acaeçiere en [CP qual cosa que fuere] segond que conuiene a aquella cosa. (Libro de las cruzes, fol. 20v) ‘And by this means you will be able to interpret each of the figures, when you find it in whatever element, in accordance with the procedure corresponding to that element.’



(73) todos los omnes de aquel lugar assi los barones como las mugeres [CP de qual gente que sean]. o de qual quier linage. o de qual quier dignidat reciba cada vno .CCtos. azotes. (Fuero Juzgo, fol. 79r) ‘Every person in that place, both men and women, whatever class of person they may be, from whatever family, irrespective of their rank, must each receive 200 lashes.’

196 

I. E. Mackenzie

An analysis of quales in (17), qui in (69) and (70), and que in (71) as genuine wh-expressions should therefore not be precluded on the grounds that such an analysis implies a violation of the doubly-filled Comp filter. Indeed, it is arguable that violations of the latter filter are a predictable component of the process whereby free relatives are reanalysed as free choice structures. This can be seen from a comparison of free relatives like qual moneda quisieres and qual uilla quisieres, taken from the earlier example (66), with the structurally closest free choice formulation shown in (74):

(74) cualquier libro que quieras ‘any book you want’

As can be seen immediately, the free choice structure has an overt que, which is lacking in the free relatives. Logically, therefore, the transformation of the free relative pattern into the free choice one requires, at some point, the overt inclusion of the element que. This innovation is most simply envisaged as a process whereby a previously silent element comes to be given phonetic realization; that is to say, given that the silent element in question is the complementizer, it is most simply envisaged as a violation, in the diachronic dimension, of the doubly-filled Comp filter.

5.3.3.3  The Particle quier Given the forgoing remarks, the most appropriate parallel for the independent particle quier in the thirteenth century (and no doubt before that time) is with the English -ever formative rather than free choice any. For, while quier emphasized the unrestricted nature of the reference of its containing DP, analogously to both -ever and any, its presence did not (or at least not always) deprive the associated wh-expression of its wh-­characteristics, a property which is exhibited by -ever but not by any. The transformation of sequences consisting in ‘wh-expression + quier’ into genuine FCIs, analogous to English any-forms, must therefore represent a later, albeit overlapping development. Moreover, this secondary process does not appear to have reached completion until after the medieval period, given that examples in which a quier-phrase undergoes

 The wh-System: Free Relatives, Double Articulation… 

197

wh-­movement can be found until at least the beginning of the seventeenth century. This circumstance is illustrated by the example below, from the second part of Mateo Alemán’s picaresque novel Guzmán de Alfarache, which was first published in 1604.

(75) empero por qualquier camino que trates de vengarla, saltaste de la sarten al fuego, fuyste huyendo de vn inconueniente y diste de cabeça en muchos. (Guzmán de Alfarache II, book II, ch. III) ‘However, by whatever means you seek to avenge it, you jumped from the frying pan to the fire, you fled one problem and ran into many.’

Here, the PP por qualquier camino construes with either trates or vengarla inside the embedded verb phrase, and hence must be analysed as having undergone wh-movement, an operation which is not in principle possible if qualquier is a free choice quantifier. In the spirit of Sect. 3.3.1, where a seamless rather than staggered model of change was preferred, it is plausible to assume that the availability of wh-movement in quier-structures overlapped for some considerable time with the occurrence of genuine free choice quier-phrases, implying there was a long period of linguistic competition between an ever-type quier and an any-type one.10 However, it is in practice impossible to quantify the evolving proportion of the quier tokens represented by each variant, given that the ever-type can only be identified unambiguously in cases in which the quier-phrase is a PP originating inside the embedded verb phrase, as in (75) together with (68) to (71). Regardless of the presence of quier, free relatives meeting that particular specification are statistically infrequent. It should also be noted that the consolidation of quier-phrases as FCIs does not appear to correlate with the morphological evolution of quier from independent particle to suffix. For, while examples like (75) imply that the ever-type variant of quier existed until at least the early seventeenth century, postnominal occurrences of the form quier, as in (67), fall away steeply after the end of the thirteenth century. The example below, from the Fuero de Briviesca (composed 1313; copied 1351–1400), is one of the last in the corpus:

198 



I. E. Mackenzie

(76) mas en qual hora quier que sopiere de algun herege. que luego lo faga saber al obispo dela tierra (Fuero de Briviesca, fol. 75r) ‘But at any time he learns of some heresy, he must report it to the local bishop’

Although orthographical agglutination of quier to qual did not become mandatory until centuries later, the effective loss of the above postnominal position in the mid- to late fourteenth century can probably be taken to diagnose that the form had acquired a suffixal status by that time. To the extent that this finding implies that the consolidation of the free choice syntax followed, rather than preceded, the formal grammaticalization of forms like qualquiera, it can be regarded as being effectively the inverse of Company’s proposal (2009: 93–94), according to which the formal grammaticalization of such forms remained ‘weak’ for some time after the semantics of the structure had crystallized. More generally, the disassociation of the morphological and syntactic processes contributing to the overall development of the -quier(a) forms underlines the fact that grammaticalization is not itself a theoretical primitive, but rather a coming together of various separate phenomena.

5.4 Conclusion Although no major typological shift appears to have occurred in the wh-­ system as Old Spanish evolved into Modern Spanish, the medieval language had a substantially different appearance from its modern counterpart. Much of the dissimilarity centres around the multifaceted wh-expression qual. The determinerless variant of qual was a rich source of free relative clauses, whose regular occurrence was a characteristic feature of the Old Spanish syntax. As used in this type of context, qual was undermined diachronically in two distinct ways. Firstly, qual-based free relative structures in which the particle quier also occurred underwent reanalysis, eventually developing into free choice DPs such as cualquier tierra ‘any land’ together with the free choice pronoun cualquiera ‘any(one)’. Secondly, in its use both as a determiner and as a pronoun, qual came to be prefixed by the definite article, a development which appears to have made it more suitable for occurrence in headed relative

 The wh-System: Free Relatives, Double Articulation… 

199

clauses as opposed to the free relative structures in which bare qual had flourished. The quantitative evolution of ‘definite article + qual’ has the asymmetric up-and-down profile associated with accidental failed changes, stylistic factors no doubt driving the steep increase and reanalysis of the surface string as a single, morphologically complex relative pronoun probably contributing to the shallowness of the decline after the peak frequency was achieved. The prefixation of the article to qual, whereby a structure that already had a D-element acquired another one, can be linked to another important theme in the syntax of Old Spanish, namely the existence of DP recursion, a phenomenon which was also manifested in the availability of possessive structures of the type el mi amigo ‘my friend’ and in the occurrence of double articulation structures. Where these latter involved a relative clause, we find surface strings that are plausibly regarded as the input material for the reanalysis of ‘definite article + que’ as a syntactically simple relative pronoun, analogous to el qual/cual. The joint emergence of the morphologically compound pronouns el que and el cual represents a further major change in the wh-system, given that the medieval language is noteworthy for its reliance on the inventory of simple wh-expressions that were inherited from spoken Latin.

Notes 1. This filter can plausibly be assimilated to the one alluded to in Sect. 4.3.2, which blocks linearly adjacent iterations of identical heads, such as two instances of the definite article (*el el) or the declarative complementizer (*que que). 2. That CP should occur in postcopular position or as the complement of a preposition is unsurprising, given that CP occurs naturally in such positions in Spanish, examples like (i) and (ii) below being very normal. (i) Lo que me preocupa es [CP que el equipo no avanza]. ‘What worries me is that the team is not progressing.’ (ii) Insisto [PP en [CP que el cambio sigue siendo posible]]. ‘I insist that change is still possible.’

200 

I. E. Mackenzie

3. This particular example also exhibits a violation of the so-called doublyfilled Comp filter of Chomsky and Lasnik (1977), which requires the complementizer, here que, to be silent if there is also an overt wh-word, here quales. However, as is discussed in Sect. 5.3.3.2, such violations were not that infrequent in Old Spanish. An additional point to note about this example is that the wh-­movement undergone by de quales quier malas costumbres indicates that qual(es) in the sequence qual(es) quier was still an independent wh-expression rather than a bound morpheme, as is the same form (spelled with a c rather a q) in the modern quantifiers cualquier and cualesquier. With these latter items, wh-movement is not possible, due to the fact that ‘cual(es)quier + NP’ is not a wh-­phrase. This is demonstrated by the ungrammaticality in modern Spanish of a structure such as (i) below, in which the PP de cualquier país de origen has been fronted from post copular position, analogously to de quales quier malas costumbres in (17). (i) *Todos los productos con marca oficial de calidad europea tienen las mismas garantías de cualquier país de origen que sean. ‘All products which bear the official European trade mark have identical guarantees whatever their country of origin.’ 4. These now obsolete uses are illustrated in (i) and (ii) below respectively: (i) E otrossi las bronchas que traye en los pechos & las sortijas de las manos & las çapatas que todo cayo aluen & finco ella desnuya qual nascio. (Estoria de España I, fol. 35v) ‘And even the jewels she wore around her neck and the rings on her fingers and her boots, everything fell away and she was left naked like the day she was born.’ (ii) & diz que se paro estonces a Semele la cara como aquellas estrannezas quando las uio. qual se para alas uezes el aer con los Nublos & los uientos & las lluuias & los truenos & los relampagos & los Rayos. (General estoria II, fol. 116v) ‘And he says that then Semele’s face became like those strange things when she saw them, just as the air is transformed sometimes by storm clouds, winds, rain, thunder and lightning.’ 5. Non-restrictive occurrences of ‘qual + NP’, analogous to las quales formas in (35), are not impossible to find, as (i) below demonstrates. However, they are so infrequent as to not really alter the picture of complementary distribution implied in the text.

 The wh-System: Free Relatives, Double Articulation… 

201

(i) Et son estos molinos conombrados sobre la presa del molino que dio el rey a don Pedro Perez notario de la reyna donna Johanna a qual molino dizien en tiempo de moros Abennaroça. (Letter from Alfonso X a Garci Martinez, 25 December 1253; Archivo de la Catedral de Sevilla, Leg. 63, Núm. 13. 37-1–14-70) ‘And the aforementioned mills are above the dam belonging to the mill which the king gave to don Pedro Perez, notary of Queen Joanna, which mill was called Abennaroça in the time of the Moors.’ 6. As per the discussion in Sect. 1.2.3, these figures are based on the reciprocals of the growth ratios shown in Table 5.1, i.e. 1/1.256 = 0.796 and 1/1.251 = 0.799. By subtracting the resultant decimal values from 1 and then converting to percentages we arrive at the rates of decay specified in the text. 7. The data on which the regression estimates in Table 5.2 are based can be found in Table A.7 in Appendix 1. 8. Etymologically, quier is commonly assumed to result from apocope of the third-person present subjunctive form quiera. However, according to Girón Alconchel (2012a), a plausible alternative source is the indicative form quiere. Either etymology is compatible with the discussion here in the text. 9. Rivero (1988: 58) assumes that the que-clauses which followed ‘qual/qui/que etc. . . quier’ sequences in Old Spanish were actually relative clauses and Company (2009: 85–86) appears to adopt the same perspective. While that analysis is no doubt correct in certain cases, specifically those in which quier was already being used in such a way that the DP containing it was no longer a wh-phrase, the observed wh-movement in examples such as (17) and (69) to (71) implies that the analysis cannot apply generally. If it did, a formulation like con qui quier que te tomasses in (70), for instance, would have to be analysed as an agrammatical structure involving two separate relative pronouns/operators. 10. In a sense, this competition is an analogue of the ambiguity posited by Rivero (1988) for Old Spanish between relative -quier constructions and quantificational ones. However, the parallel is not an exact one, given Rivero’s blanket assumption (see note 9 above) that the presence of a que-­clause after the quier-containing phrase invariably indicates that the latter is quantificational. The availability of wh-movement even in the presence of such que-clauses (see examples (17), (69), (70) and (71), as well as the later (75)) implies that Rivero’s assumption is too categorical.

202 

I. E. Mackenzie

References Alarcos Llorach, Emilio. 1994. Gramática de la lengua española. Madrid: Espasa Calpe. Barra Jover, Mario. 2007. Cambios en la arquitectura de la prosa española y romance: sintaxis y cohesión discursiva por correferencia nominal. Revista de Filología Española 87 (1): 7–43. ———. 2008. Tradición discursiva, creación y difusión de innovaciones sintácticas: la cohesión de los argumentos nominales a partir del siglo XIII.  In Sintaxis histórica del español. Nuevas perspectivas desde las tradiciones discursivas, ed. J. Kabatek, 127–150. Frankfurt: Vervuert/Iberoamericana. Bresnan, Joan, and Jane Grimshaw. 1978. The syntax of free relatives in English. Linguistic Inquiry 9: 331–391. Chomsky, Noam, and Howard Lasnik. 1977. Filters and control. Linguistic Inquiry 8 (3): 425–504. Company, Concepción. 2009. Parámetros de gramaticalización en los indefinidos compuestos del español. In Romanística sin complejos. Homenaje a Carmen Pensado, ed. Fernando Sánchez Miret, 71–103. Bern: Peter Lang. Fassi Fehri, Abdelkader. 1980. Some complement phenomena in Arabic, lexical grammar, the complementizer phrase hypothesis and the Non-accessibility condition. Unpublished MS, University of Rabat. García García, Serafina. 1992. Evolución de qual en la lengua literaria desde el siglo XIII hasta el XV. In Actas del II congreso internacional de historia de la lengua española, ed. Manuel Ariza Viguera, vol. I, 445–454. Madrid: Pabellón de España. Girón Alconchel, José Luis. 2006. Las oraciones de relativo II. Evolución del relativo compuesto el que, la que, lo que. In Sintaxis histórica de la lengua española, vol. 2, tomo 2, segunda parte: la frase nominal, ed. Concepción Company, 1477–1592. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. ———. 2012b. Gramaticalización como creación de lengua a partir del habla. Relativos e indefinidos compuestos en los Fueros de Aragón y en el Fuero de Teruel. Archivo de Filología Aragonesa 68: 15–38. Jaeggli, Osvaldo. 1980. On some phonologically-null elements in syntax. Doctoral dissertation, MIT, Cambridge, MA. Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1.

 The wh-System: Free Relatives, Double Articulation… 

203

Lapesa, Rafael. 2000. Estudios de morfosintaxis histórica del español. Madrid: Gredos. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Menéndez Benito, Paula. 2010. On universal free choice items. Natural Language Semantics 18 (1): 33–64. Pesetsky, David. 1998. Some optimality principles of sentence pronunciation. In Is the best good enough? ed. Pilar Barbosa et al., 337–383. Cambridge, MA: MIT Press. Pons Rodríguez, Lola. 2007. La qual çibdad: las relativas con antecedente adjunto del siglo XIII a hoy. Evolución de un procedimiento cohesivo. Romanistisches Jahrbuch 58: 275–305. Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. Rivero, María Luisa. 1984. Diachronic syntax and learnability: Free relatives in thirteenth-century Spanish. Journal of Linguistics 20 (1): 81–129. ———. 1988. La sintaxis de qual quiere y sus variantes en el español antiguo. Nueva revista de filología hispánica 36: 47–47. Roberts, Ian, and Anna Roussou. 2003. Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. van Riemsdijk, Henk. 2006. Free relatives. In The Blackwell companion to syntax vol. II, ed. Martin Everaert and Henk Riemsdijk, 338–382. Oxford: Blackwell.

6 Negation: Dispensing with the Clutter

6.1 Introduction Negation in Spanish, and indeed Romance generally, has attracted a good deal of attention for several reasons. Firstly, there are important theoretical questions surrounding the appropriate classification of the quantifying words involved in negation, i.e. words like Spanish nadie ‘nobody’, Catalan res ‘nothing’ and Italian nessuno ‘no/none/nobody’. In some uses, these words are like indisputably negative quantifiers such as English nothing or nobody, but in others they seem to have more in common with non-negative forms such as anything and anybody. An intersecting issue is the role of the negative morpheme no or, in Italian and in older forms of Spanish, non. While this item is capable of expressing true or ‘logical’ negation, it also occurs in so-called negative concord structures, where it functions as the concomitant of a postverbal negative quantifier rather than expressing negation in its own right. According to the Fundamental Asymmetry Hypothesis (FAH) of Longobardi 2017, this latter pattern is a reflex of the fact

© The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_6

205

206 

I. E. Mackenzie

that basic clausal negation in languages like Spanish and Italian is effected in a position to the left of the finite verb rather than to its right, as happens for example in English. According to the FAH, in Spanish and Italian-type languages the postverbal placement of negative words like nada or nessuno prevents the [+ NEG] feature of these items from ‘achiev[ing] sentential scope’ (Longobardi 2017: 118), a defect which is remedied by the insertion of no into a preverbal position. The analysis is a plausible one, but it raises the question of why Catalan, analogous as regards the locus of clausal negation, exhibits the co-occurrence of its negative morpheme with preverbal negative words, albeit on an optional basis. Issues such as these acquire a special significance in the diachronic dimension. In the first place, as Martins (2000) has observed in respect of Romance generally, the propensity for negative items like nada and ninguno to occur with a non-negative value was greater in the medieval period than it is now, a fact which has led some researchers (e.g. Poole 2011) to argue that such items underwent a transformation in their basic status. Secondly, the pattern attested in modern Catalan whereby the negative morpheme co-occurs with preverbal negative words was a salient feature of Old Romance, Spanish only fully abandoning it in the early modern period. On the face of it, this pattern might seem to be no more than a preverbal counterpart to the negative concord structure referred to above, as Posner (1984) and others have assumed. However, that approach turns out to raise more problems than it solves, suggesting that an alternative analysis may be appropriate. An additional element of this nexus of problems is the pleonastic or expletive use of the negative morpheme. As it is in a number of other Romance languages, this usage is residually available in modern Spanish, but in the medieval language it was significantly more productive, one particularly emblematic context being complement clauses governed by verbs of prohibition, denial, avoidance or doubt. Although expletive negation is usually treated as being peripheral to developments which affected negation proper, the chapter highlights important structural and diachronic linkages between the two phenomena.

  Negation: Dispensing with the Clutter 

207

6.2 Overview of the Negation System 6.2.1 N  egative Quantifiers and (Negative) Polarity Items A distinction is usually drawn between negative quantifiers (NQs) and polarity items (PIs), the latter sometimes also being referred to as ‘negative polarity items’ (Linebarger 1980). NQs, such as English nobody and nothing, are inherently negative, in the sense that their insertion into a clause suffices to ensure a negative interpretation. In contrast, PIs, prototypically exemplified by English any (excluding the ‘free choice any’ alluded to in Sect. 5.3), lack negative meaning in their own right; accordingly, they must be in the scope of another, negative expression if the clause in which they occur is to receive a negative interpretation. Examples (1) and (2) below illustrate this contrast:

(1) They did nothing.



(2) They didn’t do anything.

These two sentences have the same meaning but achieve it using, respectively, a negative quantifier (NQ) and a polarity item (PI), the latter in combination with the clausal negator not, reduced to n’t in enclitic position. In keeping with their status as not being inherently negative, PIs can appear in a recognizable but somewhat disparate class of non-negative contexts, a selection of which is shown in (3) to (8) below1: Interrogative clauses:

(3) Did they buy anything?

‘Before’-clauses:

(4) I enjoy sitting with a cup of coffee on my own before anyone wakes up.

208 

I. E. Mackenzie

Conditionals:

(5) If any water leaks out, you’ll need to reseal the joint.

After verbs of doubt, denial, prohibition etc.

(6) I doubt they’ll go anywhere else.



(7) The government denied that any new commitments had been made.



(8) They were banned from taking part in any football-related activity for two years.

Contexts of this type are a subset of what are often called polarity contexts, which for present purposes can be understood as meaning simply that they are contexts in which PIs can occur or ‘are licensed’. In addition to the nonnegative environments illustrated by (3) to (8), which Martins (2000: 195) characterizes as being broadly ‘modal’, polarity contexts also include the type of context shown in (2), which is specifically a negative polarity context. It is important to note, particularly when considering the Romance languages, that the distinction between NQs and PIs is not purely distributional; for the two types of expression differ semantically in a fundamental way. This distinction is nicely exposed by the fact that only NQs, and never PIs, can be modified by proximity-indicating adverbs such as English almost or its equivalents in other languages. Thus example (1) can be amended to (9), whereas (2) cannot be correspondingly amended to (10):

(9) They did almost nothing. (10) *They didn’t do almost anything.

Adverbs like almost can only occur in contexts that reference a limit, which may be embodied in a specific cardinal value, as with almost twenty applicants, or it may arise from universal quantification, either of the positive variety (almost every applicant) or the negative one (almost no applicants).2 This constraint rules out the insertion of almost into contexts which involve existential

  Negation: Dispensing with the Clutter 

209

quantification, given that such quantification does not reference a limit: *almost some people, *almost several books etc. The ungrammaticality illustrated by (10) can thus be interpreted as indicating that PIs are existential quantifiers. This latter point is corroborated to an extent by the fact that, in premodern Spanish, the existential quantifier alguno ‘some’ could be used as a PI, as is illustrated by (11) below, from a fifteenth century manuscript:

(11) Et por ende mandamos que este obispado de leon que non sea obediente a algund arçobispado (Crónica de 1344, fol. 44r) ‘And we therefore decree that this bishopric of Leon should not be under the jurisdiction of any archbishopric.’

In light of the foregoing remarks, I will assume that NQs express negative universal quantification (∀x ~) and that PIs express existential quantification (∃x). For example, the sentences Nobody saw Harry and Harry didn’t see anyone should be analysed as in (12) and (13) respectively: (12) ∀x (x is a person → ~ x saw Harry)

(13) ~ ∃x (x is a person & Harry saw x)

Notice that the negation operator ‘~’ in (12) reflects the [+ NEG] meaning of the NQ nobody, whereas in (13) it is the analogue of the clausal negation marker not (reduced to n’t), a by-product of the fact that PIs are not negative in their own right.

6.2.2 N-Words in Modern Spanish 6.2.2.1  NQs Versus PIs One of the striking differences between the English type of negation system and that found in languages like modern Spanish and Italian is that the negation-related quantifiers or N-words (Laka 1990) of the latter languages exhibit contrasting behaviour depending on whether they are preverbal or postverbal. When preverbal, they are exactly analogous to English-style NQs; when postverbal, however, they only have a negative

210 

I. E. Mackenzie

meaning if they are in the scope of the negative morpheme (e.g. no or non) or, less commonly, if they are in the scope of a preverbal NQ. Thus the insertion of no into immediately preverbal position in any of the sentences in (14) would result in ungrammaticality or, in the case of Nadie habló, it would force a rather awkward “double negation” reading: (14) Nadie habló/Nunca llaman/De nada me arrepiento/Jamás te olvidaré ‘Nobody spoke/They never call/I regret nothing/I will never forget you.’

Conversely, the negative morpheme no cannot be removed from the semantically equivalent sentences in (15) without harming their grammaticality:

(15) No habló nadie/No llaman nunca/No me arrepiento de nada/No te olvidaré jamás ‘Nobody spoke/They never call/I regret nothing/I will never forget you.’

In addition, like their counterparts in a variety of other Romance languages, Spanish N-words can be used non-negatively in a range of polarity contexts: Interrogative clauses:

(16) ¿Jamás te hizo daño? ‘Did he/she/it ever hurt you?’



(17) No sé si estará nada abierto el domingo. ‘I don’t know whether anything will be open on Sunday.’

‘Before’-clauses:

(18) Antes de hacer nada, consúltalo con la almohada ‘Before doing anything, have a good night’s sleep.’

  Negation: Dispensing with the Clutter 

211

Comparatives:3

(19) Siendo mediocampista, remata mejor que nadie. ‘Despite being a midfielder, he/she finishes better than anyone.’

Verbs of doubt etc:

(20) Dudo que ninguno de ellos sea quien te denunció. ‘I doubt whether any of them is the person who reported you.’

Conditionals:

(21) Si encuentras a nadie igual, cásate con él. (Bosque 1980: 28) ‘If you find anyone similar, marry him.’

It is worth pointing out that some speakers find examples like (21) to be rather unnatural, conditional clauses in modern Spanish being a somewhat marginal context for the non-negative use of N-words. The fact that, in their postverbal use, Spanish N-words require the ancillary presence of the negative morpheme (or another, preverbal N-word), combined with their ability to be used non-negatively in the types of context illustrated by (16) to (21), suggests a parallel with the English any forms. Accordingly, a number of authors (e.g. Bosque 1980; Laka 1990; Suñer 1995) have proposed that Spanish N-words are PIs rather than NQs. Conversely, if priority is given to the fact that preverbal Spanish N-words appear to be inherently negative  – analogously to English nobody, nothing etc. – the model advanced by Haegeman and Zanuttini (1991), according to which N-words are NQs, becomes more attractive. Pursuing a via media, Longobardi (2017) argues that Spanish (and Italian) N-words are ambiguous between PI and NQ uses, while Espinal (2000) adopts a partially similar approach for Catalan and Spanish, albeit she stresses (p. 578) that the relevant N-words are ‘semantically underspecified’ rather than ambiguous. Acceptance of some degree of ambiguity does in fact appear to be inescapable; although, as will become apparent presently, the ambigu-

212 

I. E. Mackenzie

ity does not align with the preverbal versus postverbal distinction. The key test in this regard is whether it is possible to insert the adverb casi, which is the direct equivalent of English almost. As was implicit in the discussion earlier, this is a semantic test discriminating between different types of quantifier, with NQs and PIs falling on opposite sides of the dividing line (cf. Español Echevarría 1994: 3–4). Accordingly, if Spanish N-words are subject to an NQ–PI ambiguity, the test should be able to expose this. It turns out that casi-insertion is fine for both preverbal N-words and for postverbal ones which follow a verb negated by no, but not for N-words used non-negatively in polarity contexts:

(22) Casi nadie vino/De casi nada me arrepiento/Casi nunca llaman/ Casi ninguna tienda estaba abierta.4 ‘Almost nobody came/I regret almost nothing/They almost never call/Almost no shops were open.’



(23) No vino casi nadie/No me arrepiento de casi nada/No llaman casi nunca/No estaba abierta casi ninguna tienda. ‘Almost nobody came/I regret almost nothing/They almost never call/Almost no shops were open.’



(24) *No sé si estará casi nada abierto/*Antes de hacer casi nada/*Remata mejor que casi nadie/*Dudo que casi ninguno de ellos sea quien te denunció. ‘*I don’t know whether almost anything will be open/*Before doing almost anything/(S)he is a better finisher than almost anyone5/*I doubt almost any of them is the person who reported you.’

Given this pattern of data, it would appear that Spanish N-words are indeed ambiguous between an NQ reading and a PI one. Crucially, however, postverbal position after a previous no is an NQ context rather than a PI one. Structures of this latter type (i.e. no . . . nada/nadie/nunca etc.) involve so-called negative concord. According to Longobardi’s (2017: 118) Fundamental Asymmetry Hypothesis, the existence of negative concord in languages like Spanish and Italian is a corollary of the fact that basic clausal negation in these languages is realized in a ‘pre-INFL’ position;

  Negation: Dispensing with the Clutter 

213

that is, to the left of the finite verb (henceforth designated as ‘VFIN’). This circumstance implies that in the relevant languages a clause can only be interpreted as expressing a negation if it contains a mark of negation to the left of VFIN. The asymmetry between preverbal and postverbal N-words as regards insertion of the negative morpheme (no in modern Spanish; non in Italian) follows straightforwardly from this constraint: preverbal placement of the N-word supplies the required mark of negation to the left of VFIN, whereas postverbal placement does not. The latter case must be supplemented by the insertion of the negative morpheme into the pre-INFL position, an operation which delivers the prototypical instance of negative concord, viz. the surface string ‘no(n) + VFIN + N-word’.

6.2.2.2  Ambiguity of no As was noted in the previous section, N-words in the no . . . nada/nadie/nunca type of structure are interpreted as NQs. Significantly, however, if the requirement for a pre-INFL mark of negation is satisfied by a preverbal NQ rather than by the negative morpheme no, any postverbal N-word must be interpreted as a PI and not as an NQ. This is demonstrated by the failure of casi-insertion in the examples in (25) below:

(25) Nadie compró (*casi) nada/Nunca ves a (*casi) nadie por aquí/ Ningún entrenador sobornó a (*casi) ningún arbitro. ‘Nobody bought (*almost) anything/You never see (*almost) anybody round here/No manager bribed (*almost) any referee.’

Preverbal NQs can presumably be regarded as being items which are indisputably [+ NEG], given that their presence alone will force a clause to be interpreted negatively. Thus the data pattern in (25) is plausibly understood as reflecting the fact that an N-word will be interpreted as a PI whenever it is in the scope of a genuine [+ NEG] item. This has an important implication for the negative morpheme no insofar as it occurs in negative concord structures. Given that, in the latter type of case, the N-word in the scope of no is interpreted as an NQ and not as a PI, it follows from (25) and the construction put upon it, that the no of negative concord is not actually a genuine [+ NEG] item. Instead, it must be seen

214 

I. E. Mackenzie

as being merely the syntactic concomitant of an NQ, combining with the latter to deliver a discontinuous expression of negation: no . . . nada, no . . . nadie etc. In more theoretical terms, and following the general approach of Roberts (2007: 68–69), Spanish no in negative concord structures can be seen as having only an ‘uninterpretable’ negative feature, i.e. a purely formal one with no semantic import. The ‘interpretable’ or semantically relevant [+ NEG] feature should then be seen as being borne by the associated NQ, with which no ‘agrees’, much as a finite verb agrees in terms of person and number with its subject. As a direct consequence of the foregoing, it must also be inferred that the no of negative concord is nonidentical with the irreducibly [+ NEG] no of clausal negation, exemplified in (26):

(26) No rechistaron. ‘They didn’t flinch.’

Here no can legitimately be viewed as a being the Spanish-specific expression of the logical negation operator ‘~’, in the sense that if rechistaron ‘they flinched’ is p, then (26) is ~ p. The distinction between this use of no and the concord-related one just considered is sharply disambiguated in fronting structures, [+ NEG] no being required (if the statement is a negative one) but negative concord no being inadmissible:

(27) Dinero no le di dinero. ‘Money I didn’t give him.’



(28) Nada (*no) le dieron nada para comer. ‘They give him nothing to eat.’

It is also revealed by the negation pattern of Standard French, which requires ne . . . pas for cases like (26), but just ne for cases which also involve an NQ:

(29) Ils n’ont pas tressailli. ‘They didn’t flinch.’

  Negation: Dispensing with the Clutter 



215

(30) Ils n’ont (*pas) vu personne. ‘They didn’t see anybody.’

The distinction between the [+ NEG] use of no and its use in negative concord structures does not in fact exhaust the functional diversity of the negative morpheme, given examples such as the following:

(31) Es mejor vacunarse ahora que no tener que ir a urgencias en invierno. ‘It’s better to get vaccinated now than to have to go to A&E in the winter.’

Here no is self-evidently not a component in a negative concord structure. Nor, however, is it [+ NEG]: what is asserted is that getting vaccinated now is better than going to A&E in the winter, not that it is better than not going. Recall, however, that comparative structures create polarity contexts, in the sense that they license the use of N-words without a negative interpretation, as was illustrated by (19), repeated below:

(19) Siendo mediocampista, remata mejor que nadie. ‘Despite being a midfielder, he/she finishes better than anyone.’

The use of no illustrated by (31) is in fact limited to polarity contexts. This restriction, together with the fact that the negative morpheme in this type of occurrence is entirely uninvolved in any actual negation, points to an analysis of the relevant no as a semantically empty reflex of the containing polarity context.6 As is discussed in more detail in Sect. 6.3 below, this expletive or pleonastic use of the negative morpheme has been lost in the majority of contexts in which it could previously be encountered in Spanish (see Sánchez López 1999: 2628–2629 for an overview of its residual presence in the modern language). In Catalan, however, it appears to be still productive across a range of contexts. Espinal (2000: 574, n. 21) reports that expletive no is licensed after verbs of doubt and prohibition, as well as témer ‘to fear’, in addition to being found in comparative structures and after subordinating conjunctions such as fins ‘until’ and abans ‘before’.

216 

I. E. Mackenzie

Indeed, an example such as (32) below, from Espinal (2000: 560), is the direct analogue of a pattern that was once commonplace in Spanish:

(32) Dubto que no mengi. ‘I doubt that he eats.’

Summarizing the key points of this section, the syntactic negative morpheme in Spanish-type languages can be analysed as having three distinct uses, as listed in (33): (33) Three types of occurrence of the negative morpheme (no/non etc.): (i) Clausal negation ([+ NEG]) (ii) Concomitant of an NQ (negative concord) (iii) Semantically empty reflex of a polarity context (expletive negation)

6.3 Negation in Old Spanish 6.3.1 F undamental Similarities with the Modern System In broad terms, the system of negation found in the High and Late Middle Ages is already similar to the modern one. In the first place, the Old Spanish negative morpheme no(n) exhibits the three-way ambiguity captured in (33) from the outset of the literary period: Clausal negation:

(34) Rachel non se enprennaua. (Fazienda de ultramar, fol. 4v) ‘Rachel could not conceive.’

Negative concord:

(35) & sil matare el qui demandare los fiadores de saluo. non peche nada ni esca enemigo. (Fuero viejo de Alcalá, fol. 20r)

  Negation: Dispensing with the Clutter 

217

‘And if he is killed by the person who requested the guarantors, that person pays no fine, nor can the victim’s family declare him to be an enemy.’

Expletive negation:

(36) E por end touo por meior de morir que non fazer ninguna destas cosas. (Estoria de España I, fol. 26r) ‘She therefore thought it better to die than to do any of these things.’

As regards the N-words themselves, it is notable that they were used more commonly as (non-negative) polarity items than their modern counterparts are and also in a wider range of (non-negative) polarity contexts (see also Poole 2011; Martins 2000). Thus in addition to there being a greater frequency of examples like (37) below, which is analogous to a modern formulation such as antes de que nadie se despierte ‘before anybody wakes up’, we also find examples like (38) and (39), in which the N-words appear in contexts from which N-words are excluded in the modern language.

(37) & mandamos la tener en tal manera. que ante que ninguno aya el regno; ante prometa por sacramento de guardar esta ley. (Fuero Juzgo, fol. 11r) ‘And we stipulate that it should be observed in this way: before any individual accedes to the throne, they must swear an oath to uphold this law.’



(38) Et si nenguna mugier passare estos cotos delas Tocas & delas otras cosas como sobredicho ca si non como yo mando. que peche .xx. maravedis en coto por quantos dias lo uistiere. (Alfonsine diploma, 6 October 1252, Seville; Archivo municipal de Alcalá de Henares, Leg. 9000/1) ‘And if any woman infringes these prohibitions on gauze and the other things specified, in accordance with my stipulation, she must pay 20 maravedis as a fine for however many days she wore these things.’

218 



I. E. Mackenzie

(39) Et qui njnguna cosa fiziere fuera ende estas njn contra njnguna destas fuera para desfazer. culpado sera. (General estoria V, fol. 122v) ‘And whoever does anything outside of these stipulations or seeks to overturn them will be called to account.’

Given the continuing possibility in modern Spanish of using N-words non-negatively in a range of polarity contexts – see (16) to (21) in Sect. 6.2.2.1 – it is examples like (38) and (39) which really purport to evince a difference between the medieval and modern situations. However, despite attracting quite a lot of attention in the literature, these kinds of example are actually rather unrepresentative of medieval usage.7 In the first two data points of the present corpus, covering the period 1250–1289 (2,826,000 words), (38) is the only instance I could find in which an N-word or a DP headed by an N-word occurs as the (non-negative) subject of a conditional clause. In contrast, there are 853 cases in which the pronoun alguno ‘someone’ or a DP headed by a form of alguno qua determiner has that type of occurrence. Examples (40) and (41) are two illustrations of this common arrangement:

(40) & si alguno mostrare desden quel escarmienten por ello (Poridat de las poridades, fol. 6r) ‘And if anyone shows defiance, he should be punished for it.’

(41) Si algun sieruo ati fuxiere. nol des a su sennor. Mas more con tigo enel logar o quiere. (General estoria I, fol. 332r) ‘And if any slave flees to you, do not return him to his master. Rather he should live with you in a place of his choosing.’

As regards the free relative structure illustrated by (39), that was actually uncommon both with N-words and with items like alguno, only two additional cases with an N-word being recoverable from the 1250–1289 section of the corpus and eleven with alguno. Despite the small number of tokens, however, there is a clear asymmetry between the N-words and alguno. The somewhat atypical nature of examples like (38) and (39), combined with the fact that N-word usage (without a negative meaning) in polarity contexts is still reasonably productive even in the twenty-first century, suggests that the differences in this area between Old Spanish

  Negation: Dispensing with the Clutter 

219

and modern Spanish are a matter of degree rather than kind. This fits in with the general drift of Romance, in which, according to Martins (2000: 215), N-words ‘evolved towards reducing their degree of variable underspecification, thereby becoming more restrictive with respect to their licensing contexts’. Accordingly, it is probably safer to assume that N-words in medieval Spanish, like their counterparts in modern usage, were NQs that could be used as PIs, rather than that they were invariably PIs on a par with the English any forms, as is proposed by Poole (2011). The view advanced by Poole receives some support from the fact, discussed in Sect. 6.3.2 below, that for much of the Old Spanish period preverbal N-words (other than nunqua/nunca ‘never’) were usually accompanied by the negative morpheme no(n). This circumstance might be taken to be an indication that the N-words of the period were more like English polarity items such as any, anybody and anything, which must co-occur with not (or a negative quantifier) in order to express a negation. However, the parallel between medieval Spanish N-words and items like English any immediately breaks down in view of the contrasting scope relations. Thus the English any-forms are required to be within the scope of the relevant [+ NEG] item, a constraint which results in the exclusion of formulations like *Anybody didn’t write a letter, whereas in the Old Spanish case it is the N-words themselves which take scope over the negative morpheme, the relevant surface string being ‘N-word + no(n) + VFIN’. The latter string is in fact amenable to better analyses in which the N-word is treated as an NQ rather than as a PI. Specifically, the relationship between the N-word and no(n) could be analysed as a form of negative concord (see Sect. 6.2.2.1 above), or, as will be suggested in Sect. 6.4, the negative morpheme could be seen as having an expletive occurrence. An additional piece of evidence which has been claimed to support the idea that Old Spanish N-words were PIs rather than NQs is the fact that no(n) could be prefixed to an N-word in contexts in which there was no verb to host the negative morpheme, as in examples like (42) and (43):

(42) E el pierde & su enemjgo non njnguna cosa & prouoca la yra de nuestro señor dios contra sy por rrazon dela vengança asy commo dicho es. (Libro de las donas, fol. 73v)

220 

I. E. Mackenzie

‘And he loses while his enemy [loses] nothing and he provokes the wrath of our lord God against himself because of his vengeance, as has been stated.’

(43) todo para ti: & no nada de que puedas dar parte. (Comedia de Calisto y Melibea, fol. 36r) ‘Everything for yourself and nothing which you could share with others.’

Camus Bergareche (2006: 1179–1180) argues that examples of this type show that Old Spanish N-words were incapable of having a negative value unless they were in the scope of an ancillary negative expression, and hence they were not genuine NQs. Certainly, the pattern illustrated in (42) and (43) is intriguing; however, there are a number of reasons why one should be cautious about using it as the basis for wider claims concerning negation in Old Spanish. In the first place, tokens of ‘no(n)–N-word’ (i.e. no(n) immediately followed by, or joined to, an N-word) are actually rather uncommon, particularly in the texts of the High Middle Ages. This is strikingly illustrated by the fact that there are only three tokens of this structure in the first two data points of the corpus, covering the period 1250–1289, which have a combined size of 2,826,000 words.8 This is particularly significant given that the earliest sections of the corpus are precisely those in which, under the ‘N-words as PIs’ hypothesis, one would expect tokens of ‘no(n)–N-word’ to be at their most productive. This point follows from the chronology relating to the evolution of the ‘N-word + no(n) + VFIN’ structure discussed in Sects. 6.3.2 and 6.4. As can be seen from Fig. 6.1, together with the discussion in Sect. 6.4, the rate of occurrence of this latter structure declines over the course of the Middle Ages, the bulk of the decrease occurring before 1500. However, this is the very process which is claimed to reveal the putative change in the status of the N-words from PIs to NQs (see Poole 2011). Thus if ‘no(n)–N-word’ was indeed a reflex of the alleged PI status of the Old Spanish N-words, one would expect its rate of occurrence to decrease over time, implying that tokens of it should be at their most abundant at the earliest sampling points in the corpus.

  Negation: Dispensing with the Clutter 

221

100% 90% 80% 70% 60% 50% 40%

Ninguno/Nenguno*

30%

Verbs of prohibition, denial, etc.

20% 10%

Jamas/Iamas Nada

Nadie/Nadi 0% 1260 1290 1320 1350 1380 1410 1440 1470 1500 1530 1560 1590

Fig. 6.1  Decline in the use of no(n) after five medieval triggers (1250–1609) (*Includes both pronominal and determiner uses)

In fact, not only is ‘no(n)–N-word’ rare in the first two data points of the corpus, as was just noted, but the structure actually seems to increase in frequency at the end of the Middle Ages, although numbers remain small. For example, there are twenty-two tokens in the 1500 sampling point, which averages data for the period 1490–1509 and has a size of 1,231,000 words. If anything, then, the quantitative evidence points towards an increase in the frequency of forms like no(n) ninguno and no(n) nada precisely at a time when, if such forms were diagnostics of the PI status of the N-words, one would expect them to be in decline. Further grounds for scepticism stem from the fact that the prefixation of no(n) was not in fact obligatory when N-words occurred in verbless contexts, as can be seen from examples like (44) and (45):

(44) Seras tu sennor de muchas naciones. et njnguno de ti. si assi fizieres & uisquieres como te yo ensenno. (General estoria I, fol. 328r) ‘You shall be master of many nations, but none [will be master] of you, if you behave in this way and live as I teach you.’

222 

I. E. Mackenzie



(45) E si fueren en diuerso estado desto que dixiemos; significa que aura pocos hermanos o ninguno. (Judizios de las estrellas, fol. 168v) ‘And if they were arranged differently from what we said, this means that he will have few brothers or none.’



(46) & que atales ay dellas que quando las matan. que non les fallan baço. si non muy pequenna sennal por el. o aun que en tales ya que njnguna. (General estoria I, fol. 243r) ‘And that there are examples of them in which, when they are killed, a spleen cannot be found, except a very small residue of one, or even, in such specimens, no [residue] at all.’

It is also worth noting that, to judge from some of the later texts, forms like no(n) ninguno and no(n) nada could occur without a negative meaning. This possibility is illustrated in (47) below, where no ningun cuero occurs as a comparative phrasal complement.

(47) Avn que vna vez me dixo vn doctor que mayor causa era para fazer piedra en los rriñones el mejor pescado del mun do: que no ningun cuero avn que fuesse de toro. (Cura de la piedra, fol. 6v) ‘Although a doctor once told me that the best fish in the world was a greater cause of kidney stones than any animal skin, even a bull’s.’

While expletive no(n) routinely appeared in comparative structures in Old Spanish (cf. (36) above), it was usually omitted in cases in which the term of the comparison also contained an N-word (used as a PI). In other words, the pattern in (48) below was the more usual one:

(48) vos erades matador delos puercos monteses & delos osos qujer de Cauallo qujer de pie meJor que njnguno otro; (Crónica de 1344 I, fol. 94r) ‘You were a better slayer of wild boars and bears, either on horse or on foot, than anyone else.’

In light of this fact, no ningun cuero in (47) could potentially be analysed as being simply a stylistic alternative to ningun cuero, possibly having a more emphatic value. An analysis along similar lines for the [+ NEG] variant of ‘no(n)–N-word’, illustrated in examples like (42) and (43),

  Negation: Dispensing with the Clutter 

223

would be go some way towards explaining the somewhat non-systematic properties of this structure, viz. (i) the optionality of no(n) in such formulations, (ii) their infrequency in the medieval corpus and (iii) the fact that they become more frequent precisely at a time when, given the evolution of the system as a whole, one would expect them to be in decline. Regardless of whether or not one accepts this latter analysis, the facts reported above concerning ‘no(n)–N-word’ seriously call into question the utility of this structure as a diagnostic for assessing the status of the Old Spanish N-words. Accordingly, I will assume that its existence does not undermine in any significant way the conclusion drawn earlier, viz. that the safest analysis is to regard the Old Spanish N-words as being, like their modern counterparts, NQs that could also be used as PIs. Now Longobardi (2017: 113) proposes three ‘core dimensions of parametrization’ for negation systems generally, viz. (i) does the basic clausal negator precede or follow the finite verb? (ii) can the negative morpheme express logical negation or is it merely a scope marker (like French ne)? and (iii) are the N-words NQs or PIs? The assessment just made concerning the Old Spanish N-words, combined with the continuity between medieval no(n) and modern no alluded to earlier, implies that the Old Spanish negation system had the same settings in terms of Longobardi’s framework as the modern system does. This is spelled out explicitly in (49): (49) Convergent parametric settings of Old and Modern Spanish negation: (i) Clausal negation is realized to the left of VFIN; (ii) The negative morpheme is not merely an empty scope marker but is also capable of functioning as the logical negation operator; (iii) N-words are NQs (which may or may not also be capable of functioning as PIs) rather than invariable PIs.

6.3.2 Use of no(n) with Preverbal N-Words Despite the foregoing continuity overall, there are two ways in which negation in the medieval language differed from what is observed in the modern period. The first of these relates to the availability in Old Spanish

224 

I. E. Mackenzie

of a negative concord-like structure in which the finite verb in a clause that also contained a preverbal N-word was formally negated by no(n), resulting in the surface string ‘N-word  +  no(n)  +  VFIN’ (where ‘VFIN’ means ‘finite verb’). The majority of the examples of this structure involve some form of ninguno/nenguno, given that (i) preverbal nunqua/nunca almost never co-occurred with the negative morpheme and (ii) both nadie/nadi and preverbal nada, together with iamas/jamas when used as an N-word, are relatively infrequent in the manuscripts until the late medieval period, by which time the structure was in decline. Notwithstanding this lexical imbalance, illustrations involving various different N-words are given in (50) to (53) below.

(50) solamientre nenguna delas partes non sea destoruada. por grandes uozes. nin por grandes bueltas. (Fuero Juzgo, fol. 17r) ‘Just that neither of the parties should be at a disadvantage due to shouting or disorder.’



(51) [. . .] mas que non de espantar los nin de foyr le ellos. commo lo farien quando sopiessen que nada non tenie. (General estoria IV, fol. 12r) ‘. . . rather than frightening them or making them flee him, as they would do if they knew he had nothing.’



(52) de auer soltura dellas es ya de fiar enel onbre que jamas non curara dello njn avn delo pensar (Mostrador y enseñador de los turbados, fol. 132r) ‘And if they were forgiven, it can be assumed of man that he will never be cured of this nor even of thinking about it.’



(53) ca nunqua non uiemos en esta nuestra sazon nenguno que dixiesse que llego fata ella; (Cánones de Albateni, fol. 6v) ‘For we never saw, in the present era, anyone who said they reached it.’

As was indicated above, the pattern in this last example is rare, the usual one for nunqua/nunca being the one shown in (54)9:

  Negation: Dispensing with the Clutter 



225

(54) Et si pusieren un poco della en mucha dagua & ruciaren con ella los maderos; nunqua se podreçran ni torceran. (Lapidario, fol. 74r) ‘And if a little of it is placed in a lot of water and then planks of wood are coated with it, they will never rot or warp.’

Camus Bergareche (1986, 2006) argues that ‘N-word + no(n) + VFIN’ reflects the existence of an essentially homogeneous medieval system, which he labels ‘Negación de Tipo Medieval’. Under his analysis, this system remained in place until the mid-fifteenth century, at which point the use of no(n) with preverbal N-words was lost in the period of a few decades. However, even if we ignore the fact that, self-evidently, the N-word nunqua/nunca must have been outside the system envisaged by Camus Bergareche, it is worth noting that variation in terms of the presence/absence of no(n) is detectable already in the thirteenth century. For example, preverbal ninguna has a similar occurrence in each of (55) and (56) below, both from the same text, and yet in the first example it appears without non but in the second it occurs with it:

(55) & por esto que uiene en tod en todo. & non es sabudo quando; por ninguna guisa es la cosa que ell omne mas deue temer. (General estoria IV, fol. 234r) ‘And because of the fact that it always comes, but one does not know when, it is in no way the thing that man must fear most.’



(56) Los de Macedonia quando uieron a los elefantes daquella guisa ouieron grand miedo & por ninguna guisa non se atreuieron de yr a ellos. (General estoria IV, fol. 224r) ‘The Macedonians, when they saw the elephants arranged in that way, took fright and did not dare in any way to charge at them.’

Camus Bergareche (1986: 113; 2006: 1196) also recognizes the existence of medieval examples in which preverbal N-words do not occur with no(n), but regards these as exceptions to an otherwise general rule. In contrast, the quantitative data adduced in Sect. 6.4 below, together with the S-curve corresponding to ninguno shown in Fig. 6.1, suggest that the relevant medieval examples reflect variation of the kind which is normal

226 

I. E. Mackenzie

during a period of syntactic change. This latter approach also turns out to be more consistent with the profile of the change at the end of the Middle Ages, when, as Fig. 6.1 also indicates, the loss of ‘N-word + no(n) + VFIN’ was not quite as abrupt as Camus Bergareche suggests (see also note 14 below). Turning now to the structural analysis of the ‘N-word + no(n) + VFIN’ pattern, the standard assumption (see e.g. Posner 1984) is that it involves negative concord, implying that the N-word is an NQ and the associated negative morpheme is a syntactic marker rather than an expression of logical negation.10 This is clearly a possibility, although it raises the question of why preverbal negative concord should be diachronically unstable while negative concord involving a postverbal N-word has been completely inert over the course of at least eight centuries. A synchronic analogue of this asymmetry can also be highlighted in modern Catalan. The latter language has generalized negative concord for postverbal N-words and it also allows structures like the one in (57): (57) Ningú no ha menjat postres. (Espinal et al. 2016: 147) ‘Nobody ate dessert.’

However, unlike the no which construes with a postverbal N-word, the no in examples like (57) can be dropped without harming grammaticality.11 Both the modern Catalan facts and the diachronic Spanish ones suggest that one should be wary of assuming that the relationship between the negative morpheme no(n) and an associated N-word is the same regardless of whether the N-word is preverbal or postverbal. I return to this issue in Sect. 6.4, after consideration has been given to the subject of expletive negation, i.e. the type (iii) use of no(n) in terms of the taxonomy given earlier in (33).

6.3.3 Expletive Negation The expletive or pleonastic use of medieval no(n) was significantly more productive than is the corresponding usage in the modern language. In particular, no(n) commonly appeared in subordinate clauses, both finite

  Negation: Dispensing with the Clutter 

227

and nonfinite, as a function of the meaning of the governing verb, relevant semantic triggers being prohibition, avoidance, doubt and denial. Representative examples are given below in (58) to (61):

(58) les deffendio que non aorassen njnguna destas cosas en ydolo njn sin ydolo. (General estoria I, fol. 181v) ‘He forbad them to worship any of these things either made into an idol or not.’



(59) empero razones y ha en que algunos omnes se podrien escusar de no caer en la pena del danno que les podrie auenir en razon de sus cosas. (Libro de las leyes, fol. 2v) ‘Nevertheless, there are reasons for which certain people could escape the punishment which should befall them in view of their actions.’



(60) & quiludema non dubdaua que non matassen los mas de la yente dell otra part que uinie con aquell adelantado. (General estoria IV, fol. 33r) ‘And Quiludema did not doubt that they would kill the majority of the soldiers in the opposing band who accompanied that general.’



(61) & el sieruo començo de luego a negar que no fiziera aquel mal quel dizien. (Estoria de España I, fol. 87v) ‘And the slave began immediately to deny that he had done that thing that was being attributed to him.’

Verbs that express the relevant type of meaning – verbs of prohibition, doubt etc. – create polarity contexts, i.e. contexts which admit PIs like English anyone or N-words (nada, nadie etc.) used as PIs. This latter point is actually exemplified in (58), where njnguna in the object phrase njnguna destas cosas occurs as a PI rather than an NQ. N-words continue to be capable of being used in this type of context in modern Spanish, as was illustrated earlier by example (20), repeated below:

(20) Dudo que ninguno de ellos sea quien te denunció. ‘I doubt whether any of them is the person who reported you.’

228 

I. E. Mackenzie

In both Old Spanish and Modern Spanish, then, subordinate environments created by verbs in the relevant meaning class constitute polarity contexts. Thus what sets the medieval language apart in this respect is the fact that the embedded verb could also be (syntactically) negated in such contexts. However, in keeping with the general analysis of expletive negation proposed in Sect. 6.2.2.2, in this particular use no(n) is most sensibly regarded as being a purely formal element, a semantically empty reflex of the containing polarity context. Now it was also observed in Sect. 6.2.2.2 that modern Catalan retains this use of the syntactic negative morpheme, the example given at that point being (32), which is now repeated below:

(32) Dubto que no mengi. ‘I doubt that he eats.’

The availability of this structure in modern Catalan is interesting, given that the latter language, as was observed in the previous section, also allows its syntactic negative morpheme no to be used in conjunction with preverbal N-words. We thus have a situation in which both Old Spanish and modern Catalan have what might be termed a productive system of expletive negation and also allow their syntactic negative morpheme to co-occur with preverbal N-words, whereas modern Spanish exhibits neither of these characteristics. The fact that the two characteristics appear to have a tendency to co-exist and also to disappear together suggests that they may be linked. The nature of such a possible linkage, together with the quantitative evolution of both of these negation-related phenomena, is discussed in the next section.

6.4 Quantitative Evolution of the System As is implied by the discussion in the previous sections, the principal developments in the negation system since the High Middle Ages are (i) the loss of no(n) with preverbal N-words and (ii) the loss of (expletive) no(n) in subordinate clauses governed by verbs of prohibition, denial, doubt etc.12 Figure 6.1 in Sect. 6.3.1 shows the quantitative decline in these

  Negation: Dispensing with the Clutter 

229

two uses of no(n) over the course of the time window covered by the present corpus, the underlying numerical data being viewable in Table A.8 in Appendix 1. It will be noticed that the curves for nada, nadie/nadi and jamas/iamas in the figure are incomplete. This is due to the fact that these items, either in general or specifically as preverbal N-words, do not provide enough tokens for statistical analysis until the latter stages of the Middle Ages. This is strikingly illustrated by nadie, together with its early variant nadi, which have a combined total of just seven tokens in the first twelve data points of the corpus (7,852,000 words), spanning the period 1250–1489.13 Despite discrepancies of this type, it can be seen that all of the N-words gravitate at around the same time towards the same steady state as regards their propensity to co-occur with the negative morpheme in preverbal position. That is to say, such co-occurrence comes to be permanently excluded at the end of the sixteenth century. However, for the bulk of the preceding period, the decline of no(n) with preverbal N-words is expressed fundamentally by the decline of no(n) with preverbal ninguno/nenguno. This latter process, as the graphic shows, describes a rather neat (inverted) S-curve, whose midpoint is located somewhere between the 1440 and 1460 data points.14 As regards the use of no(n) in complements of verbs of prohibition, denial, doubt etc., it can be seen from the figure that this usage declines over the same period that ‘ninguno/nenguno + no(n) + VFIN’ does, although the curve for the verbal complements becomes much less steep towards the end of the Middle Ages. The difference in the overall steepness of the respective curves is reflected in the differing decadal odds ratios associated with the two processes, which can be seen in Table 6.1. The ORs shown in the table for these two contexts imply that the likelihood of no(n) being used in conjunction with preverbal ninguno/nenguno decreased by 31.6% per decade, whereas the corresponding decrease in the complements of the relevant verbs was 7% per decade. Given Kroch’s Constant Rate Effect (see Sect. 1.2.1), these sharply differing rates of decline might appear to suggest that the corresponding processes should not be viewed as manifestations of a single linguistic change. However, as was discussed in Sect. 3.3.2, the materialization of a common rate of change across two or more contexts is dependent to a

230 

I. E. Mackenzie

Table 6.1  Rate of decline in use of no(n) in five contexts (1250–1609) Context After preverbal ninguno/nenguno After preverbal jamas/iamas After preverbal nada After preverbal nadie/nadi In complements of certain verbsa

Odds ratio (per decade)

Percentage equivalent

0.684

31.6

0.637 0.900 0.451 0.930

36.3 10.0 54.9 7.0

Verbs of prohibition, denial, doubt etc.

a

large extent on the contexts being defined fairly broadly. Clearly, however, the ninguno/nenguno context is very narrowly defined, being specific to a single lexical item. Self-evidently, this state of affairs implies that the observable rate of change is likely to be subject to a lexically-conditioned effect. Indeed, as the remainder of Table 6.1 shows, the rate of decline in the use of no(n) is actually not consistent across the four N-words that regularly co-occurred with no(n) in preverbal position. Moreover, it should not be forgotten that in the case of the N-word nunqua/nunca, there has been almost no change at all as far as co-occurrence (in preverbal position) with no(n) is concerned, given that no(n)-usage in this case was already very rare at the beginning of the time window examined here. The rate of decay in the use of no(n) with preverbal N-words is thus subject to a significant degree of lexical bias, meaning that any comparison with the decay rate for no(n) in other contexts, such as in complements of verbs of prohibition, doubt etc., is statistically unreliable. Accordingly, the differing rates of decline shown in Table 6.1 should not be assumed to rule out the possibility that the loss of the no(n) that was a concomitant of verbs in the relevant meaning class and the loss of the no(n) that was a concomitant of preverbal N-words represent a single underlying change. What we do see, on the other hand, is that these two processes were roughly synchronous, albeit the loss of no(n) with preverbal N-words reached completion at a time when no(n) could still be used ‘expletively’ in complements of the relevant verbs. In light of this approximate synchronicity and given that both processes have an identical surface manifestation, viz. the decline and

  Negation: Dispensing with the Clutter 

231

eventual loss of the form no(n), it is not implausible to wonder whether the item no(n) does not in fact have the same type of occurrence in both contexts. The usual assumption, in terms of taxonomy (33), repeated below, is that the no(n) in the string ‘N-word + no(n) + VFIN’ represents a type (ii) occurrence of the negative morpheme, while the no(n) triggered by verbs of prohibition, doubt etc. involves a type (iii) occurrence. (33) Three types of occurrence of the negative morpheme (no/non etc.): (i) Clausal negation ([+ NEG]) (ii) Concomitant of an NQ (negative concord) (iii) Semantically empty reflex of a polarity context (expletive negation)

However, the distinction between types (ii) and (iii) assumes tacitly that the NQ specified for type (ii) is postverbal, negative concord then arising to supply a preverbal mark of negation, as per Longobardi’s Fundamental Asymmetry Hypothesis (see Sect. 6.2.2.1 above). Under these assumptions, the negative morpheme in the type (ii) occurrence forms part of a discontinuous expression of negation, thereby contrasting with its type (iii) usage, in which it is not involved in negation (in the strict sense) at all. In the present case, however, the NQ (i.e. the N-word) is preverbal, meaning that there are no solid conceptual grounds for assuming that the distinction between types (ii) and (iii) is actually a real one. That is to say, the position of the N-word in the medieval string ‘N-word + no(n) + VFIN’ should in principle suffice to ensure a negative interpretation of the clause, implying that no(n) in this instance is structurally redundant. Matters would of course be different if Old Spanish had not been a language in which basic clausal negation was realized in a position to the left of the finite verb, as it is in the modern language. However, to restate the point made at the end of Sect. 6.3.1, in terms of this fundamental parametric setting, Old Spanish and Modern Spanish are identical. Given the foregoing train of thought, it can be wondered whether there is anything which preverbal N-words/NQs have in common with expletive negation triggers. In fact, the two types of item coincide in very important way. To see this, consider the characterization of expletive

232 

I. E. Mackenzie

negation assumed in (33), according to which expletive no(n) is a phonological or semantically empty reflex of a containing polarity context. Now polarity contexts are simply those contexts in which polarity items (PIs) can occur or, to use the favoured term, are ‘licensed’. The finding, reported in Sect. 6.2.2.2, that a postverbal N-word in the scope of a preverbal NQ is necessarily interpreted as a PI is therefore significant, given that it implies that preverbal NQs create polarity contexts. The discussion in Sect. 6.2.2.2 related to modern Spanish, but exactly the same principle holds in Old Spanish, the tendency for preverbal N-words to be accompanied by the negative morpheme no(n) having no effect in this regard. For example, while the preverbal N-words in (62) and (63) below are obviously NQs, the postverbal N-words ninguno and nunca can only mean ‘anyone’ and ‘ever’ respectively, implying that they occur as PIs rather than as NQs.

(62) ca ellos ningun mal non faran a ninguno si non al qui lo fizo. (General estoria IV, fol. 23v) ‘For they will do no harm to anyone except to whoever does harm himself.’



(63) a este linage delos ebreos que uos ueedes. njnguna pestilencia non les uiene nunca que danno les tenga de tod en todo. (General estoria I, fol. 305v) ‘This race of the Hebrews that you see, no pestilence ever afflicts them which harms them in any way.’

In abstract terms, examples like (62) and (63) can be seen as realizations of the schema ‘NQ . . . no(n) . . . PI’, where ‘NQ’ and ‘PI’ represent, respectively, the preverbal and post-verbal N-words highlighted in bold in the examples. Obviously enough, the schema ‘NQ . . . no(n) . . . PI’ is a special instance of ‘N-word + no(n) + VFIN’, specifically the instance in which there is an N-word after VFIN. Accordingly, the point made earlier about the status of no(n) in the string ‘N-word + no(n) + VFIN’ applies also to its occurrence in ‘NQ . . . no(n) . . . PI’. That is to say, in both cases no(n) is structurally redundant. Keeping that in mind, we can turn to the basic schema realized in examples involving expletive negation, such as (58), which is now repeated below.

  Negation: Dispensing with the Clutter 



233

(58) les deffendio que non aorassen njnguna destas cosas en ydolo njn sin ydolo. (General estoria I, fol. 181v) ‘He forbad them to worship any of these things either made into an idol or not.’

Here the trigger for the occurrence of non is the verb form deffendio. If this item is represented by ‘α’, then (58) can be seen as instantiating the schema ‘α . . . no(n) . . . PI’. Given that no(n) here is also structurally redundant, the schema ‘α . . . no(n) . . . PI’ can be regarded as differing from the schema ‘NQ . . . no(n) . . . PI’ solely in terms of the item which licenses the PI. Notice, furthermore, that this analysis is independent of the actual occurrence of a PI. The PIs in the above examples and their associated schemata merely provide empirical confirmation that the relevant contexts are polarity contexts. Obviously enough, such contexts remain polarity contexts even when they do not explicitly host a polarity item. In its general format, then, the parallel is between ‘α . . . no(n)’ and ‘NQ . . . no(n)’, where α is any expletive negation trigger. In light of this parallel, there is no obvious reason – other than the force of tradition, perhaps – for not treating NQ and α in these schemata as essentially having the same effect, viz. creating a polarity context, of which the overt presence of no(n) is no more than an empty phonological reflex. This conclusion is of course entirely consonant with the point made earlier with respect to the taxonomy of no(n) uses summarized in (33), namely that the distinction between the (ii) type of use (negative concord) and the (iii) type (expletive negation) is difficult to maintain in the case when the NQ of the would-be negative concord structure occurs before the finite verb rather than after it. There is in fact a third consideration supporting this perspective. For if ‘N-word + no(n) + VFIN’ really did involve negative concord, it is perplexing that this specific pattern should have declined and then disappeared while the supposedly related pattern ‘no(n) + VFIN + N-word’ has enjoyed an unvarying, categorical rate of occurrence. Posner (1984: 13–16) and others have suggested that prescriptive influence is responsible for the loss in question, citing Nebrija’s implicit condemnation in his Gramática castellana of the practice of ‘putting two negations for one’ (fol. 51r).

234 

I. E. Mackenzie

However, this view has a number of rather obvious flaws. In the first place, if prescriptive influence aimed at stamping out ‘double negation’ (interpreted as single negation) was involved, it is odd that the offending structure should have been eliminated solely from the preverbal field. This is even more surprising in light of the fact that Nebrija’s oft-cited comment targets the pattern which survived rather than the pattern which actually disappeared, his specific example being No quiero nada (ibid.). Secondly, as Fig. 6.1 indicates, the process whereby no(n) was lost with preverbal N-words begins in the High Middle Ages and is already very well advanced by the mid-fifteenth century, whereas Nebrija’s grammar was published in 1492 and Valdés, another commentator cited as a possible prescriptive influence in this regard, published his Diálogo de la lengua in the 1530s. Chronologically, then, it does not seem possible for the relevant grammarians to have exercised any significant influence on the development in question. Finally, there is the intrinsic difficulty in seeing how the isolated comments of one or two grammarians could influence the behaviour of an entire speech community. These problems disappear if the no(n) in ‘N-word + no(n) + VFIN’ is analysed as having an expletive occurrence. Expletive no(n) has declined steadily since the High Middle Ages, disappearing in some contexts altogether, the complements of verbs of prohibition, doubt, denial etc. being a case in point. Under an expletive negation-based analysis, the disappearance of no(n) after preverbal N-words simply adds to the list of contexts from which expletive negation was lost and hence it does not require any explanation in itself. As regards the reason why expletive negation in general has been progressively eliminated, it is worth recalling that this type of no(n) has neither a semantic value nor a syntactic function. In the latter regard, it differs from more famous expletive elements such as the English subject-expression there, which is customarily analysed as being inserted to satisfy Chomsky’s EPP, a condition requiring clauses to have overt material in the structural subject position. We can therefore agree with Martins’s observation (2000: 214) that expletive negation per se is somewhat at variance with Minimalist assumptions about natural language.15 In particular, it is unusual to encounter wordlevel items which appear to contribute nothing to the structures in which they occur. The progressive loss of expletive negation which has taken

  Negation: Dispensing with the Clutter 

235

place in Spanish, as in Romance generally, may thus represent the pruning of an essentially parasitic element, a purging of extraneous linguistic clutter. To sum up: the loss of no(n) after preverbal N-words coincides to a large extent with the loss of expletive no(n) in complements of verbs of prohibition, doubt and denial. A common rate of decline across the two contexts cannot be demonstrated empirically, but this is likely to be a by-­product of the lexically compartmentalized distribution of the data, insufficiently coarse-grained definitions of contexts tending to undermine the principle of the Constant Rate Effect (see Sect. 3.3.2). Given the overlapping diachrony of the two processes, together with their common endpoint– i.e. loss of no(n) – it was examined whether they could be viewed as contextually-specific manifestations of a single underlying change. The proposal has been that they can, given that preverbal N-words create polarity contexts, exactly as do verbs in the relevant meaning class. Moreover, as it is defined here, expletive negation is no more than the redundant phonological reflex of a polarity context. Analysing the no(n) which follows preverbal N-words as having an expletive occurrence is also consistent with the Fundamental Asymmetry Hypothesis (FAH) of Longobardi 2017. According to the FAH, negative concord is an epiphenomenon that arises when an N-word has a postverbal occurrence in a language in which negation requires a formal mark of negation in a pre-­INFL position. Finally, the proposed analysis obviates the need to explain why preverbal “negative concord” was lost while the corresponding pattern in which the N-word is postverbal has remained entirely stable. The negative concord approach relies on the unlikely influence of prescriptivists whereas, under the proposal here, the loss of no(n) after preverbal N-words is seen as part of the wider decline of expletive negation.

6.5 Transient Surge in Preverbal Nada As was first observed and documented by Octavio de Toledo (2014), the Enlightenment period, together with adjacent centuries, is the locus of a big up-and-down quantitative movement in the fronting of the N-word nada in

236 

I. E. Mackenzie

its non-subject use. The structure in question is illustrated by the two examples below from the statesman and author Gaspar Melchor de Jovellanos.

(64) En cuanto a las operaciones de Wellesley, nada sabemos de oficio en detalle; (Letter from Jovellanos to Lord Holland, Seville, 21 May 1809) ‘Concerning Wellesley’s manoeuvres, officially we know nothing in detail.’



(65) Las primeras Cortes de nada tratarán primero que de arreglar la representación para las sucesivas. (Letter from Jovellanos to Lord Holland, Seville, 7 June 1809) ‘The first sessions of parliament will deal with nothing other than agreeing the representation at successive ones.’

Although the big increase (and subsequent decline) in nada-fronting is very much a post-medieval phenomenon, it is useful to mention it in this book as it intersects with a number of the themes that have been touched on, in particular the nature of ‘failed changes’. The findings reported by Octavio de Toledo are based on a large number of tokens (54,108) mined from the Real Academia Española’s Corpus Diacrónico del Español, with data points located at fifty year intervals. That is already an impressive empirical contribution. However, given that the phenomenon postdates the advent of printing, we can avail ourselves of the Google Books N-gram Spanish corpus made available by Brigham Young University (Davies 2011) to get a finer-grained picture based on an even larger number of tokens. Accordingly, Fig. 6.2 below provides a visualization of the relevant bell curve based on 4,170,175 tokens, with data points distributed primarily at decadal intervals (the corresponding numerical data can be found in Table A.9 in Appendix 1). The figure also includes logistic trendlines, shown as dotted lines, for the upward and downward trends.16 It will be recalled from Fig. 2.1 (Chap. 2) that objects and prepositional phrases generally had fronting rates of around 5% and 15% respectively at the end of the sixteenth century, having decreased from peak values in the fifteenth century. The data represented in Fig. 6.2 subsumes both object and PP uses of fronted nada. Thus the initial value shown in

  Negation: Dispensing with the Clutter 

237

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1600 1630 1660 1690 1720 1750 1780 1810 1840 1870 1900 1930 1960 1990 Fig. 6.2  Rise and decline of fronted nada (1600–2000)

the figure, viz. 12.5% in 1600, is likely to be within a few percentage points of the notional ‘background’ fronting rate which existed before the big increase set in. Accordingly, it is probably safe to regard Fig. 6.2 as capturing the fundamentals of the up-and-down event which affected preverbal nada. A key feature of this event is the peak in the rate of fronting, which occurs at the 1810 data point, where the rate is 67.3% (3703/5505). This in line with the findings in Octavio de Toledo (2014), according to which the peak is located in the 1751–1800 period, albeit with a higher percentage rate (76%; 1355/1780).17 Interestingly, the rate of the increase which precedes the peak in Fig. 6.2 is very similar to the rate of the decrease which follows the peak. This is partially obscured in the curve for the raw data by the fluctuations in the rate of fronting that characterize the seventeenth century and the first half of the eighteenth century, fluctuations which are a direct consequence of the fact that the sample size decreases the further back in time one goes. However, a close similarity in the rates of increase and decrease is indicated by the upward and downward trendlines shown in the figure, which are

238 

I. E. Mackenzie

Table 6.2  Rates of growth and decay in nada-fronting (1600–2000) Rising OR (per decade)

Percentage equivalent

Decaying OR (per decade)

Percentage equivalent

1.133

13.3

0.889

11.1

roughly symmetrical. This visual impression is confirmed by the rates of increase and decrease shown in Table 6.2, which are based on regression analysis of the data in Table A.9. The odds ratio (OR) for the upward trend is 1.133, equating to an increase of 13.3% per decade. The exactly equivalent rate of decline is given by the reciprocal of the OR for the increase (see Sect. 1.2.3), which in this case is 1/1.133 = 0.883, equating to a decay rate of 11.7% per decade. Self-evidently, this rate is very close to the observed rate of decay, viz. 11.1% per decade, corresponding to an OR of 0.889. The rise and decline of nada-fronting thus appear to constitute a symmetrical event from the quantitative perspective. Octavio de Toledo (2014) attributes the increase in the fronting of nada to a Latinizing influence, which condemned the “illogical” use of two negations and hence favoured a structure in which the N-word did not co-occur with the negative morpheme. While this analysis cannot be ruled out, it suffers from analogous problems to those mentioned in Sect. 6.4 in connection with Posner’s (1984) prescriptivism-based explanation for the loss of no(n) with preverbal N-words. Moreover, as was shown in Sect. 2.5, there was towards the end of the Middle Ages a similar, albeit less spectacular increase in fronting generally, in which prescriptivism or a dislike of supposedly illogical structures presumably cannot be a factor. In this latter case, the driver of the increase was assumed to be stylistic, given the well-established link between fronting to the preverbal field and such notions as emphasis and focus. In light of that event, and given that the grammar itself appears to be unaffected, it seems reasonable to treat the rise and subsequent decline of nada-fronting as a stylistic trend, the operation in question initially delivering greater expressivity but over time becoming hackneyed. From that perspective, nada-fronting is analogous to a cultural meme, whose propagation will decay as a consequence of the meme losing its novelty value.

  Negation: Dispensing with the Clutter 

239

As was mentioned in Sect. 2.5, the mathematical modelling known as Rumour Propagation Theory (Piqueira 2010) suggests that the propagation of a meme evolves as a symmetric bell curve just in the case in which the probability of an ‘ignorant’ becoming a ‘spreader’ is approximately equal to the probability of a spreader becoming a ‘stifler’. In terms of nada-fronting, the ignorants in Piqueira’s model correspond to those speakers who are susceptible to adopting the fronting tendency; the spreaders are those speakers who exhibit the tendency and who, for whatever reason (prestige, influence, social networks etc.), cause other s­peakers to adopt it; and the stiflers are speakers who have been exposed to the tendency but either failed to adopt it or have abandoned it. Accordingly, the symmetry manifested by the rise and decline of nada-fronting suggests that the underlying dynamic was one in which the rate of recruitment of new nada-fronters was, over the course of the relevant period, approximately equal to the rate at which nada-fronters joined the ‘stifler’ population. In this regard, the event in question comes close to being an ‘inherent failed change’ as identified in the work of Postma (2010, 2017), in the sense that the factor driving the increase was counterbalanced by a roughly equal opposing force. However, the equilibrium in question stems from the internal dynamics of the relevant speech community rather than from any structural constraint dooming the innovation to failure from the outset. That is to say, the relevant equilibrium does not in itself provide evidence for Postma’s (2010: 285) notion of ‘a change that includes from its very beginning the ingredient of failure’.

6.6 Conclusion Although the negation patterns of Old Spanish are superficially rather distinct from those which are found in modern Spanish, a deeper inspection, supported by relevant quantitative data, points towards a considerable degree of continuity. In terms of the three ‘core dimensions of parametrization’ proposed by Longobardi (2017), the transition from medieval Spanish to the modern language does not appear to involve any significant change. The locus of basic clausal negation has remained constant, as has the capability for the negative morpheme to express logical

240 

I. E. Mackenzie

negation and the status of the N-words as, fundamentally, negative quantifiers. Although their basic status is unaltered, the Spanish N-words have followed the general diachronic pattern in Romance, identified by Martins (2000), whereby such items have become progressively less ‘underspecified’. This implies that their ability to occur with a non-negative value in polarity contexts has become more restricted over time, making them more clearly identifiable as negative items. However, as is implied by the remarks just above, this process has not led to any change in the ­fundamental status of the N-words: a basic analysis of these items as negative quantifiers, albeit with a secondary capability to occur as polarity items, appears to be warranted from at least the thirteenth century. A further diachronic trend has seen the expletive use of the negative morpheme undergo a severe retrenchment, surviving now vestigially in only a small number of environments. This contrasts with medieval usage, in which expletive negation was both more frequent and enjoyed a wider syntactic distribution. Most noticeably, it was routinely triggered in the complements of verbs of prohibition, doubt, denial and avoidance. The process leading to the disappearance of no(n) from this latter context co-existed for several centuries with the process which brought about the disappearance of no(n) from clauses in which an N-word occupies a preverbal position. In view of this long period of overlap and given important structural similarities between these two uses of no(n), it has been proposed here that they represent contextually specific manifestations of the same phenomenon, viz. expletive negation. The chapter also examined the striking quantitative bell curve described by the evolution of nada-fronting in the Enlightenment and adjacent centuries. Unlike the other failed changes considered in this book, the rise and decline of nada-fronting has the symmetric profile required under Postma’s conceptualization of such events as reflecting the interaction of two approximately equal but opposing forces. However, both the growth in nada-fronting and its subsequent decay appear to be stylistic rather than structural, implying that the geometry of the bell curve reflects the internal dynamics of the propagating population rather than the operation of some overriding syntactic constraint.

  Negation: Dispensing with the Clutter 

241

Notes 1. Comparative structures should also be included, but they are omitted here due to an English-­specific ambiguity. This is discussed at note 3 below, after the relevant Spanish data have been introduced. 2. This also explains why almost can co-occur with free choice any, as can Spanish casi ‘almost’ with free choice cualquier(a): (i) Fred will do almost anything for money. (ii) Fred hará casi cualquier cosa por el dinero. As is pointed out in Sect. 5.3.3.1, free choice items are a type of universal quantifier. Accordingly, they implicitly reference a limit, whence the insertability of adverbs meaning ‘almost’. 3. Given examples like (i) below, English also appears to allow PIs in comparative structures: (i)

María sings better than anybody I know

In view of the discussion in Sect. 6.2.1, it may seem strange to observe that almost can be inserted into (i), yielding the perfectly grammatical sentence María sings better than almost anybody I know. However, the any of the latter sentence can be inferred to be free choice any rather than PI any. This any-related ambiguity is disambiguated in Spanish, free choice cualquier(a) being required if casi ‘almost’ is inserted: (ii) María canta mejor que casi cualquier persona que conozco. ‘María sings better than almost any person I know.’ 4. As regards the temporal adverb nunca, the casi-insertion test does not work for the ‘blanket denial’ type of use, exemplified in (i) below, as (ii) demonstrates: (i)

Nunca se lo dije/No se lo dije nunca. ‘I never said that to him.’

(ii) *Casi nunca se lo dije/*No se lo dije casi nunca. ‘*I almost never said that to him.’

242 

I. E. Mackenzie

The reason for the exclusion of casi is that this item can only be inserted into contexts which reference a continuous scale, whereas the above use of nunca implies a dichotomous choice: either the event happened or it did not. 5. As regards the absence of an asterisk in this particular English translation, see note 3 above. 6. According to an alternative approach, advanced by Espinal (1992, 2000) for Catalan, the expletive negation marker does have a negative property, but this is absorbed by the logical specification of the item that creates the expletive negation context, which in the case of (31) would be the comparative form mejor. 7. As noted by Poole (2011), they appear to be commoner in NavarroAragonese texts than they are in the Castilian ones. 8. All three of the tokens involve a form of ninguno. 9. A common assumption, given the pattern in (54), is that the distribution of medieval nunqua/nunca in relation to no(n) represents a continuation of the situation which existed in Classical Latin, in which numquam (like other negative words) did not co-occur with the negative morpheme. Under that account, however, one would also expect postverbal nunqua/nunca to have been incompatible with no(n) in Old Spanish, which is not the case. It is also interesting to note that one of the late Latin examples which appear to foreshadow the Romance pattern of negation, viz. the example below from Pope Saint Gregory I, actually involves the co-occurrence of numquam and non to the left of the finite verb: (i) reliquis numquam praeponi non debuit ‘He was never to be placed before others.’ (Greg. Magnus, Epist. 5,4; cited Molinelli 1988: 38) 10. Camus Bergareche (2006) also applies the term ‘negative concord’ (‘concordancia negativa’) to this medieval structure. However, given that he also appears to regard Old Spanish N-words as PIs rather than NQs, I assume that he uses the term ‘negative concord’ simply as a descriptive label, without assuming that the N-word in a negative concord structure is, by definition, an NQ. This descriptive use of the term is problematic, however, as it potentially conflates negative concord, as in e.g. Spanish no . . . nada, with negative polarity, as in English not . . . anything. 11. Espinal et al. (ibid.) observe that if ningú receives a special fall-rise intonation, (57) – and similar examples – can also be interpreted as expressing

  Negation: Dispensing with the Clutter 

243

a double negation, i.e. ‘Everybody ate dessert’ (= ‘Nobody did not eat dessert’). The same double negation reading does not appear to be available if the N-word is postverbal, further underlining the general asymmetry between ‘N-word + no(n) + VFIN’ and ‘no(n) + VFIN + N-word’. 12. The specific verbs examined here are the following: defender ‘forbid’, dudar ‘doubt’, embargar ‘prohibit’, estorbar ‘prevent’, evitar ‘avoid’, excusar ‘exempt’, guardar ‘refrain’, negar ‘deny’, prohibir ‘prohibit’ and vedar ‘forbid’, together with orthographic variants. 13. Interestingly, there appear to be twelve tokens of nadi in the Poema de mio Cid (although none of nadie), suggesting that, initially at least, nadi/nadie was commoner in verse than in prose. In the prose manuscripts of the early literary period, the usual vehicle for expressing the idea of ‘nobody’ or ‘no one’ was the masculine singular form of ninguno/nenguno, used as a pronoun. 14. Camus Bergareche (2006: 1197) also identifies the 1440–1460 period as being crucial in terms of the loss of no(n) with preverbal N-words. However, in contrast to what is shown here in Fig. 6.1, he also contends that usage before and after this period was close to being categorical, the omission of no(n) being practically non-existent before 1440 and its insertion being exceptional after 1460. While the abrupt change model envisaged by Camus Bergareche is not without precedent, the long-term, sigmoidal evolution captured in Fig. 6.1, manifested most obviously by the ninguno/nenguno curve, is more in keeping with what is known about syntactic change. 15. Martins herself (ibid.) finesses this issue by treating expletive negation as the overt manifestation of the [+ modal] head of a polarity phrase. However, to the extent that the proposed head appears to be required solely to account for expletive negation, the analysis leaves itself open to the accusation of being ad hoc. 16. The parameters for the two trendlines, estimated by logistic regression, are as follows. Upward trendline: intercept (−1.9158), slope (0.1249); downward trendline: intercept (0.7921), slope (−0.118). 17. For the previous period, 1701–1750, Octavio de Toledo proposes two possible rates, viz. 59% and 78%, the latter including a large number of tokens from Feijoo, which appear to skew the data. Given the findings reported here, the lower of the two rates appears to be the more reliable of the two.

244 

I. E. Mackenzie

References Bosque, Ignacio. 1980. Sobre la negación. Madrid: Cátedra. Camus Bergareche, Bruno. 1986. Cronología y extensión de un cambio en la expresión de la negación en español. Revista de Filología de la Universidad de La Laguna 5: 111–122. ———. 2006. La expresión de la negación. In Sintaxis histórica del español, vol. 1, tomo 2, primera parte, ed. Concepción Company, 1165–1252. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. Davies, Mark. 2011. Google books (Spanish) corpus (45 billion words, 1500–2000). Available online at http://googlebooks.byu.edu/ Español Echevarría, Manuel. 1994. A Typology for NPI-Licensing. Unpublished manuscript. Los Angeles: University of California. Espinal, M.  Teresa. 1992. Expletive negation and logical absorption. The Linguistic Review 9 (4): 333–358. ———. 2000. On the semantic status of N-words in Catalan and Spanish. Lingua 110: 557–580. Espinal, M. Teresa, Susagna Tubau, Joan Borràs-Comes, and Pilar Prieto. 2016. Double negation in Catalan and Spanish. Interaction between syntax and prosody. In Negation and polarity: Experimental perspectives, ed. Pierre Larrivée and Chungmin Lee, 145–176. Cham: Springer. Haegeman, Liliane, and Raffaella Zanuttini. 1991. Negative heads and the NEG criterion. The Linguistic Review 8: 233–251. Laka, Itziar. 1990. Negation in syntax: On the nature of functional categories and projections. Doctoral dissertation, MIT, Cambridge, MA. Linebarger, Marcia. 1980. The grammar of negative polarity. Doctoral dissertation, MIT. Longobardi, Giuseppe. 2017. Theory and experiment in parametric minimalism: The case of Romance negation. Revista Linguí∫tica 13 (2): 108–157. Martins, Ana Maria. 2000. Polarity items in Romance: Underspecification and lexical change. In Diachronic syntax: Models and mechanisms, ed. Susan Pintzuk et al., 191–219. Oxford: Oxford University Press. Molinelli, Piera. 1988. Fenomeni della negazione dal latino all’italiano. Florence: La Nuova Editrice. Octavio de Toledo, Álvaro S. 2014. Entre gramaticalización, estructura informativa y tradiciones discursivas: algo más sobre nada. In Procesos de gramaticalización en la historia del español, ed. José Luis Girón Alconchel and Daniel Sáez Rivera, 263–319. Frankfurt: Iberoamericana/Vervuert.

  Negation: Dispensing with the Clutter 

245

Piqueira, José Roberto. 2010. Rumor propagation model: An equilibrium study. Mathematical Problems in Engineering, 631357. https://www.hindawi.com/ journals/mpe/2010/631357/ Poole, Geoffrey. 2011. Focus and the development of N-words in Spanish. In Romance languages and linguistic theory 2009. Selected papers from ‘Going Romance’ (Nice 2009), ed. Janine Berns et al., 291–303. Amsterdam: John Benjamins. Posner, Rebecca. 1984. Double negatives, negative polarity and negative incorporation in Romance: A historical and comparative view. Transactions of the Philological Society 82 (1): 1–26. Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. ———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Sánchez López, Cristina. 1999. La negación. In Gramática descriptiva de la lengua española, vol. 2: Las construcciones sintácticas fundamentales. Relaciones temporales, aspectuales y modales, ed. Violeta Demonte and Ignacio Bosque, 2561–2634. Madrid: Espasa Calpe. Suñer, Margarita. 1995. Negative elements, island effects and resumptive no. The Linguistic Review 12: 233–273.

7 Conclusion: Change and Continuity

This book has examined a range of syntactic phenomena in Old Spanish, with a particular focus on structures that have an important theoretical dimension or which present interesting patterns of change. The findings fall into two categories, those which relate to the more general issue of how syntactic change is manifested quantitatively and those which relate specifically to our understanding of Old Spanish and how it evolved.

7.1 F indings Relating to Quantitative Evolution Generally As regards the first category, one important conclusion to be drawn is that while Old Spanish had its fair share of failed changes, the data point strongly towards an accidental model of failure rather than the ‘inherent’ model propounded by Postma (2010, 2017). In the latter model, the “change” is seen as being doomed from the start, due to the fact that the innovative structure is incompatible with some general syntactic constraint. The resultant bell curve  – by hypothesis a symmetrical one  – is then analysed as

© The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9_7

247

248 

I. E. Mackenzie

being approximately proportional to the first derivative of the logistic curve described by the successful change of which the failure is assumed to be a by-product. None of the failed changes identified in the book appear to fit this type of pattern. This is strikingly demonstrated by the history of clitic linearization, where the successful advance of proclisis in tensed main clauses was accompanied by three failed changes in other contexts. The corresponding bell curves, according to the data adduced, are all asymmetric and they bear no relation to the first derivative of the logistic function that models the successful advance of proclisis in main clauses. In addition, rather than indicating the inherent unviability of the advancing structure, the failures appear to reflect the operation of factors unleashed by system-internal developments. Thus in the most spectacular of the clitic-related failures, that which took place in infinitival clauses introduced by a preposition, there is no Spanish-external reason why the upward trend in proclisis could not have been successful; modern French, for example, illustrates precisely that outcome. Instead, the reversal of the proclitic trend in this particular context can be linked to a reanalysis of infinitival clauses more generally as disallowing proclisis, a reanalysis that was caused, somewhat ironically, by the loss of enclisis in structures that brought a clitic into an immediate pre-infinitive position. Other accidental failures derive their initial impetus from stylistic or expressive needs, which cause a particular structure to advance quantitatively for a sustained period, only for the growth trend to be subsequently reversed by some unrelated event. A case in point is the prefixation of the definite article to the determiner and pronoun qual, an operation made possible by the syntax of the Old Spanish DP but whose late medieval quantitative increase was driven by the enhanced functionality of the resultant structure vis-à-vis the earlier bare qual. This upward trend was reversed from about 1460, producing the characteristic bell curve of a failed change. However, the curtailment of the growth phase was essentially fortuitous, being no more than a by-product of the more general loss of the syntax which enabled the double determiner pattern, a loss expressed most saliently by the decline in possessive structures like la mi mugier or un su hermano. Tokens of el cual, la cual etc. are of course still capable of being gener-

  Conclusion: Change and Continuity 

249

ated, but these no longer involve the iteration of D-elements, the erstwhile definite article having been reanalysed as a purely formal component of a compound relative pronoun. Stylistic factors were also crucial in the one genuinely symmetric failure discussed in the book, viz. the rise and decline of nada-fronting. This up-and-down event actually occurred outside of the period which is the book’s main focus, but its relevance to some of the book’s core themes is obvious. Using Google N-gram technology, it can be shown that there is a striking peak in the preverbal placement of non-subject nada at close to 70% in the early 1800s. It turns out that the growth trend to the left of the peak and the decay trend to the peak’s right are very close to being exact converses of one another, implying that the overall event had the symmetric profile associated with inherent failures. However, the symmetric bell curve can also be seen as being a reflex of random population dynamics. Piqueira (2010) shows that in rumour propagation, the evolution of the spreader population evolves as a bell curve if the conversion rate from ‘ignorant’ to ‘spreader’ is roughly equal to the conversion rate from ‘spreader’ to ‘stifler’. Substituting ‘nada-fronters’ for ‘spreaders’, the same model would predict a symmetrical up-and-down evolution for nada-fronting whenever the rate of recruitment to the fronting paradigm was roughly equivalent to the rate of its abandonment. However, the mechanism delivering this result is non-linguistic. Thus although the forces whose interaction delivers the bell curve are (roughly) equal and opposite, as is predicted by the inherent failure model, this is a random circumstance and not the result of a general syntactic constraint condemning nada-fronting to failure from the outset. Overall, then, the Spanish data examined here do not provide any strong evidence for the existence of inherently failing changes. Instead, the up-and-down quantitative events that have occurred are largely aleatory. When structural factors are involved, the rising and declining flanks of the relevant bell curves represent unrelated changes which happen to abut in diachronic space. When the growth and decay are purely stylistic, expressive or functional, the quantitative evolution reflects the interaction of population dynamics. The latter forces can in principle deliver a symmetrical bell curve, but the equilibrium required for this is not in any way predetermined.

250 

I. E. Mackenzie

Turning now to quantitative change itself, the data presented here provide robust support for the hypothesis of cross-contextual uniformity in the rate of syntactic change. Constant Rate Effects (CREs) are apparent in a variety of changes that affected Old Spanish, including the late medieval increase in V2-style fronting, the general advance of pronominal proclisis and the loss of DP recursion. Among other things, the CRE principle appears to be independent of the unit of measurement by which the synchronic frequency of the advancing structure is gauged. Thus while it applies in the case in which synchronic use is expressed as a percentage or probability, as is expected, it also applies when usage is measured in terms of absolute token counts (normalized to the volume of discourse). This latter measure was required in the case mentioned earlier of prefixation of the definite article to qual, owing to the impossibility of defining a closed set of competitor constructions. The diachronic growth rates for prefixation across the two contexts of qual qua determiner and qual qua pronoun turned out to be almost identical, exactly as would be expected under a standard operationalization of the Constant Rate Effect. The cases examined here also confirm that what Kauhanen and Walkden (2017: 1.3) call ‘time separation between contexts’ is an integral component of syntactic change. A striking example of this can be seen in the loss of determiners with prenominal possessives, which began in the early fifteenth century. At the start of the change, ‘D + possessive + NP’ was much more frequent (in relative terms) when D was indefinite than when it was definite. Thus although the structure decayed at the same rate across the two contexts, the initial disparity caused the decline to always be more advanced in the definite case than in the indefinite one. This led inexorably to a situation in which definite DPs like la mi mugier were close to extinction even as the corresponding indefinite structure, un su hermano etc., was still productive. It can be surmised that the survival of the latter pattern in modern Guatemalan Spanish and other former colonial varieties is an oblique reflex of that early modern state of affairs. In light of the time separation phenomenon, Kauhanen and Walkden (2017: 2.2) propose a remodelling of the CRE in which each of the contexts affected by a change is assigned a specific ‘production bias’ which either favours or inhibits what they conceptualize as the incoming ­grammatical system. From that perspective, the CRE is epiphenomenal,

  Conclusion: Change and Continuity 

251

a mathematical by-product of grammar competition (Yang 2000) modulated by context-specific production biases. However, the findings here indicate that an observable CRE cannot always be linked to grammar competition. For example, while the advance of pronominal proclisis manifests a strong CRE across twelve data points (1260–1480) in the two main contexts affected, in one of these contexts, that of post-­ prepositional infinitival clauses, the change eventually fails and enclisis replaces proclisis. Despite the existence of a CRE, this bifurcating pattern of change obviously cannot be modelled in terms of an incoming, ‘proclitic’ grammar which displaces an existing ‘enclitic’ one. The advance of proclisis must therefore be treated separately in each context, as in Kroch’s original conception of the CRE. Time separation is no more than a reflex of the fact that the baseline frequencies at the start of a change may be unequal across contexts. Just as a common rate of interest will never level the difference between non-­ identical investments, so a uniform rate of change cannot, in itself, eliminate cross-contextual disparities in synchronic usage. That syntactic changes do not necessarily start from a uniform value, usually assumed to be zero (cf. Kroch 1989: 204–205), conflicts with the popular notion that syntactic change involves the emergence of an entirely new structure, which over time replaces an existing one. However, the data examined here suggest that this latter conception references the exception rather than the norm. In the majority of the cases looked at, the change involves the disruption of stable variation rather than the introduction into the system of a new structure. For example, the above-mentioned loss of determiners with prenominal possessives occurred after a long period during which variation between ‘D + possessive + NP’ and its competitor structures was in a steady state. In general, then, the findings here are consistent with the classic conception of the CRE, which assumes common rates of change but divergent baseline values, an approach which Kauhanen and Walkden (2017: 4.6) label ‘same slopes, different intercepts’. It must be said, however, that CREs appear to be subject to a granularity restriction, in the sense that the contexts over which a uniform rate of change holds need to be defined fairly broadly. The rich data from clitic linearization provide an i­nteresting demonstration of this phenomenon. For while the advance of proclisis has a uniform rate across the two broad contexts alluded to above, more narrowly

252 

I. E. Mackenzie

circumscribed contexts, defined with respect to factors which are independently known to affect clitic placement, manifest a wide degree of variation in the rate of change. Thus in a sub-classification of main clauses according to the element preceding the finite verb, it was found that the odds of proclisis grew at rates ranging from below 10% per decade (after an adverbial) to over 24% (after a coordinating conjunction). An analogous phenomenon is apparent in relation to the syntax of negation, where the loss of no(n) with preverbal N-words was found to proceed at a different rate depending on which particular N-word was involved. Rather than calling into question the principle of the CRE, disparities of this kind suggest that narrowly defined micro-contexts may be susceptible to interference. In the case of Spanish proclisis, for example, it seems likely that those micro-contexts which were lagging behind main clauses overall (due to the time separation phenomenon) underwent a quantitative adjustment towards the end of the change process, bringing them into line with the situation in the containing context. This can be seen with the postcoordinator micro-context, in which the rate of proclisis was barely above 17% in the early sixteenth century while the rate in main clauses overall stood at close to 72%. From this point onwards the diachronic curve for the post-coordinator context departs from its existing trajectory and becomes appreciably steeper, suggesting that speakers accommodated their usage in the micro-context in view of what they correctly perceived to be the dominant option in the containing macro-­context. If this type of distorting effect is a general one, it may be the case that, for CRE purposes, contextual dissection is subject to a law of diminishing returns.

7.2 F indings Relating to the History of Old Spanish As regards the second category of finding mentioned at the start of this chapter, a major theme of the analyses presented here is the continuity in syntax between Old Spanish and modern Spanish. Superficially, the medieval language looks very different to its modern counterpart, but very often the differences turn out not to involve important dimensions of parametric classification.

  Conclusion: Change and Continuity 

253

This is strikingly highlighted in the area of word order, where the relative abundance of what Lapesa (1981: 217) called the ‘régimen antepuesto’ has led many to assume that Old Spanish was a V2 language, implying that in this regard it had more in common with continental Germanic and mainland Scandinavian than with modern Ibero-Romance. Detailed examination of medieval prose manuscripts indicates that this perspective overestimates the extent to which Old Spanish differed from its modern counterpart. The structures in question are indeed V2 structures, but only in the trivial sense in which Ā-movement to the CP edge, accompanied by obligatory verb–subject inversion, by definition involves V2 syntax. Isolated or restricted V2 syntax of this latter type is common in many languages which are not V2 languages, including modern English and modern Spanish. It is certainly true that Old Spanish made greater use of this type of syntax than its modern counterpart does, but this is a matter of degree rather than kind. What gives the medieval situation its peculiar character is the fact that restricted V2 syntax – used more intensively than in the modern language – co-existed with a fortuitous or ‘accidental’ type of V2 linear order, the latter being a by-product of the prevalence of VS(O) in the medieval language. Analogously to what happens in modern Arabic, VS(O) in Old Spanish when combined with a preverbal constituent delivered a linear order in which the finite verb occupied the second position. However, such structures fail the basic tests of true V2 syntax. Firstly, they were not structures in which verb–subject inversion was obligatory, given that they were also compatible with preverbal subjects, positioned between the first-position element and the finite verb, ­delivering V3 word order. Secondly, the clitic linearization for such structures was enclisis rather than proclisis, the latter being the linearization required in Old Spanish in true V2 structures. In short, the assumption that Old Spanish was subject to a generalized V2 constraint, as is exhibited by many varieties of modern Germanic, results from undue focus on a minority structure allied to the misreading of another, superficially similar construction. A further area in which the medieval grammar is more similar to the modern one than is commonly thought is negation. Here the consistent use of the negative morpheme no(n) in conjunction with preverbal

254 

I. E. Mackenzie

N-words (other than nunqua/nunca), combined with the possibility of using N-words as polarity items in conditional and related constructions, gives the impression that the Old Spanish negation system was fundamentally different from its modern counterpart. However, in terms of the core axes of parametrization – the locus of clausal negation, the possibility for the negative morpheme to express logical negation and the status of the N-words  – there have been no actual changes since at least the mid-thirteenth century. The status of the N-words is the issue that offers the best opportunity for highlighting differences. However, as with the V2 structures mentioned above, the differences in this regard between the medieval and modern instantiations of the language are quantitative rather than qualitative. That is to say, rather than evincing a change in their syntactic category – from out-and-out polarity items to NQs – the diachronic evolution of the N-words consists in their becoming progressively more specialized as NQs with the passage of time, without (as yet) losing their ability to also be used as PIs. Separately, but in the same spirit, it was noted that analysing the co-­ occurrence of the negative morpheme with preverbal N-words as a now defunct, preverbal type of negative concord is problematic from various viewpoints. In particular, the conceptual basis for negative concord in Spanish-type languages dissolves if the N-word is preverbal. This objection might not apply if the Old Spanish negation system differed from its modern counterpart in terms of the core dimensions of parametrization; however, as was just noted, this is not the case. There is, moreover, a stark contrast between the diachronic instability of the alleged preverbal pattern of negative concord and the extreme inertness of the pattern in which the N-word is post-verbal. It turns out, in fact, that the use of no(n) with preverbal N-words has more in common with expletive negation. For both expletive negation triggers and preverbal N-words create polarity contexts and in neither case can a syntactic or semantic function be identified for the negative morpheme, which thus appears to be no more than a phonological reflex of the polarity context. Diachronically, too, the phenomena appear to parallel one another, expletive negation surviving only residually in the modern language and having been lost entirely in some of the contexts in which it previously enjoyed considerable vitality.

  Conclusion: Change and Continuity 

255

In contrast to the foregoing cases, the DP constitutes an important locus of change over the course of the Middle Ages. Unlike modern Spanish, Old Spanish allowed two or more D elements be merged recursively, an operation which delivered the compound possessives mentioned above, as well as double article structures reminiscent of modern French superlatives and the linear strings from which the modern relative pronouns el que and el cual were reanalysed. The loss of DP recursion, reflected in the quantitative data from the early decades of the fifteenth century, represents an important parametric switch, which has received less attention than it merits, perhaps owing to the tendency to consider each of its surface manifestations in isolation. Also in the DP, we see a shift away from a medieval system that was rich in the occurrence of free relatives, such as qual marido escogiese ‘whichever husband she would choose’ and qui tanxiere omne muerto ‘whoever touches a corpse’, to a modern-looking one based more consistently on headed relatives. An important by-product of this process is the emergence of the series of free choice quantifiers characterized morphologically by the suffix -quier(a). The advent of these items was made possible by the apparent immunity of Old Spanish free relatives to Chomsky and Lasnik’s doublyfilled Comp filter, together with the medieval tendency to insert the intensifying particle quier into such structures. Free relatives involving both quier and a doubly-filled Comp violation were reanalysed as headed relatives, the overt complementizer que being reinterpreted as a relativizing element and the material to its left as the antecedent of a relative clause. The other area of the grammar which underwent significant change was clitic linearization. In the system which had come into existence in the High Middle Ages, the placement of weak pronouns was determined by syntax. The principal constraint in this so-called Tobler–Mussafia system was that weak pronouns could not occur immediately after a CP boundary, enclisis arising whenever proclisis would violate this restriction. Over time, an inherently proclitic weak pronoun out-competed the older Tobler–Mussafia clitic, although reversals of this process occurred in (post-prepositional) infinitival clauses and with gerunds and positive imperatives. The combined effect of these developments was the loss of the syntax-based principle for regulating clitic linearization and its replacement by one based primarily on verbal morphology.

256 

I. E. Mackenzie

In broad terms, then, that the structure of the DP and clitic linearization are the two principal loci of syntactic change in Old Spanish. Constituent order and negation, though contributing in important ways to the overall character of medieval Spanish, do not in themselves show any major structural reorganization.

References Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Lapesa, Rafael. 1981. Historia de la lengua española. 9th ed. Madrid: Gredos. Piqueira, José Roberto. 2010. Rumor propagation model: An equilibrium study. Mathematical Problems in Engineering, 631357. https://www.hindawi.com/ journals/mpe/2010/631357/ Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. ———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Yang, Charles D. 2000. Internal and external forces in language change. Language Variation and Change 12: 231–250.

Appendix

Appendix 1: Data Tables

© The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9

257

258 Appendix Table A.1  Evolving rates of fronting (1250–1609)

Period

Objecta

PPa

Adverb of manner or quantity

1250–1269

5.9% (59/1000) 6.2% (62/1000) 5.1% (51/1000) 5.4% (54/1000) 5.3% (53/1000) 4.8% (48/1000) 3.9% (39/1000) 4.1% (41/1000) 4.6% (46/1000) 6.7% (67/1000) 6.0% (60/1000) 5.9% (59/1000) 5.9% (59/1000) 5.2% (52/1000) 3.6% (36/1000) 4.2% (42/1000) 3.9% (39/1000) 3.7% (37/1000)

8.4% (84/1000) 7.8% (78/1000) 8.8% (88/1000) 7.8% (78/1000) 9.6% (96/1000) 11.3% (113/1000) 13.5% (135/1000) 15.2% (152/1000) 16.8% (168/1000) 19.2% (192/1000) 18.9% (189/1000) 19.0% (190/1000) 18.7% (187/1000) 17.1% (171/1000) 16.4% (164/1000) 15.8% (158/1000) 13.7% (137/1000) 12.6% (126/1000)

2.8% (6/215) 2.8% (11/388) 3.4% (2/58) 4.3% (2/47) 2.4% (1/42) 5.5% (3/55) 6.3% (2/32) 7.5% (6/80) 8.2% (4/49) 7.4% (6/81) 16.5% (20/121) 16.3% (48/295) 17.9% (127/711) 15.8% (98/622) 13.0% (52/399) 12.4% (114/923) 7.9% (65/827) 7.9% (24/304)

1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

Predicative adjective or noun

Non-finite verb form

11.0% (226/2053) 11.4% (424/3704) 10.2% (70/684) 9.2% (51/552) 8.1% (40/491) 7.3% (48/657) 6.3% (25/400) 8.2% (101/1232) 9.7% (78/807) 9.1% (107/1171) 16.3% (278/1703) 11.7% (308/2622) 10.2% (269/2634) 10.3% (214/2078) 10.1% (145/1432) 8.6% (265/3086) 8.7% (241/2766) 7.1% (78/1101)

1.6% (99/6190) 2.0% (221/11,225) 1.8% (52/2868) 1.3% (31/2388) 1.5% (30/1998) 1.2% (40/3363) 1.3% (30/2326) 2.1% (122/5789) 3.4% (98/2888) 4.0% (154/3885) 3.8% (200/5252) 3.3% (250/7571) 3.1% (229/7453) 1.2% (74/6201) 0.7% (26/3852) 0.6% (54/9030) 0.8% (66/8194) 1.2% (37/3023)

Rates for objects and PPs based on random samples (1000 tokens per time period)

a

 Appendix 

259

Table A.2  Evolving rates of proclisis (1250–1609) (excluding negative contexts) Period

Main clauses

Post-prepositional infinitivals

Imperatives

Gerundials

1250–1269

24.7% (2491/10,084) 24.5% (4457/18,191) 26.1% (800/3067) 32.1% (698/2173) 37.7% (538/1428) 39.7% (511/1288) 59.7% (602/1009) 45.6% (3465/7598) 59.9% (1366/2281) 83.3% (3253/3905) 57.5% (4431/7705) 77.3% (7922/10,249) 71.7% (8683/12,111) 85.3% (12,082/14,164) 97.2% (10,674/10,982) 90.5% (22,836/25,233) 92.5% (21,031/22,736) 97.3% (8254/8483)

34.2% (130/380) 19.1% (326/1711) 20.2% (104/515) 26.2% (85/325) 29.2% (90/308) 41.8% (69/165) 33.5% (52/155) 75.6% (559/739) 86.0% (361/420) 78.5% (547/697) 70.1% (368/525) 73.0% (1194/1635) 45.9% (1243/2708) 43.0% (1276/2970) 24.5% (441/1797) 15.3% (667/4348) 4.6% (131/2825) 0.6% (9/1452)

10.5% (14/133) 10.8% (27/249) 5.9% (3/51) 6.8% (3/44) 4.2% (2/48) 2.8% (1/36) 3.7% (1/27) 9.6% (17/177) 8.3% (6/72) 2.1% (2/94) 4.2% (4/96) 4.8% (7/147) 8.3% (12/145) 30.3% (46/152) 47.9% (45/94) 38.3% (70/183) 42.6% (52/122) 22.5% (9/40)

0.0% (0/403) 0.0% (0/744) 0.0% (0/142) 0.0% (0/112) 0.0% (0/107) 0.0% (0/90) 0.0% (0/83) 0.0% (0/291) 7.1% (6/85) 9.5% (60/633) 7.1% (51/722) 2.4% (8/327) 0.7% (6/809) 1.6% (14/870) 2.3% (28/1199) 1.2% (49/4124) 1.3% (61/4664) 0.0% (0/2776)

1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

260 Appendix Table A.3  Evolving rates of proclisis in main-clausal subcontexts (1250–1609) Period

Post-adverbial

Post-subject

Post-pause

Post-coordinator

1250–1269

50.3% (91/181) 51.2% (168/328) 56.0% (28/50) 59.4% (19/32) 66.7% (18/27) 70.8% (17/24) 88.2% (30/34) 64.7% (123/190) 81.1% (77/95) 93.1% (121/130) 81.3% (248/305) 77.6% (152/196) 78.6% (320/407) 90.6% (403/445) 98.3% (290/295) 93.5% (686/734) 94.4% (728/771) 99.1% (320/323)

30.8% (149/483) 32.7% (289/885) 34.0% (55/162) 42.2% (49/116) 50.8% (60/118) 59.0% (46/78) 66.7% (72/108) 57.0% (263/461) 79.0% (94/119) 97.0% (194/200) 97.0% (387/399) 93.5% (1252/1339) 90.4% (501/554) 96.3% (875/909) 99.6% (1169/1174) 98.1% (1570/1600) 98.7% (1471/1490) 98.7% (543/550)

1.5% (19/1261) 1.6% (36/2242) 1.4% (4/294) 2.3% (4/173) 2.8% (3/109) 4.3% (3/69) 13.7% (10/73) 1.4% (2/139) 22.0% (18/82) 39.3% (33/84) 2.1% (5/237) 12.7% (21/166) 9.4% (88/937) 53.9% (496/921) 85.3% (464/544) 72.2% (925/1282) 79.7% 1026/1288) 95.1% (731/769)

1.9% (84/4435) 2.2% (176/7986) 1.2% (17/1450) 1.8% (16/882) 2.6% (20/762) 4.0% (21/525) 6.3% (21/333) 6.1% (105/1724) 13.4% (68/507) 30.3% (117/386) 12.7% (101/797) 24.1% (358/1488) 17.2% (238/1380) 55.0% (341/620) 86.4% (152/176) 73.8% (375/508) 81.2% (406/500) 88.9% (193/217)

1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

 Appendix 

261

Table A.4  Differential evolution of ‘clitic–VINF’ according to contexta (1250–1609) Period

Preposition + clitic + VINF

VFIN + clitic + VINF

1250–1269

34.6% (133/384) 20.5% (360/1752) 22.3% (118/530) 26.6% (87/327) 29.7% (92/310) 42.5% (71/167) 34.4% (54/157) 75.8% (573/756) 86.1% (365/424) 78.2% (555/710) 70.0% (375/536) 73.4% (1228/1673) 46.5% (1307/2809) 44.1% (1372/3109) 25.6% (479/1871) 16.3% (748/4577) 4.7% (140/2987) 0.6% (9/1536)

22.7% (117/516) 22.8% (554/2431) 25.9% (190/735) 22.2% (97/436) 18.8% (79/420) 15.8% (35/221) 22.3% (51/229) 3.2% (33/1034) 2.1% (12/561) 7.7% (74/962) 7.8% (57/727) 4.7% (109/2318) 7.9% (298/3774) 3.7% (154/4150) 1.1% (27/2476) 2.5% (159/6351) 2.1% (84/4019) 0.9% (18/2048)

1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609 a

Includes negative tokens

262 Appendix Table A.5  Interpolation versus proclisis (1250–1609) Interpolation

Proclisis

Period

Finite

Infinitival

Finite

Infinitival

1250–1269

7.9% (1053/13,332) 6.6% (1618/24,521) 8.1% (389/4800) 6.8% (216/3174) 5.7% (163/2860) 4.8% (145/3026) 4.1% (122/2981) 4.4% (604/13,739) 1.6% (74/4500) 0.8% (74/8889) 2.6% (349/13,336) 2.3% (630/27,269) 0.6% (130/22,484) 0.3% (94/31,278) 0.0% (4/26,881) 0.0% (2/54,425) 0.0% (1/45,080) 0.0% (0/18,219)

4.3% (6/139) 13.9% (58/418) 1.7% (2/120) 4.4% (4/91) 2.1% (2/94) 4.1% (3/74) 5.3% (3/57) 4.2% (25/598) 4.2% (16/381) 4.6% (27/582) 6.0% (24/399) 3.9% (50/1278) 3.7% (50/1357) 0.6% (8/1380) 0.2% (1/480) 0.7% (5/753) 2.1% (3/143) 0.0% (0/9)

61.3% (12,279/20,032) 62.5% (22,903/36,623) 65.6% (4411/6720) 59.2% (2958/4997) 58.6% (2697/4602) 76.6% (2881/3759) 87.6% (2859/3264) 76.1% (13,135/17,268) 82.9% (4426/5341) 93.1% (8815/9468) 79.9% (12,987/16,260) 92.0% (26,639/28,969) 86.7% (22,354/25,785) 93.3% (31,184/33,428) 98.9% (26,877/27,184) 95.2% (54,423/57,153) 96.0% (45,079/46,967) 98.7% (18,219/18,450)

34.6% (133/384) 20.5% (360/1752) 22.3% (118/530) 26.6% (87/327) 29.7% (92/310) 42.5% (71/167) 34.4% (54/157) 75.8% (573/756) 86.1% (365/424) 78.2% (555/710) 70.0% (375/536) 73.4% (1228/1673) 46.5% (1307/2809) 44.1% (1372/3109) 25.6% (479/1871) 16.3% (748/4577) 4.7% (140/2987) 0.6% (9/1536)

1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

 Appendix 

263

Table A.6 Evolving rates of occurrence of three recursive DP structures (1250–1609)

Period 1250–1269 1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

Definite article + possessive + NP

Indefinite determiner + possessive + NP

Demonstrative + possessive + NP

17.2% (2269/13,194) 17.4% (4301/24,720) 15.3% (696/4552) 15.9% (557/3502) 14.3% (458/3204) 13.9% (579/4169) 13.6% (173/1275) 17.8% (993/5577) 9.8% (455/4647) 25.2% (1385/5495) 12.7% (828/6516) 13.2% (2378/18,017) 7.1% (840/11,835) 7.0% (1243/17,756) 0.5% (53/10,535) 0.3% (68/22,550) 0.1% (20/19,548) 0.1% (7/7412)

96.5% (55/57) 95.1% (98/103) 95.8% (23/24) 94.4% (17/18) 95.2% (20/21) 97.8% (44/45) 92.9% (13/14) 94.9% (93/98) 95.5% (42/44) 86.8% (46/53) 54.5% (12/22) 69.1% (105/152) 69.4% (43/62) 52.6% (71/135) 34.8% (32/92) 27.4% (62/226) 28.6% (63/220) 24.5% (24/98)

99.4% (163/164) 99.7% (323/324) 96.8% (60/62) 93.8% (45/48) 93.6% (44/47) 96.8% (60/62) 80.0% (4/5) 98.3% (113/115) 96.9% (31/32) 66.7% (8/12) 97.6% (40/41) 96.6% (85/88) 95.6% (151/158) 96.7% (59/61) 96.0% (97/101) 89.3% (175/196) 79.6% (133/167) 86.8% (33/38)

264 Appendix Table A.7 Three wh-structures: tokens per million words (1250–1609) Period

Definite article + pronominal qual

Definite D + NP + definite article + qual + NP article + relative clause

1250–1269 1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609

8.1 8.5 12.4 11.2 13.7 37.1 133.4 136.5 157.6 379.9 380.3 207.9 143.7 89.7 136.7 131.9 224.4 221.8

39.7 44.0 58.4 60.1 57.7 124.8 575.0 652.3 707.0 1623.3 2724.4 2485.6 2001.4 1486.6 1356.3 1338.2 1918.5 1463.8

49.8 52.4 37.2 29.3 35.6 25.2 14.3 11.2 26.8 25.1 14.3 14.1 19.4 21.6 12.5 12.1 10.6 13.5

Table A.8  Decline of no(n) in five contexts (1250–1609)

Period 1250–1269 1270–1289 1290–1309 1310–1329 1330–1349 1350–1369 1370–1389 1390–1409

Preverbal nenguno/ ninguno 98.6% (208/211) 98.8% (328/332) 98.4% (60/61) 100.0% (34/34) 100.0% (30/30) 97.2% (103/106) 93.8% (45/48) 93.2% (55/59)

Preverbal Preverbal Preverbal Complements of jamas/iamas nada nadie/nadi certain verbsa 89.7% (52/58) 89.8% (106/118) 86.4% (19/22) 88.2% (15/17) 81.8% (9/11) 76.2% (16/21) 64.3% (9/14) 37.8% (14/37) (continued)

 Appendix 

265

Table A.8 (continued)

Period 1410–1429 1430–1449 1450–1469 1470–1489 1490–1509 1510–1529 1530–1549 1550–1569 1570–1589 1590–1609 a

Preverbal nenguno/ ninguno

Preverbal Preverbal Preverbal Complements of jamas/iamas nada nadie/nadi certain verbsa

74.3% (26/35) 60.8% (79/130) 41.4% (53/128) 16.4% (47/287) 15.6% (19/122) 13.1% (26/199) 0.9% (2/221) 0.3% (1/287) 0.7% (1/146) 0.0% (0/80)

79.3% (111/140) 35.7% (5/14) 38.0% (30/79) 7.7% (6/78) 1.1% (2/186) 1.5% (1/66) 3.1% (4/130) 3.8% (4/106) 0.0% (0/67)

58.3% (21/36)

20.8% (5/24) 22.6% (7/31) 31.3% (5/16) 6.3% (1/16) 15.8% (6/38) 21.4% (6/28) 0.0% (0/40)

37.5% (9/24) 11.1% (2/18) 0.6% (1/155) 0.9% (2/218) 0.6% (1/163) 0.0% (0/50)

57.7% (41/71) 74.2% (23/31) 54.3% (19/35) 52.5% (21/40) 37.3% (22/59) 30.3% (10/33) 40.0% (6/15)

Verbs of prohibition, denial, doubt etc.

Table A.9  Evolving rates of nada-fronting (1600–2000) Period

Rate

Period

Rate

1600–1609 1610–1619 1620–1639 1640–1649 1660–1669 1670–1679 1680–1699 1700–1729 1730–1739 1740–1749 1750–1759 1760–1769 1770–1779 1780–1789 1790–1799 1800–1809 1810–1819 1820–1829

12.5% (2/16) 19.0% (4/21) 21.4% (3/14) 9.4% (6/64) 25.0% (5/20) 2.7% (2/74) 46.1% (41/89) 62.5% (10/16) 23.7% (56/236) 52.9% (55/104) 25.8% (104/403) 64.5% (729/1130) 61.2% (2815/4598) 61.6% (5240/8502) 60.0% (7063/11,764) 59.7% (5916/9910) 67.3% (3703/5505) 65.2% (9508/14,590)

1830–1839 1840–1849 1850–1859 1860–1869 1870–1879 1880–1889 1890–1899 1900–1909 1910–1919 1920–1929 1930–1939 1940–1949 1950–1959 1960–1969 1970–1979 1980–1989 1990–1999 2000–2009

62.6% (11,291/18,036) 58.3% (23,673/40,599) 58.8% (28,006/47,626) 51.5% (32,348/62,781) 51.6% (22,889/44,357) 48.5% (31,190/64,337) 47.1% (28,175/59,769) 44.7% (31,683/70,888) 45.7% (33,597/73,585) 41.2% (39,132/94,940) 36.9% (46,294/125,300) 36.1% (72,979/202,063) 33.8% (88,785/263,050) 28.3% (115,270/407,613) 24.5% (124,406/507,010) 23.3% (120,758/519,043) 16.2% (104,575/643,644) 11.8% (102,051/868,478)

266 Appendix

Appendix 2: Texts Included in the Main Corpus Abbreviations BL British Library BNE Biblioteca Nacional de España BNF Bibliothèque Nationale de France Esc. Escorial HSA Hispanic Society of America RAH Real Academia de la Historia

Thirteenth Century Cánones de Albateni (Paris: Arsenal 8322), Documentos castellanos de Alfonso X (various ms; CD-ROM prepared by Herrera et  al. 1999), Estoria de España I (Esc. Y-I-2), Estoria de España II (Esc. X-I-4), Formas e imágenes (Esc. h-I-16), Fuero Juzgo (HSA B2567), Fuero real (Esc. Z-III-16), General estoria I (BNE MSS/816), General estoria IV (Vatican Urb lat 539), General estoria V (Esc. I-I-2), Judizios de las estrellas (BNE MSS/3065), Lapidario (Esc. h-I-15), Libro de ajedrez, dados y tablas (Esc. T-I-6), Libro de las animalias (BNE RES/270), Libro de las cruzes (BNE MSS/9294), Libro de las leyes (BL Add. 20787), Libro del cuadrante señero (Paris: Arsenal 8322), Libros del saber de astronomía (Madrid: Biblioteca Universitaria Complutense 156), Picatrix (Vatican Reg. Lat. 1283), Poridat de las poridades (Esc. L-III-2), Tablas de Zarquiel (Paris: Arsenal 8322).

Fourteenth Century Crónica de veinte reyes (Esc. Y-I-12), Espéculo (BNE MSS/10123), Estoria de España II (Esc. X-I-4), General estoria II (BNE MSS/10237), General estoria V (Esc. I-I-2), General estoria VI (Toledo Catedral 43-20), Fuero de Briviesca (BNE MSS/9199), Libro de la caza (BNE MSS/6376), Libro de la montería (Esc. Y.II.19), Libro de las armas (BNE

 Appendix 

267

MSS/6376), Libro de las tres creencias (BNE MSS/9302), Libro de los estados (BNE MSS/6376), Libro de los fueros de Castilla (BNE MSS/431), Libro del caballero y del escudero (BNE MSS/6376), Libro infinido (BNE MSS/6376), Ordenamiento de Alcalá (BNE VITR/15/7), Sumas de la historia troyana (BNE MSS/9256), Tratado de cetrería (Real Academia Española 9), Tratado de la Asunción de la Virgen (BNE MSS/6376), Tratado de las enfermedades de las aves de caza (Esc. V-II-­ 19), Visita y consejo de médicos (BNE MSS/18052).

Fifteenth Century Arte cisoria (Esc. f-IV-1), Arte de bien morir (Esc. 32-V-194), Atalaya de las Corónicas (BL Egerton: MS 287), Axioco (BNF: Espagnol 458), Biblia romanceada judío cristiana (RAH, códice 87), Breve confesionario (Esc. 32-V-194), Caída de príncipes (HSA B1196), Cárcel de amor (BNE INC/2134), Castigos y documentos para bien vivir (BNE MSS/6559), Claros varones de Castilla (BNE INC/96), Comedia de Calisto y Melibea (HSA), Compendio de medicina (Biblioteca general histórica de la Universidad de Salamanca 2262), Compilación de las batallas campales (BNE INC/249(2)), Conde Lucanor (BNE MSS/6376), Conjuración de Catilina (Esc. g.III.11), Corbacho (Esc. h-III-10), Crónica abreviada (BNE MSS/1356), Crónica de 1344 I (Biblioteca Francisco de Zabálburu y Basabe 11-109), Crónica de Alfonso X (BNE MSS/829), Crónica de España (BNE INC/1732), Crónica de Sancho IV (BNE MSS/829), Cuaderno de las leyes nuevas de la hermandad (BNE R/2462/4), Cura de la piedra (RAH, INC San Román 21), De los oficios (BNE MSS/7815), Doce trabajos de Hércules (BNE MSS/27), Exemplario contra los engaños y peligros del mundo (BNE INC/1994), Enrique fi de Oliva (Österreichische Nationalbibliothek Ink. 5.G.39), Escritura de cómo y por qué razón no se debe dividir partir ni enajenar (BNE INC/249(2)), Esopete ystoriado (John Rylands Library [Manchester] JRULM 19562), Espejo de medicina (BNE MSS/3384), Generaciones y semblanzas (Fundación Lázaro Galdiano Inv. 14998), Gramática castellana (BNE INC/2142), Historia de la linda Melosina (Bibliothèque royale de

268 Appendix

Belgique Inc. B 840), Historia del gran Tamorlán (BNE MSS/9218), Introductiones latinae (BNE INC/2652), Invencionario (BNE INC/2652), Las pronósticas (BNE INC/2438), Letra sobre los matrimonios y casamientos entre los reyes de Castilla (BNE INC/249(2)) Libro de albeitería (BNE INC/2342), Libro de cetrería (BNE MSS/21549), Libro de la caza (BNE MSS/6376), Libro de la caza de las aves (BL Add. 16392), Libro de la montería (Esc. Y.II.19), Libro de las armas (BNE MSS/6376), Libro de las donas (Esc. h-III-20), Libro de los estados (BNE MSS/6376), Libro infinido (BNE MSS/6376), Libro del caballero y del escudero (BNE MSS/6376), Libro del caballero Zifar (BNF Richelieu: Espagnol 36), Libro del Consejo y de los Consejeros (Esc. Z-III-4), Libro del Kuzari (BNE MSS/17812), Libro llamado Infancia Salvatoris (BNE INC/1424), Lilio de medicina (BNE INC/2438), Los siete sabios de Roma (BNE MSS/6052), Menor daño de medicina (Esc. b-IV-34), Morales de Ovidio (BNE MSS/10144), Mostrador y enseñador de los turbados (BNE MSS/10289), Nobiliario vero (BNE INC/2299), Oliveros de Castilla (HSA), Oracional de Fernán Pérez de Guzmán (BNE INC/249(3)), Ordenanzas reales (BNE INC/1338), Retórica (Esc. T.II.12), Secretos de la medicina (Madrid: Real biblioteca II/3063), Sermones contra los iudios e moros (Soria: Biblioteca pública del estado 25-H), Siete partidas (BNE INC/766), Suma de la flor de cirugía (BNE MSS/3338), Suma de las Crónicas de España (Esc. h-II-­ 22), Teseida (BNE MSS/7553), Tratado de astrología (BNE RES/2), Tratado de la adivinanza (BNE MSS/6401), Tratado de la Asunción de la Virgen (BNE MSS/6376), Tratado de la fisionomía en breve suma contenida (BNE INC/51), Tratado de la música (Esc. ç.III.23), Tratado de la peste (BNE INC/51), Tratado de la reformación de la ánima (BNF Richelieu: Espagnol 458), Tratado de las armas (Biblioteca Nazionale Casanatense 1098), Tratado de las fiebres (Esc. M-I-28), Defensa de virtuossas mugeres (BNE MSS/1341), Triunfo de amor (BNE MSS/22019), Triunfo de las donas y Cadira de onor (Madrid: Ed. Nacional, 1982), Universal vocabulario de latín en romance (BNE INC/448  V.1 INC/449  V.2), Valerio de las historias escolásticas y de España (BNE INC/249(1)), Vida de Sanct Isidoro (Esc. b.III.1), Visión delectable (BNE INC/2442).

 Appendix 

269

Sixteenth and Early Seventeenth Century Arnalte y Lucenda (Biblioteca Trivulziana Cod. Triv. 940), Carta de fray Toribio de Motolinía al Emperador Carlos V (Mexico City: Andrade, 1858), Cartas de Pedro de Valdivia (Madrid: Atlas, 1960), Hernán Cortés: Cartas de relación (México: Porrúa, 1985), Crónica de la Nueva España (Madrid: Atlas, 1971), De los nombres de Cristo (Madrid: Católica, 1951), Derrotero al Estrecho de Magallanes (Madrid: Historia 16, 1987), Diálogo de las cosas acaecidas en Roma (Madrid: Espasa Calpe, 1969), Diálogo de Mercurio y Carón (Madrid: Castalia, 1993), Discurso político al rey Felipe III al comienzo de su reinado (Barcelona: Anthropos, 1990), El concejo y consejeros del príncipe (Valencia: Alfonso el Magnánimo, 1952), El Crotalón (Madrid: Bibliófilos Españoles, 1871), El Libro de los proverbios glosados (Kassel: Reichenberger, 1994), El patrañuelo (Madrid: Castalia, 1979), Epistolario espiritual (Madrid: La Lectura, 1962), Examen de ingenios para las ciencias (Madrid: La Rafa, 1930), Floresta española (Madrid: SBE, 1953), Guerra de Granada (Madrid: Rivadeneyra, 1852), Guzmán de Alfarache II (Barcelona: Sebastian de Cormellas, 1605), Historia de la conquista de México (Mexico City, Pedro Robredo, 1943), Historia de la invención de las Yndias (Bogotá: Caro y Cuervo, 1965), Historia de los indios de la Nueva España (Mexico City: Andrade, 1858), Historia natural y moral de las Indias (Madrid: Atlas, 1954), Jardín de flores curiosas (Madrid: Castalia, 1982), La Galatea (Alcalá de Henares: Centro de Estudios Cervantinos 1994), La lozana andaluza (Venice, 1528), Las seiscientas apotegmas (Madrid: Espasa Calpe, 1972), Lazarillo de Tormes (Burgos: Juan de Junta, 1554), Leyes del estilo (Bodleian Library Inc. d. S6. 1500. 1), Libro de la oración y meditación (Salamanca: Portonariis, 1569), Libro de las Fundaciones (Madrid: Biblioteca de autores cristianos, 1997), Libro del ejercicio corporal y de sus provechos (León: Lancia, 1996), Libro primero de las epístolas familiares (Madrid: Aldus, 1950–1952), Lumbre del alma (Seville: Cromberger, 1542), Luz del alma cristiana (Seville: Montesdoca, 1555), Naufragios (Valladolid, 1555), Obra de Agricultura (Alcalá: Brocar, 1513), Peregrinación de la vida del hombre (Medina del Campo: Millis, 1552), Philosofía secreta (Madrid, 1585),

270 Appendix

Primaleón (Salamanca: Juan de Porras, 1512), Recibimiento que hizo la muy noble y muy leal ciudad de Sevilla a D. Felipe II (Seville: Escrivano, 1570), Segunda Celestina (Venice: Sabio, 1536), Sermones de Fray Dionisio Vázquez (Madrid: Espasa-Calpe, 1956), Sevillana medicina (BL C.63.h.23.(2.)), Summa de tratos y contratos (Madrid: Ministerio de Economía, 1977), Tratado de la tribulación (Madrid: Tello, 1877), Tratado y discurso sobre la moneda de vellón (Madrid: Ministerio de Economía, 1987).

 ppendix 3: Texts Cited but Not Included A in the Main Corpus Fazienda de ultramar (Biblioteca Universitaria de Salamanca 1997) Fuero de Madrid (Augustín Millares Carlo, Madrid 1932) Fuero viejo de Alcalá (Alcalá de Henares, AMA [H] F.V.A) Poema de mio Cid (Biblioteca Nacional de España VITR/7/17) Sumario de la medicina (Biblioteca Nacional de España INC/1169)

References

Alarcos Llorach, Emilio. 1994. Gramática de la lengua española. Madrid: Espasa Calpe. Alexiadou, Artemis. 2014. Multiple determiners and the structure of DPs. Amsterdam: John Benjamins. Alexiadou, Artemis, and Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order, V-movement and EPP-checking. Natural Language and Linguistic Theory 16: 491–539. Alonso, Martín. 1962. Evolución sintáctica del español. Madrid: Aguilar. Altmann, Gabriel, Haro Buttlar, Walter Rott, and Udo Strauss. 1983. A law of change in language. In Historical linguistics, ed. B.  Brainerd, 104–115. Bochum: Studienverlag Dr. N. Brockmeyer. Barbosa, Pilar. 2001. On inversion in wh-questions in Romance. In Subject inversion in Romance and the theory of universal grammar, ed. A. Hulk and J.-Y. Pollock, 20–59. Oxford: Oxford University Press. ———. 2009. Two kinds of subject. Studia Lingüística 63 (1): 2–58. Barra Jover, Mario. 2007. Cambios en la arquitectura de la prosa española y romance: sintaxis y cohesión discursiva por correferencia nominal. Revista de Filología Española 87 (1): 7–43. ———. 2008. Tradición discursiva, creación y difusión de innovaciones sintácticas: la cohesión de los argumentos nominales a partir del siglo XIII.  In

© The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9

271

272 References

Sintaxis histórica del español. Nuevas perspectivas desde las tradiciones discursivas, ed. J. Kabatek, 127–150. Frankfurt: Vervuert/Iberoamericana. Bartra-Kaufmann, Ana. 2009. Some remarks about the grammaticalization process of the DP functional domain in Old Romance. In Proceedings of the IV Nereus international workshop on definiteness and DP structure in Romance languages, Arbeitspapier, ed. María Teresa Espinal et  al., vol. 124, 1–26. Konstanz: University of Konstanz. ———. 2013. Indirect evidence of D development in Romance: ‘Internal’ D. Paper given at the Workshop European Dialect Syntax VII, 13–15 June, University of Konstanz. Batllori, Montserrat. 2003. La sintaxis de los determinantes en el Calila e Dimna. In Actas del IV congreso de lingüística general (Cádiz, 3–6 de abril de 2000), vol. II, ed. María Dolores Muñoz Núñez et al., 191–203. Cádiz: Servicio de Publicaciones de la Universidad de Cádiz. ———. 2010. La periferia izquierda del sintagma nominal: artículo ante posesivo en español medieval. In Actes du XXVe congrès international de linguistique et de philologie romanes, ed. Maria Iliescu et al., 419–430. Berlin: Walter de Gruyter. Batllori, Montserrat, and Francesc Roca. 1998. Los determinantes y los cuantificadores en el español medieval y moderno. Paper delivered at the 28th symposium of the Sociedad Española de Lingüística (Relaciones entre Semántica y Sintaxis), Madrid, 14–18 December. ———. 2000. The value of definite determiners from Old Spanish to Modern Spanish. In Diachronic syntax: Models and mechanisms, ed. Susan Pintzuk et al., 241–254. Oxford: Oxford University Press. Battlori, Montserrat, Carlos Sánchez, and Avel·lina Suñer. 1995. The incidence of interpolation on the word order of Romance languages. Catalan Working Papers in Linguistics 4 (2): 185–209. Belletti, Adriana. 2004. Aspects of the low IP area. In The structure of CP and IP.  The cartography of syntactic structures, ed. Luigi Rizzi, vol. 2, 16–51. New York: Oxford University Press. Benincà, Paola. 2006. A detailed map of the left periphery in medieval Romance. In Crosslinguistic research in syntax and semantics: Negation, tense and clausal architecture, ed. Raffaella Zanuttini et  al., 53–86. Washington, DC: Georgetown University Press. Bosque, Ignacio. 1980. Sobre la negación. Madrid: Cátedra. Bosque, Ignacio, and M. Carme Picallo. 2012. Articles as partitives. In Functional heads: The cartography of syntactic structures, ed. Laura Brugé et  al., vol. 7, 138–149. Oxford: Oxford University Press.

 References 

273

Bossong, Geor. 2006. La sintaxis de las Glosas Emilianenses en una perspectiva tipológica. In Actas del VI Congreso Internacional de Historia de la Lengua Española, ed. José Jesús de Bustos and José Luis Girón, vol. 1, 529–544. Madrid: Arco Libros. Bouzouita, Miriam. 2008a. At the syntax-pragmatics interface: Clitics in the history of Spanish. In Language in flux: Dialogue coordination, language variation, change and evolution, ed. Robin Cooper and Ruth Kempson, 221–263. London: College Publications. ———. 2008b. The diachronic development of Spanish clitic placement. Doctoral dissertation, King’s College London. Bresnan, Joan, and Jane Grimshaw. 1978. The syntax of free relatives in English. Linguistic Inquiry 9: 331–391. Burzio, Luigi. 1986. Italian syntax: A government-binding approach. Dordrecht: Reidel. Camus Bergareche, Bruno. 1986. Cronología y extensión de un cambio en la expresión de la negación en español. Revista de Filología de la Universidad de La Laguna 5: 111–122. ———. 2006. La expresión de la negación. In Sintaxis histórica del español, vol. 1, tomo 2, primera parte, ed. Concepción Company, 1165–1252. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. Carlier, Anne, and Béatrice Lamiroy. 2014. The grammaticalization of the prepositional partitive in Romance. In Partitive cases and related categories, ed. Silvia Luraghi and Tuomas Huumo, 477–522. Berlin: Walter de Gruyter. Castillo Lluch, Mónica. 1996. La posición del pronombre átono en la prosa hispánica medieval. Doctoral dissertation, Universidad Autónoma de Madrid. ———. 1997. La interpolación en español antiguo. In Actas del IV Congreso Internacional de Historia de la Lengua Española, ed. Claudio García Turza et al., 409–422. La Rioja: Universidad de La Rioja/Servicio de Publicaciones. ———. 2001. Cronología lingüística y evoluciones pendulares: el caso de la construcción preposición  +  infinitivo  +  pronombre átono. Cahiers de Linguistique Hispanique Médiévale 24: 181–198. Chenery, W.H. 1905. Object pronouns in dependent clauses: A study in Old Spanish word-order. Publications of the Modern Language Association of America 20: 1–151. Chomsky, Noam. 1981. Some concepts and consequences of the theory of government and binding. Cambridge, MA: MIT Press. ———. 1995. The minimalist program. Cambridge, MA: MIT Press.

274 References

———. 2008. On phases. In Foundational issues in linguistic theory, ed. Robert Freidin et al., 133–166. Cambridge, MA: MIT Press. ———. 2013. Problems of projection. Lingua 130: 33–49. Chomsky, Noam, and Howard Lasnik. 1977. Filters and control. Linguistic Inquiry 8 (3): 425–504. Cinque, Guglielmo. 1990. Types of Ā-dependencies. Cambridge, MA: MIT Press. ———. 1993. A null theory of phrase and compound stress. Linguistic Inquiry 24 (2): 239–297. Clavería Nadal, Gloria. 1992. La construcción artículo+posesivo en los siglos XIV y XV.  In Actas del II Congreso Internacional de Historia de la Lengua Española, vol. 1, 347–359. Madrid: Pabellón de España. Company, Concepción. 1991. La extensión del artículo en el español medieval. Romance Philology 44 (4): 402–424. ———. 2006. Artículo + posesivo + sustantivo y estructuras afines. In Sintaxis histórica de la lengua española, vol. 2, tomo 1, segunda parte: la frase nominal, ed. Concepción Company, 759–880. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. ———. 2009. Parámetros de gramaticalización en los indefinidos compuestos del español. In Romanística sin complejos. Homenaje a Carmen Pensado, ed. Fernando Sánchez Miret, 71–103. Bern: Peter Lang. Company, Concepción, and Julia Pozas Loyo. 2006. Los indefinidos compuestos y los pronombres genérico-impersonales omne y uno. In Sintaxis histórica de la lengua española, vol. 2, tomo 2, segunda parte: la frase nominal, ed. Concepción Company, 1073–1222. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México. Contreras, Heles. 1986. Spanish bare NPs and the ECP. In Generative studies in Spanish syntax, ed. Ivonne Bordelois, Heles Contreras, and Karen Zagona, 25–49. Dordrecht/Riverton: Foris. Cornilescu, Alexandra. 1992. Remarks on the determiner system of Rumanian: The demonstratives AL and CEL. Probus 4: 189–260. Cruschina, Silvio, and Ioanna Sitaridou. 2009. From Modern to Old Romance: The interaction of information structure and word order. Paper presented at XI Diachronic Generative Syntax Conference (DiGS), August, Universidade de Campinas. Danford, Richard K. 2002. Preverbal accusatives, pronominal reduplication and information packaging: A diachronic analysis of Spanish. Doctoral dissertation, Ohio State University. Davies, Mark. 1997. The evolution of Spanish clitic climbing: A corpus-based approach. Studia Neophilologica 69 (2): 251–263.

 References 

275

———. 2011. Google books (Spanish) corpus (45 billion words, 1500–2000). Available online at http://googlebooks.byu.edu/ de Haan, Germen J., and Fred P. Weerman. 1986. Finiteness and verb fronting in Frisian. In Verb second phenomena in Germanic languages, ed. Hubert Haider and Martin Prinzhorn, 77–110. Dordrecht: Foris. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Drubig, Hans Bernhard. 2003. Toward a typology of focus constructions. Linguistics 41 (1): 1–50. Eberenz, Rolf. 2000. El español en el otoño de la Edad Media. Sobre el artículo y los pronombres. Madrid: Gredos. Ebert, Christian, Cornelia Ebert, and Stefan Hinterwimmer. 2014. A unified analysis of conditionals as topics. Linguistics and Philosophy 37 (5): 353–408. Eide, Kristine G., and Ioanna Sitaridou. 2013. Contrastivity and information structure in the old Ibero-Romance languages. In Information structure and syntactic change in Germanic and Romance languages, ed. Kristin Bech and Kristine G. Eide, 377–412. Amsterdam: John Benjamins. Ellegård, Alvar. 1953. The auxiliary do: The establishment and regulation of its use in English. Stockholm: Almqvist and Wiksell. Elsig, Martin. 2017. New insights into an old form: A variationist analysis of the pleonastic possessive in Guatemalan Spanish. Language Variation and Change 29 (2): 157–186. England, John. 1980. The position of the direct object in Old Spanish. Journal of Hispanic Philology 5 (1): 1–23. ———. 1983. Word order in Old Spanish prose: The indirect object. Neophilologus 67: 385–394. Español Echevarría, Manuel. 1994. A typology for NPI-licensing. Unpublished manuscript. Los Angeles: University of California. Espinal, M.  Teresa. 1992. Expletive negation and logical absorption. The Linguistic Review 9 (4): 333–358. ———. 2000. On the semantic status of N-words in Catalan and Spanish. Lingua 110: 557–580. Espinal, M. Teresa, Susagna Tubau, Joan Borràs-Comes, and Pilar Prieto. 2016. Double negation in Catalan and Spanish. Interaction between syntax and prosody. In Negation and polarity: Experimental perspectives, ed. Pierre Larrivée and Chungmin Lee, 145–176. Cham: Springer. Fassi Fehri, Abdelkader. 1980. Some complement phenomena in Arabic, lexical grammar, the complementizer phrase hypothesis and the Non-accessibility condition. Unpublished MS, University of Rabat.

276 References

———. 1993. Issues in the structure of Arabic clauses and words. Dordrecht: Kluwer. Fernández Ordóñez, Inés. 2008–2009. Orden de palabras, tópicos y focos en la prosa alfonsí. Alcanate 6: 139–172. ———. 2011. La lengua de Castilla y la formación del español. Madrid: Real Academia Española. Fischer, Susann. 2014. Revisiting stylistic fronting in Old Spanish. In Left sentence peripheries in Spanish: Diachronic, variationist and comparative perspectives, ed. Andreas Dufter and Álvaro S. Octavio de Toledo, 53–76. Amsterdam: John Benjamins. Fontana, Josep. 1993. Phrase structure and the syntax of clitics in the history of Spanish. Doctoral dissertation, University of Pennsylvania. Gago Jover, Francisco. 2011a. Prose works of Alfonso X el sabio. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www. hispanicseminary.org/t&c/ac/index-en.htm. Last consulted: October 2018. ———. 2011b. Spanish legal texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/lex/index-en.htm. Last consulted: October 2018. ———. 2011c. Spanish medical texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/med/index-en.htm. Last consulted: October 2018. ———. 2011d. Spanish chronicle texts. Digital library of Old Spanish texts. Hispanic seminary of medieval studies. http://www.hispanicseminary.org/ t&c/cro/index-en.htm. Last consulted: October 2018. Gallego, Ángel. 2007. Phase theory and parametric variation. Doctoral dissertation, Universitat Autònoma de Barcelona. García García, Serafina. 1992. Evolución de qual en la lengua literaria desde el siglo XIII hasta el XV. In Actas del II congreso internacional de historia de la lengua española, ed. Manuel Ariza Viguera, vol. I, 445–454. Madrid: Pabellón de España. Giorgi, Alessandra, and Giuseppe Longobardi. 1991. The syntax of noun phrases. Cambridge: Cambridge University Press. Girón Alconchel, José Luis. 2006. Las oraciones de relativo II. Evolución del relativo compuesto el que, la que, lo que. In Sintaxis histórica de la lengua española, vol. 2, tomo 2, segunda parte: la frase nominal, ed. Concepción Company, 1477–1592. Mexico City: Fondo de Cultura Económica/Universidad Nacional Autónoma de México.

 References 

277

———. 2012a. Los relativos compuestos españoles y su interés para la teoría de la gramaticalización. In Actas del VIII congreso internacional de historia de la lengua española, ed. Emilio Montero Cartelle and Carmen Manzano Rovira, vol. 1, 57–78. Madrid: Meubook. ———. 2012b. Gramaticalización como creación de lengua a partir del habla. Relativos e indefinidos compuestos en los Fueros de Aragón y en el Fuero de Teruel. Archivo de Filología Aragonesa 68: 15–38. Granberg, Robert. 1988. Object pronoun position in medieval and early modern Spanish. Doctoral dissertation, University of California, Los Angeles. ———. 1999. Clitic position in thirteenth-century Spanish: Sentences with preverbal subject. La Corónica 27 (2): 89–113. Haegeman, Liliane, and Raffaella Zanuttini. 1991. Negative heads and the NEG criterion. The Linguistic Review 8: 233–251. Hernanz, María Lluisa. 2001. ¡En bonito lío me he metido!: notas sobre la afectividad en español. Moenia 7: 93–109. ———. 2006. Emphatic polarity and C in Spanish. In Studies in Spanish syntax, ed. Laura Brugè, 105–150. Venice: Università Ca’ Foscari. Herrera, María Teresa, María Nieves Sánchez, María Estela González de Fauve, and María Purificación Zabía. 1999. Textos y concordancias electrónicos de documentos castellanos de Alfonso X. CD-ROM. Madison: Hispanic Seminary of Medieval Studies. Hirschbühler, Paul, and Marie Labelle. 2001. Evolving Tobler–Mussafia effects in the placements of French clitics. In New approaches to old problems: Issues in Romance historical linguistics, ed. Steven Dworkin and Dieter Wanner, 165–182. Amsterdam: John Benjamins. Höhle, Tilman N. 1992. Über Verum-Fokus im Deutschen. In Informationsstruktur und Grammatik, ed. Joachim Jacobs, 112–141. Opladen: Westdeutscher Verlag. Holmberg, Anders. 2000. Scandinavian stylistic fronting: How any category can become an expletive. Linguistic Inquiry 31: 445–483. ———. 2015. Verb second. In Syntax – Theory and analysis. An international handbook, ed. Tibor Kiss and Artemis Alexiadou, chapter 12. Berlin: Walter de Gruyter (e-book). Ishikawa, Masataka. 1995. On categorial evolution: A case study in Spanish possessives. In Historical linguistics 1993: Selected papers from the 11th international conference on historical linguistics, ed. Henning Andersen, 205–216. Amsterdam: John Benjamins.

278 References

———. 1997. A note on reference and definite articles in Old Spanish. Word 48 (1): 61–68. Jaeggli, Osvaldo. 1980. On some phonologically-null elements in syntax. Doctoral dissertation, MIT, Cambridge, MA. Kadmon, Nirit, and Fred Landman. 1990. Polarity sensitive any and free choice any. In Proceedings of the 7th Amsterdam Colloquium, ed. Martin Stokhof and Leen Torenvliet, 227–252. Amsterdam: Instituut voor Taal, Logica en Informatie. Kauhanen, Henri, and George Walkden. 2017. Deriving the constant rate effect. Natural Language and Linguist Theory. (Open access) 36: 483. https://doi. org/10.1007/s11049-017-9380-1. Kayne, Richard S. 2008. Some preliminary comparative remarks on French and Italian definite articles. In Foundational issues in linguistic theory, ed. Robert Freidin et al., 291–322. Cambridge, MA: MIT Press. Kermack, William O., and Anderson G. McKendrick. 1927. Contributions of mathematical theory to epidemics. Proceedings of the Royal Society Series A 115: 700–721. Kiss, Katalin É. 1998. Identificational focus versus information focus. Language 74: 245–273. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. ———. 1994. Morphosyntactic variation. Chicago Linguistic Society (The parasession on variation in linguistic theory) 30 (2): 180–201. Kuchenbrandt, Imme. 2016. Prosody and object clitic placement: A comparison of Old and Modern Spanish. Lingua 181: 81–98. Laka, Itziar. 1990. Negation in syntax: On the nature of functional categories and projections. Doctoral dissertation, MIT, Cambridge, MA. Lambrecht, Knud. 1994. Information structure and sentence form. Topic, focus, and the mental representations of discourse referents. Cambridge: Cambridge University Press. ———. 2000. When subjects behave like objects: An analysis of the merging of S and O in sentence focus constructions across languages. Studies in Language 24 (3): 611–682. Lambrecht, Knud, and Laura A. Michaelis. 1998. Sentence accent in information questions: Default and projection. Linguistics and Philosophy 21 (5): 477–544. Landau, Idan. 2007. EPP extensions. Linguistic Inquiry 38 (3): 485–523. Lapesa, Rafael. 1981. Historia de la lengua española, 9th ed. Madrid: Gredos.

 References 

279

———. 2000. Estudios de morfosintaxis histórica del español. Madrid: Gredos. Lasnik, Howard, and Nicholas Sobin. 2000. The who/whom puzzle: On the preservation of an archaic feature. Natural Language and Linguistic Theory 18: 343–371. Leonetti, Manuel, and Victoria Escandell Vidal. 2009. Fronting and verum focus in Spanish. In Focus and background in Romance languages, ed. Andreas Dufter and Daniel Jacob, 155–204. Amsterdam: John Benjamins. ———. 2010. Las anteposiciones inductoras de foco de polaridad. In La renovación de la palabra en el bicentenario de la Argentina: los colores de la mirada lingüística, ed. Victor M.  Castel and Liliana Cubo de Severino, 733–743. Mendoza: Editorial FFyL, UNCuyo. Linebarger, Marcia. 1980. The grammar of negative polarity. Doctoral dissertation, MIT. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. ———. 2017. Theory and experiment in parametric minimalism: The case of Romance negation. Revista Linguí∫tica 13 (2): 108–157. Mackenzie, David. 1997. A manual of manuscript transcription for the dictionary of the Old Spanish Language. (5th edition, revised and expanded by Ray Harris-Northall). Madison: Hispanic Seminary of Medieval Studies. Mackenzie, Ian. 2010. Refining the V2 hypothesis for Old Spanish. Bulletin of Hispanic Studies 87 (4): 379–396. ———. 2017. The rise and fall of proclisis in Old Spanish postprepositional infinitival clauses: A quantitative approach. Bulletin of Hispanic Studies 94 (2): 127–146. Mackenzie, Ian, and Wim van der Wurff. 2012. Relic syntax in Middle English and Medieval Spanish: Parameter interaction in language change. Language 88 (4): 846–876. Martínez Gil, Fernando. 1989. Las inversiones del orden de palabras en el Romancero. Hispania 72 (4): 895–908. Martins, Ana Maria. 2000. Polarity items in Romance: Underspecification and lexical change. In Diachronic syntax: Models and mechanisms, ed. Susan Pintzuk et al., 191–219. Oxford: Oxford University Press. ———. 2003. From unity to diversity in Romance syntax: Portuguese and Spanish. In Aspects of multilingualism in European language history, ed. Kurt Braunmüller and Gisella Ferraresi, 201–233. Amsterdam: Benjamins. Matute Martínez, Cristina. 2013. Hacia una caracterización dialectal de la interpolación en el castellano de la Edad Media. In Actas del XXVI congreso inter-

280 References

nacional de lingüística y filología romances, 2010. Universitat de València, ed. Casanova Herrero and Césareo Calvo Rigual, 155–164. Berlin: Walter de Gruyter. Menéndez Benito, Paula. 2010. On universal free choice items. Natural Language Semantics 18 (1): 33–64. Menéndez Pidal, Ramón. 1926. Orígenes del español, 6th ed. Madrid: Espasa-Calpe. Meyer-Lübke, Wilhelm. 1897. Zur Stellung der tonlosen Objektspronomina. Zeitschrift für romanische Philologie 21 (3): 313–334. Molinelli, Piera. 1988. Fenomeni della negazione dal latino all’italiano. Florence: La Nuova Editrice. Mussafia, Adolfo. 1886. Una particolarità sintattica della lingua italiana dei primi secoli. In Miscellanea di filologia e linguistica, dedicata alla memoria di Napoleone Caix e Ugo Angelo Canello, ed. G.I. Ascoli et al., 255–261. Florence: Le Monnier. O’Neill, John. 1999. Electronic texts and concordances of the Madison Corpus of early Spanish manuscripts and printings. CD-ROM.  Madison: Hispanic Seminary of Medieval Studies. Octavio de Toledo, Álvaro S. 2014. Entre gramaticalización, estructura informativa y tradiciones discursivas: algo más sobre nada. In Procesos de gramaticalización en la historia del español, ed. José Luis Girón Alconchel and Daniel Sáez Rivera, 263–319. Frankfurt: Iberoamericana/Vervuert. Ordóñez, Francisco. 1997. Word order and clause structure in Spanish and other Romance languages. Doctoral dissertation, CUNY. Penny, Ralph. 2002. A history of the Spanish language, 2nd ed. Cambridge: Cambridge University Press. Pesetsky, David. 1998. Some optimality principles of sentence pronunciation. In Is the best good enough? ed. Pilar Barbosa et al., 337–383. Cambridge, MA: MIT Press. Piqueira, José Roberto. 2010. Rumor propagation model: An equilibrium study. Mathematical Problems in Engineering, 631357. https://www.hindawi.com/ journals/mpe/2010/631357/ Plank, Frans. 2003. Double articulation. In Noun phrase structure in the languages of Europe, ed. Frans Plank, 337–395. Berlin: Mouton de Gruyter. Poletto, Cecilia. 2014. Word order in Old Italian. Oxford: Oxford University Press. Pons Rodríguez, Lola. 2007. La qual çibdad: las relativas con antecedente adjunto del siglo XIII a hoy. Evolución de un procedimiento cohesivo. Romanistisches Jahrbuch 58: 275–305.

 References 

281

Poole, Geoffrey. 2011. Focus and the development of N-words in Spanish. In Romance languages and linguistic theory 2009. Selected papers from ‘Going Romance’ (Nice 2009), ed. Janine Berns et al., 291–303. Amsterdam: John Benjamins. ———. 2013. Interpolation, verb-second and the low left periphery in Old Spanish. Iberia: International Journal of Theoretical Linguistics 5 (1): 69–98. ———. 2017. Feature inheritance in Old Spanish: (Re)Visiting V2. In Order and structure in syntax I: Word order and syntactic structure, ed. Laura R. Bailey and Michelle Sheehan, 49–68. Berlin: Language Science Press. Posner, Rebecca. 1984. Double negatives, negative polarity and negative incorporation in Romance: A historical and comparative view. Transactions of the Philological Society 82 (1): 1–26. Postma, Gertjan. 2010. The impact of failed changes. In Continuity and change in grammar, ed. Anne Breitbarth et  al., 269–302. Amsterdam: John Benjamins. ———. 2017. Modelling transient states in language change. In Micro-change and macro-change in diachronic syntax, ed. Éric Mathieu and Robert Truswell, 75–93. Oxford: Oxford University Press. Pountain, Christopher J. 2001. A history of the Spanish language through texts. London: Routledge. Quer, Josep. 2002. Edging quantifiers: On QP-fronting in Western Romance. In Romance languages and linguistic theory 2000, ed. Claire Beyssade et al., 253–270. Amsterdam: John Benjamins. Rini, Joel. 1990. Dating the grammaticalization of the Spanish clitic pronoun. Zeitschrift für romanische Philologie 106: 354–370. Rivero, María Luisa. 1984. Diachronic syntax and learnability: Free relatives in thirteenth-century Spanish. Journal of Linguistics 20 (1): 81–129. ———. 1986. Parameters in the typology of clitics in Romance and Old Spanish. Language 64: 774–807. ———. 1988. La sintaxis de qual quiere y sus variantes en el español antiguo. Nueva revista de filología hispánica 36: 47–47. ———. 1991. Clitic and NP climbing in Old Spanish. In Current studies in Spanish linguistics, ed. Héctor Campos and Fernando Martínez Gil, 241–282. Washington, DC: Georgetown University Press. ———. 1993. Long head movement versus V2, and null subjects in Old Romance. Lingua 89: 217–245. Rizzi, Luigi. 1990. Speculations on verb-second. In Grammar in progress: Essays in honour of Henk van Riemsdijk, ed. Joan Mascaró and Marina Nespor, 375–386. Groningen: Foris.

282 References

Roberts, Ian. 2007. Diachronic syntax. Oxford: Oxford University Press. Roberts, Ian, and Anna Roussou. 2003. Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rosemeyer, Malte, and Andrés Enrique Arias. 2016. A match made in heaven: Using parallel corpora and multinomial logistic regression to analyze the expression of possession in Old Spanish. Language Variation and Change 28 (3): 307–334. Sánchez López, Cristina. 1999. La negación. In Gramática descriptiva de la lengua española, vol. 2: Las construcciones sintácticas fundamentales. Relaciones temporales, aspectuales y modales, ed. Violeta Demonte and Ignacio Bosque, 2561–2634. Madrid: Espasa Calpe. Serradilla Castaño, Ana. 2010. “Artículo + posesivo + nombre” frente a “posesivo + nombre” como variante invisible en un texto medieval. Epos 26: 53–76. Sitaridou, Ioanna. 2011. Word order and information structure in Old Spanish. Catalan Journal of Linguistics 10: 159–184. Sobin, Nicholas. 1997. Agreement, default rules, and grammatical viruses. Linguistic Inquiry 28: 318–343. Stark, Elisabeth. 2008. Typological correlations in nominal determination in Romance. In Essays on nominal determination: From morphology to discourse management, ed. Henrik Høeg Müller and Alex Klinge, 45–61. Amsterdam: John Benjamins. Sundquist, John D. 2011. Negative movement in the history of Norwegian: The evolution of a grammatical virus. In Grammatical change: Origins, nature, outcomes, ed. Diane Jonas et al., 293–312. Oxford: Oxford University Press. Suñer, Margarita. 1982. The syntax and semantics of Spanish presentational sentence-­types. Washington, DC: Georgetown University Press. ———. 1995. Negative elements, island effects and resumptive no. The Linguistic Review 12: 233–273. Tagliamonte, Sali A. 2012. Variationist sociolinguistics: Change, observation, interpretation. Oxford: Wiley-Blackwell. Tobler, Adolf. 1875. Review of Jules Le Coultre, De l’ordre des mots dans Crestien de Troyes. Göttingische Gelehrte Anzeigen 34: 1057–1082. Travis, Lisa. 1984. Parameters and effects of word order variation. Doctoral dissertation, MIT, Cambridge, MA. van Riemsdijk, Henk. 2006. Free relatives. In The Blackwell companion to syntax vol. II, ed. Martin Everaert and Henk Riemsdijk, 338–382. Oxford: Blackwell. Vergnaud, Jean Roger, and Maria Luisa Zubizarreta. 1992. The definite determiner and constructions in French and in English. Linguistic Inquiry 23: 595–652.

 References 

283

Wanner, Dieter. 1991. The Tobler–Mussafia law in Old Spanish. In Current studies in Spanish linguistics, ed. Héctor Campos and Fernando Martínez Gil, 313–378. Washington, DC: Georgetown University Press. ———. 2006. The power of analogy: An essay on historical linguistics. The Hague: Mouton de Gruyter. ———. 2007. Wild variation, random patterns and uncertain data. In Historical linguistics 2007. Selected papers from the 18th international conference on historical linguistics, ed. Monique Dufresne et al., 185–194. Amsterdam: John Benjamin. ———. 2008. Pronombres átonos de objeto con infinitivo o la pertinencia de la periferia. In Actas del VII congreso internacional de historia de la lengua española, ed. Concepción Company and José G.  Moreno de Alba, 197–224. Madrid: Arco Libros. Weerman, Fred. 1993. The diachronic consequences of first and second language acquisition: The change from OV to VO. Linguistics 31: 903–931. Wolfe, Sam. 2015. The nature of Old Spanish verb second reconsidered. Lingua 164: 132–155. Wright, Roger. 1994. Early Ibero-Romance. Twenty-one studies on language and texts from the Iberian Peninsula between the Roman Empire and the thirteenth century. Newark: Juan de la Cuesta. ———. 1999. Periodization and how to avoid it. In Essays in Hispanic linguistics dedicated to Paul M. Lloyd, ed. Robert J. Blake et al., 25–41. Newark: Juan de la Cuesta. ———. 2012. Studying Ibero-Romance before 1200. Paper given at the 2nd Cambridge Colloquium on the ‘Histories of the Ibero-Romance languages’ (‘Norman MacColl Symposium 2012’), Queens’ College, March. ———. 2013. The prehistory of written Spanish and the thirteenth-century nationalist zeitgeist. In A political history of Spanish. The making of a language, ed. José del Valle, 31–43. Cambridge: Cambridge University Press. Yang, Charles D. 2000. Internal and external forces in language change. Language Variation and Change 12: 231–250. Zagona, Karen. 2002. The syntax of Spanish. Cambridge: Cambridge University Press. Zamparelli, Roberto. 2002. Definite and bare kind-denoting noun phrases. In Romance languages and linguistic theory 2000: Selected papers from ‘Going Romance’ 2000, ed. Claire Beyssade et  al., 305–343. Amsterdam: John Benjamins. Zubizarreta, María Luisa. 1998. Prosody, focus and word order, Linguistic inquiry monograph 33. Cambridge, MA: MIT Press.

Index1

A

B

Ā-dependency, 28–35, 56 Adjective, 23, 26, 30, 53, 55, 56, 60, 61, 64, 103, 135, 136, 165, 175, 186, 191 Adverb, 25, 30, 52, 54, 57, 60–62, 81, 103, 165, 208, 212, 241n2, 241n4 Adverbial adverbial clause, 5, 36–38 circumstantial adverbial, 34, 38, 41, 64, 86–87 post-adverbial context, 87, 94, 95 Ā-movement, 41, 42, 58, 64, 66n7, 66n9, 67n11, 81, 87, 102, 127, 170 Antecedent, 165–167, 169, 175, 187, 190, 192, 255 Apposition, 147, 152, 156n4

Bell curve, 16, 57, 61–65, 76, 97, 98, 153, 182, 236, 239, 240, 247–249 C

Catalan, 1, 50, 205, 206, 211, 215, 226, 228, 242n6 Change failed, 15–17, 57, 75–117, 145, 153, 182, 199, 236, 239, 240, 247, 248 rate of, 11–13, 15, 60, 95, 96, 116, 124, 149–151, 155, 181, 182, 229, 230, 251, 252 successful, 75–117, 248 syntactic, 10, 12, 17, 60, 76, 94, 183, 226, 243n14, 250, 251, 256

 Note: Page numbers followed by ‘n’ refer to notes.

1

© The Author(s) 2019 I. E. Mackenzie, Language Structure, Variation and Change, https://doi.org/10.1007/978-3-030-10567-9

285

286 Index

Clitic clitic climbing, 77, 78, 107 clitic left dislocation (CLLD), 5, 35, 65n1, 67n10, 68n14 clitic linearization, 75–118 clitic placement, 33, 76, 84, 86, 87, 89, 115, 117, 252 See also Enclitic; Proclisis; Pronoun Comparative, 186, 211, 215, 222, 241n1, 241n3, 242n6 Competition, 10, 83, 91, 143, 149, 155, 156n1, 158n12, 180, 197, 201n10, 251 Complementizer, 31, 69n21, 80, 130, 137, 138, 164, 187, 195, 196, 199n1, 200n3, 255 Conditional clause, 131, 154, 211, 218 Constant Rate Effect (CRE), 10, 12, 60, 64, 76, 91, 94–96, 102, 116, 124, 151, 155, 182, 229, 235, 250–252 Context macro-context, 94–97, 99, 252 micro-context, 94–97, 99–101, 117n7, 118n8, 252 time separation between contexts, 124, 183, 250 Coordinating conjunction, 68n12, 82, 84, 252 Coordination, 79, 84, 113, 117n2 Copula, 54, 55, 125, 170 Copy, 66n6, 85, 127 CP recursion, 124, 136, 155, 187 CRE, see Constant Rate Effect

D

Decay, 10, 15–17, 63, 65, 150, 182, 183, 188, 201n6, 230, 238, 240, 249 Definite article definite article + possessive, 9, 148–150, 153–154, 158n12, 263 definite article + qual, 179, 184, 185, 199, 264 definite article + que, 190, 199 expletive definite article, 134, 154 See also Double articulation; Expletive article Definiteness, 48, 129, 139–142, 157–158n10 D-element ordering of D-elements, 139–142 outermost D-element, 139 overt D-element, 123, 175 true D-element, 190 Demonstrative, 42–50, 68n18, 124, 128, 140, 146–153, 156n1, 156n4, 184, 185, 263 Derivative, 97–99, 117, 248 Determiner, 43, 123, 125, 136, 139, 154, 170, 173, 177, 179 indefinite, 128, 142, 143, 145, 155, 263 possessive, 9, 139, 140, 142 stacked determiners, 129, 134–142, 154, 178 See also D-element; Null D; Partitive Determinerless, 123, 131, 134, 154, 198 Differential object marking (DOM), 145

 Index 

Discourse-given, 42, 44, 46–48, 52, 53, 55, 56, 61, 165 Do-support, 12–14, 16, 19n9 Double articulation, 136–139, 141, 155, 163–199 Doubly-filled Comp, 193–196, 200n3, 255 DP recursion, 124, 136–139, 146–152, 155, 164, 178, 179, 182–185, 187–190, 199, 250, 255

287

Free choice, 163–199, 207, 241n2, 241n3, 255 Free choice items (FCI), see Free choice Free relative, 163–199, 218, 255 French, 1, 18n2, 92, 107, 142, 144, 145, 155, 163, 185, 214, 223, 248, 255 Fundamental Asymmetry Hypothesis (FAH), 205, 212, 231, 235 G

E

Edge edge feature, 42, 59 edge of CP, 31, 32, 34, 35, 40–42, 56, 58, 64, 86, 165 edge of TP, 41, 81, 88 Enclitic, 5, 34, 35, 39, 40, 77, 78, 80, 82, 83, 85, 89, 104, 105, 107, 207, 251 Existential interpretation, 125, 127, 130, 132, 156n5, 157n6 Expletive article, 130–134, 157n6 See also Definite article

Germanic, 35, 36, 38, 40, 136, 253 Grammaticalization, 142, 198 Granularity restriction, 94–97, 99, 116, 251 Growth, 10 exponential, 11, 14, 17, 59, 181, 188, 189 growth rate, 11–13, 15, 16, 64, 91, 94–96, 102, 182, 250 linear, 11 See also Logistic I

F

Filter, 137, 140, 141, 155, 168, 169, 171, 193–196, 199n1, 200n3, 255 Focus affective focus, 53, 55, 69n22 focus-affected reading, 50, 51 new-information focus, 53–55 predicate focus, 54, 87 verum focus, 69n21

Imperative, 29, 83, 91, 97, 99–102, 107, 108, 116, 118n7, 118n11, 255, 259 Indefinite, 89, 123, 124, 128, 139, 142–146, 149–153, 155, 157n6, 158n10, 188, 250 Indirect question, 168, 169, 172, 179 Infinitive bare, 77 fronted, 55, 61 post-prepositional, 90, 102, 103, 107, 116

288 Index

Inherently proclitic, 83, 87, 89, 93, 99, 101, 102, 116, 255 Intercept, 93, 96, 243n16, 251 Interpolation, 108–115, 117n1, 262 Interrogative, 5, 28, 31, 42, 163, 164, 178, 179, 191, 207, 210 Intonation, 33, 34, 66n2, 82, 100, 101, 126, 166, 242n11 Inversion, 12, 31, 32, 36, 40, 64, 253 Italian, 68n13, 107, 125, 135, 136, 142, 144, 145, 155, 157n6, 163, 205, 206, 209, 211–213 L

Latin, 3–8, 19n6, 23, 61, 123, 144, 154, 155, 177, 178, 199, 242n9 Latin American, 151, 190 Left dislocation, 32, 64, 65n1, 67n10, 68n17, 89, 137, 138, 155, 179, 187 See also Clitic Left periphery, 24, 31, 102, 154 Lexical government, 125–127, 129, 131, 132, 135, 156n2, 156n5, 174, 175 Logistic curve, 12, 96, 97, 248 function, 62, 117, 248 growth, 12–15 regression, 14, 15, 19n10, 59, 103, 158n13, 243n16 N

Negation clausal, 206, 209, 212, 214, 216, 223, 231, 239, 254 double, 6, 210, 234, 243n11

expletive, 112, 206, 216, 217, 226–228, 231–235, 240, 242n6, 243n15, 254 logical, 205, 214, 223, 226, 240, 254 in Old Spanish, 216–228 Negative concord, 205, 206, 212–216, 219, 224, 226, 231, 233, 235, 242n10, 254 morpheme, 6, 205, 206, 210, 211, 213, 215, 216, 219, 223, 224, 226, 228, 231, 232, 238–240, 242n9, 253, 254 Negative quantifier (NQ), 6, 7, 205, 207–214, 216, 219, 220, 223, 226, 227, 231–233, 240, 242n10, 254 Noun abstract, 51, 146 bare, 125, 127, 132 count, 125, 132, 144, 188 mass, 125, 144, 146, 175, 188 proper, 133 NQ, see Negative quantifier Null D, 125–127, 129–134, 143, 154, 156n5, 157n6, 174, 175, 188 N-words in modern Spanish, 209–216, 218, 228 no(n)–N-word, 220–223 Old Spanish N-words, 219, 220, 223, 242n10 postverbal N-words, 213, 226, 232 preverbal N-words, 211, 212, 219, 223–226, 228–232, 234, 235, 238, 243n14, 252–254

 Index  O

Object definite, 48 fronted, 25, 30, 46, 47, 65n1, 66n9 possessive, 46 quantified, 49–50 See also Demonstrative Odds ratio (OR), 12–15, 44, 48, 50, 59, 68n18, 91, 93, 95, 99, 102, 149, 150, 181, 229, 230, 238 OR, see Odds ratio P

Parameter, 40, 41, 57, 63, 64, 93, 98, 243n16 Partitive, 142–146, 155, 158n10 PI, see Polarity, polarity item Pied-piping, 167, 175 Polarity polarity context, 154, 208, 210, 212, 215–218, 227, 228, 231–233, 235, 240, 254 polarity item (PI), 191, 207–209, 217, 219, 232, 233, 240, 254 Population dynamics, 63, 240, 249 Possessive, 8, 9, 128, 134–136, 139, 140, 142, 146, 148–155, 157n7, 158n12, 164, 168, 178, 179, 183–185, 199, 248, 250, 251, 255 See also Definite article Predicative, 26, 27, 30, 53–56, 60, 61, 64, 103, 125, 132, 165, 173, 258 Prepositional phrase (PPs), 25, 27, 42, 44, 48–50, 52, 56, 60, 61, 103, 132, 194, 236, 258 Preposition deletion, 170–172

289

Prescriptive, 233, 234 Primary linguistic data (PLD), 104, 106, 107, 116, 146 Probability absolute, 12–14 relative, 13, 19n7, 59, 149, 153 Proclisis advance of, 76, 91, 93, 95, 97, 100, 101, 103, 105–107, 109, 115, 248, 250, 251 decline of, 92, 99, 107, 108 gerundial, 102, 103, 107, 108 infinitival, 104, 106, 107, 116 obligatory, 77, 90 systematic, 29, 39, 78 See also Inherently proclitic; Tobler–Mussafia Pro-drop, 32, 33, 35, 88, 89 Pronoun relative, 164–166, 173, 175, 184, 186, 190, 192, 199, 201n9, 249, 255 resumptive, 32, 35, 67n10, 88, 176 silent, 88 weak, 5, 33–35, 39, 40, 64, 75, 77, 80, 83, 84, 87, 89, 93, 99, 101, 102, 105, 108–110, 113, 114, 116, 117n1, 255 See also Clitic; Demonstrative; Whp-value, 103, 149–151, 182, 185, 188–190 Q

Quantifier, 42, 49–53, 67n10, 69n21, 128, 129, 131, 142, 164, 188, 191, 192, 197, 200n3, 205, 209, 212, 241n2, 255

290 Index

inversion, 12, 31, 32, 36, 40, 64, 253 preverbal, 33, 66n7, 67n10, 87–89, 127, 131, 135, 174, 253 structural, 35, 126, 234 true, 35, 88, 89 See also Inversion

Quantitative, 8, 10, 12, 15–17, 24, 27, 42, 56–64, 75, 91–110, 116, 145–155, 180–185, 189, 199, 221, 225, 228–235, 238–240, 247–256 Quier, 164, 171, 191–198, 200n3, 201n8–10, 255 R

T

Reanalysis, 107, 116, 124, 143, 173, 184, 190, 198, 199, 248 Reciprocal, 15, 60, 62, 158n14, 201n6, 238 Relative clause, 46, 87, 163, 165–167, 170, 173, 175, 176, 180, 185, 187, 189, 190, 192, 193, 195, 198–199, 201n9, 255 See also Free relative Romance, 1, 3, 7, 8, 19n4, 33, 75, 88, 123, 125, 127, 142, 145, 155, 163, 205, 206, 208, 210, 219, 235, 240, 242n9

Tobler–Mussafia, 5, 33, 34, 40, 80, 82, 83, 90, 92, 93, 100, 102, 255 Topic, 32–35, 41, 42, 44, 48–49, 53–56, 58, 64, 67n10, 68n17, 69n23, 81, 88, 89, 112, 117n1, 117n5 Topic–comment, 48, 54

S

Scope, 131, 191, 192, 207, 210, 213, 219, 220, 223, 232 Semantic feature, 140, 141, 179 Slope, 14, 15, 19n9, 70n24, 93, 243n16, 251 Spellout, 88, 137, 140, 141, 170, 176, 179 Stylistic, 63, 65, 79, 103, 115, 155, 157n7, 183, 189, 199, 222, 238, 240, 248, 249 Subject, 35, 53, 64, 88, 174

V

Verb-second (V2) accidental V2, 39–41, 64 residual V2, 36, 40 restricted V2, 40, 253 true V2, 39, 40, 64, 253 V2 language, 35, 36, 38, 40, 64, 112, 253 V2 syntax, 36, 40, 41, 59, 61, 64, 253 See also Inversion; Word order Variation, 5, 7, 33, 66n4, 76, 80, 82–90, 115, 117n4, 117n6, 151, 225, 251, 252 Verbs of prohibition, 206, 227–231, 234, 235, 240, 265 V2, see Verb-second Virus, 111–115

 Index  W

Whwh-expression, 163, 176, 191–194, 196, 198, 199, 200n3 wh-movement, 28–32, 167, 175, 176, 193–195, 197, 200n3, 201n9, 201n10

291

wh-phrase, 29, 31, 164, 165, 169, 170, 200n3, 201n9 wh-pronoun, 164, 168, 176 Word order, 4, 23, 24, 31, 37, 39, 41, 54, 61, 64, 85, 186, 253