Introduction: changing views of syntactic change

Introduction: changing views of syntactic change THERESA BIBERAUER & GEORGE WALKDEN Department of Theoretical & Applied Linguistics, University of Cambridge, and Department of Linguistics & English Language, University of Manchester NB: This is a post-print version reflecting changes made in the peer review and editing process, but not the publisher’s PDF. This chapter appeared in Biberauer & Walkden (eds.) (2015), Syntax over time, 1–13 (Oxford University Press); when citing, please use the page numbers given there. The publisher should be contacted for permission to re-use or reprint the material in any form. 1. INTRODUCTION There is no such thing as syntactic change, and this is a book about it.1 In some approaches and frameworks, change is seen as integral to the very nature of language: for instance, Bailey’s (1981, 1996) Developmental Linguistics, or Hopper’s (1987) Emergent Grammar. By contrast, generative work on historical linguistics as represented at the Diachronic Generative Syntax (DiGS) conferences has led to a picture in which, in at least three respects, syntactic change does not exist.2 1 The first of these respects relates to the object of study itself. Within the Chomskyan tradition since Aspects (Chomsky 1965), this object has always been the competence of the individual speaker-hearer, or I-language (Chomsky 1986); proposed larger-scale entities, such as Saussure’s (1979 [1916]) langue, or the ‘community grammar’ of Weinreich, Labov & Herzog (1968), have always been firmly rejected as incoherent or undefinable. But if such entities do not exist, then how can they change? Hence, Lightfoot (1991: 162) takes the position that when we speak about a change in, say, the syntax of French, we are adopting ‘a convenient fiction permitting the statement of certain generalizations’. Hale (1998: 2–3) argues that change should be understood simply as a set of differences between two individual grammars. In the introduction to a previous DiGS proceedings volume, Crisma & Longobardi (2009: 5) attempt to understand these relations between grammars and partially formalize them as historically significant relations, or ‘H-relations’.3 However, this relational conception of ‘change’ bears little resemblance to our pretheoretical understanding of the notion. In principle, of course, individual grammars are subject to change, and in this limited sense, syntactic change can be said to exist. However, one of the results of research into language acquisition has been the finding that, after the acquisition period, the possibility for the internalized grammar 2 of an individual to change is severely restricted (see e.g. Clahsen 1991; Meisel 1995, 2011). Examples of syntactic change during the lifespan of individual speakers have been presented: Wagner & Sankoff (2011) present findings indicating that speakers of Montreal French have moved away from the periphrastic aller ‘to go’ + infinitive in favour of the inflectional future tense over a 13-year period, Tagliamonte & D’Arcy (2007) have demonstrated an increase in the frequency of use of be like quotatives in Ontario speakers of English over 7 years (see Haddican, Zweig & Johnson, this volume, on change in the syntactic and semantic components of this construction). This indicates that some plasticity remains; nevertheless, “virtually all of the evidence for ongoing linguistic change in adulthood derives from an increase in frequency” of one of two pre-existing synonymous options (Tagliamonte & D’Arcy 2009: 61) rather than from the introduction or loss of a discrete grammatical possibility (see in this connection also the striking findings of Heycock, Sorace, Hansen & Wilson (2013) regarding the on-going loss of V-to-T movement in modern-day Faroese; as Heycock et al. show, this option is in fact initially produced and accepted by pre-school children, before declining as they get older, with ten-year olds exhibiting adult production and judgements). So true syntactic change within individual grammars, while a logical possibility, may not exist in practice either. 3 Even assuming that some notion of syntactic change beyond the individual is coherent, another perspective on the non-existence of syntactic change is provided by the Inertial Theory of Longobardi (2001). Under this perspective, it is hypothesized that syntactic change must be ‘a well-motivated consequence of other types of change’ and ‘may only originate as an interface phenomenon’ (2001: 278).4 In other words, when syntactic change occurs, it is not primitive but always parasitic on a change in other linguistic domains, for instance a phonological, morphological, or semantic innovation. This is a specific instantiation of a general principle summarized by Keenan (2002: 149) as ‘Things stay as they are unless acted upon by an outside force or DECAY’ (see also Keenan 1994, 2009). This idea, though not uncontroversial (see Walkden 2012, and Maling & Sigurjónsdóttir, this volume), has spurred 21st-century diachronic syntacticians to investigate interactions between different linguistic modules, and the study of these interactions is a major focus of this volume. Thirdly and finally, theoretical developments within the Principles & Parameters framework have independently led to the conclusion that syntactic change does not exist. Early Principles & Parameters work hypothesized that there existed a set of syntactic principles that could be parameterized: for instance, Subjacency (stated in Chomsky 1973 as a condition; later reformulated as a principle) states that there exist 4 bounding nodes inhibiting movement, and is parameterized as to which nodes count as bounding nodes in a given language. With the move to Minimalism, Chomsky (1995a: 131) adopted a suggestion originally made by Borer (1984): the syntactic computational component of human language is invariant, with inter-speaker variation existing only within the lexicon, and perhaps only within the functional subset of the lexicon. This proposal, dubbed the Borer-Chomsky Conjecture by Baker (2008: 353), has an immediate conceptual advantage, lucidly stated by Borer (1984: 29): The inventory of inflectional rules and of grammatical formatives in any given language is idiosyncratic and learned on the basis of input data. If all interlanguage variation is attributable to that system, the burden of learning is placed exactly on that component of grammar for which there is strong evidence of learning: the vocabulary and its idiosyncratic properties. The consolidation of linguistic variation in the lexical component has an obvious consequence for the study of syntax diachronically: if there can be no such thing as lexiconindependent syntactic variation, then there can be no such thing as lexicon-independent syntactic change, and developments previously classed as syntactic changes must be 5 reconceptualized as lexical changes. Interestingly, the same conclusion results if we adopt Chomsky’s more recent argument in favour of the idea that parameters may in fact be exclusively located at PF, the interface leading to the sensorymotor systems. On this view, there is once again, no variation internal to the syntactic component, with ‘diversity resulting from complex and highly varied modes of externalization, which, furthermore, are readily susceptible to historical change’ (Berwick & Chomsky 2011). Here, then, ‘syntactic change’ is actually morphophonological. If, as we have suggested here, developments within generative syntactic theory have led to the view that syntactic change does not exist, one might conclude that the theory is worthless for the investigation of diachronic issues, or, at the very least, question the validity of continuing to adopt it. We would argue, however, that the opposite is true, and that the perspectives afforded by these theoretical developments have shed new light on the phenomenon of syntactic change as pretheoretically understood. In the remainder of this introduction we will explain why we think this is the case. Mirroring the structure of the volume, the three following sections outline three main directions in which we feel progress is being made: lexical, morphological, and informationstructural interactions. 6 2. SYNTAX AND THE LEXICON Though the adoption of the Borer-Chomsky Conjecture, in which syntactic variation is understood as lexical, may at first glance seem to be a trivial terminological matter, we would argue that it has in fact led to advances in understanding and to a widening of the field’s scope of inquiry. In early DiGS work, the focus was overwhelmingly on macro-level properties of languages; thus the proceedings volume from the first DiGS conference, Battye & Roberts (1995), had clause structure as its theme, and was heavily focused on the verb-second property, the null subject parameter, the Head Parameter and related issues. These properties are still the subject of detailed work in the current literature, though the focus has shifted towards the interpretive correlates of these different constructions (see section 4). Complementing these big-picture issues, however, the DiGS literature has increasingly seen investigations of the syntactic properties of individual lexical items and their interactions with the larger system of which they form a part, with authors making existential, rather than universal, claims about the languages under investigation. This development, a natural consequence of the Borer-Chomsky approach to syntactic variation, is paralleled by a renewed focus on microvariation in synchronic syntax along the lines sketched in Kayne (2005a). It also opens the door to a better formal understanding of 7 processes of grammaticalization, which fell largely outside the remit of traditional ‘macrodiachronic’ syntax: foundational works here, arguing that grammaticalization can be understood as change in the featural makeup of lexical items, include Roberts & Roussou (1999, 2003) and van Gelderen (2004, 2011). This focus is a hallmark of 21st-century DiGS work, with an entire proceedings volume (Batllori et al. 2005) themed around the relationship between grammaticalization and parametric variation. This ‘microdiachronic syntax’ is also well represented in the present volume by the six chapters that constitute Part 1, Syntax and the Lexicon. Light, for instance, is a treatment of expletives in the West Germanic languages, with particular reference to Early New High German and Old English, and is a fine example of how careful analysis of the properties of individual lexical items can inform bigger-picture issues. The Early New High German da and its Old English cognate þær are both analysed as SpecTP expletives. As a consequence, Light argues that the EPP, as a requirement for subjects to fill SpecTP, was more broadly active across early Germanic than previously thought. The key tool of the DiGS trade has always been careful philological and corpus-based study of historical texts. Finegrained microdiachronic research, however, necessitates the development of new ways of studying syntactic variation and 8 change. Maling & Sigurjónsdóttir contribute to a lively debate about the Icelandic ‘new’ construction, which has been analysed as both a passive and an impersonal. Their aim is to assess various more nuanced predictions of the two analyses, taking into account how the new construction interacts with different lexical items such as bound anaphors and nonagentive verbs. To this end, they employ the methodology of dialect syntax, basing their conclusions on acceptability judgement questionnaires. The results – demonstrating extensive variability – inform our understanding not only of modern Icelandic, but of syntactic change in general, on the basis that ‘the same mechanisms which operated to produce the large-scale changes of the past may be observed operating in the current changes taking place around us’ (Labov 1972: 161). Haddican, Zweig & Johnson similarly focus on a present-day construction undergoing change, namely the be like quotative in English. Their chapter supplements earlier corpus-based studies with experimental results, demonstrating that be like, originally stative, is increasingly acceptable for younger speakers with an eventive, quotative function; nevertheless, they argue that the spread of stative and eventive be like is part of a single process of change along the lines of the Constant Rate Hypothesis of Kroch (1989). The analysis of the construction – involving a null lexical item SOMETHING (see Kayne 2007) – furthermore has interesting implications for the 9 interaction of syntax, semantics and the lexicon. Clearly, both dialect-syntactic and experimental methodology constitute a valuable addition to the diachronic syntactician’s toolbox. Prepositions are of special interest from the point of view of the Borer-Chomsky Conjecture, since they have always occupied an uneasy middle ground between functional and lexical categories. Within Minimalism, the study of the fine structure of prepositions and prepositional phrases has flourished, particularly in work by Svenonius (2008, 2010), applied to the history of English by Waters (2009). Hegedűs extends this approach to the grammaticalization of postpositions from nouns in the history of Hungarian, arguing that here, too, we see a clear case of an ‘adposition cycle’. Using a structure proposed by Svenonius, Hegedűs demonstrates that the grammaticalization process does not proceed directly from noun to postposition but rather through an intermediate category of AxialPart. This chapter, then, offers an excellent illustration of how a theoretical approach that draws fine-grained featural (categorial) distinctions allows us to understand grammaticalization in a more articulated fashion than is usual in more traditional approaches. As pointed out in Roberts (2010), this type of approach may also facilitate novel insight into notoriously challenging synchronic and diachronic questions of gradience and gradualness. 10 Negation has always been the poster-child for formal approaches to cyclical change, and the literature on the subject from the 21st century alone is extensive: see Larrivée (2011) and Willis, Breitbarth & Lucas (2013) for overviews. In the present volume, too, negation is well represented. The chapters by É. Kiss and Martins both seek to understand the emergence of negative indefinites, in Hungarian and Portuguese respectively. The se-indefinites discussed by É. Kiss originally arose through univerbation of a negative particle with a (nonnegative) indefinite, but over time their negative semantics were bleached, such that in Modern Hungarian they are required to occur in a Negative Concord configuration with an independent negative particle. É. Kiss argues that word order changes played a role in this development. The indefinite quantifier algum discussed by Martins, on the other hand, develops into a negative indefinite without overt univerbation, and is argued to be so by virtue of its incorporation (along with the noun) into an abstract DP-internal negative head; the proposal thus has consequences for the debate surrounding the cartography of the nominal domain generally (see Alexiadou, Haegeman & Stavrou 2009, and Alexiadou 2014 for overview discussion) and the internal structure of negative DPs more specifically (see i.a. Déprez 2000, Martins 2000, Poletto 2008, and Biberauer & Roberts 2011). 11 It is clear, then, that, far from stifling inquiry into syntactic change, adoption of a lexicocentric perspective on variation has enriched diachronic syntax in terms of both methodology and empirical coverage. We began this section by highlighting the ‘macro’ orientation of the earliest DiGS work, which was very centrally concerned with properties associated with parameters – the verb-second parameter, the null subject parameter, and the head parameter, for example. Interestingly, the term ‘parameter’ is not mentioned in any of the six contributions in this section of the volume. On the one hand, one might interpret this as indicative of the notion having outlived its usefulness (see Newmeyer 2004, Boeckx 2010) Alternatively, this can be seen as a reflection of the way in which the notion has been refined in the context of Chomsky’s (2005) ‘three factors’ approach to language design (see Biberauer 2008, Roberts & Holmberg 2010, Biberauer & Roberts 2014, and the work of the “Rethinking Comparative Syntax” project more generally). On this latter view, all of the phenomena discussed in this section would instantiate lexically specified parameters of different sizes (see again Biberauer & Roberts 2014 on the question of ‘parameter size’). 3. SYNTAX AND MORPHOLOGY Exploring the interactions between morphology and syntax is not a new endeavour. Within Principles and Parameters, a 12 series of works have proposed strong (often biconditional) relationships between aspects of morphological paradigms and syntactic parameter settings. For instance, the richness of case morphology has been linked to the possibility of certain word orders or word-order variation, implementing ‘the traditional idea that inflectional morphology and positional constraints are functionally equivalent elements of grammatical structure’ (Kiparsky 1997: 461; see also Weerman 1988, 1997). Similarly, V-to-T movement has been linked to the richness of verbal morphology (Roberts 1985, 1993; Bobaljik & Thráinsson 1998; Rohrbacher 1999; Vikner 1997), as has the availability of null subjects (Taraldsen 1978; Rohrbacher 1999; Müller 2005, 2008; Tamburelli 2006). A full review of this literature can be found in Biberauer & Roberts (2010), who suggest that the properties licensing V-to-T and null subjects should be kept apart, though they retain the connection between morphology and both properties. The diachronic consequences of such deterministic interrelationships between morphology and syntax are obvious: if certain morphological changes occur, then syntactic changes will necessarily follow. Thus, if we assume (following Biberauer & Roberts 2010) that V-to-T movement requires rich tense morphology, then if such morphology is lost then V-to-T movement must also be lost. Since morphological change can often be motivated independently (e.g. through regular sound 13 change, or language contact), this gives us a ready-made causal chain for syntactic developments. A whole DiGS volume (Lightfoot 2002) has previously been dedicated to precisely this issue. This simple picture, in which morphological change triggers syntactic change, is intuitively appealing, but evidence has emerged in recent years that suggests that it is in fact too good to be true. For instance, in modern standard English, indirect objects precede direct objects in neutral declaratives, but in the history of the language this was not always the case; a strong version of the morphology-syntax link would then predict that word-order variability was possible as long as case morphology distinguished accusative and dative. Allen (2001), however, shows that the modern order did not become fixed until the last quarter of the fourteenth century, whereas the dative/accusative distinction had been lost in most dialects by the first quarter of the thirteenth. Unless we want to posit that for a period of 150 years learners entertained morphologically unsupported and thus impossible grammars, then, the deterministic relationship between morphology and syntax must be rethought. Similarly, empirical difficulties with deterministic proposals lead Holmberg (2005: 560) to propose that the correlation between null subjects and rich agreement is ‘due not to a rule or condition of (narrow) syntax … but to sentence processing’ (see also Ackema & Neeleman 2007). 14 The five contributions collected in part two of this volume all point towards the conclusion that the traditional hypothesis, in which morphology drives syntax, is at best only partially correct. Reintges, for instance, observes that there exists another logical possibility in a Minimalist framework: syntactic changes may lead to morphological changes, the opposite of the traditional view. He illustrates this ‘exogenous’ morphological change with reference to two developments in Later Egyptian, arguing that the emergence of discourseconfigurationality (itself contact-induced) triggers the grammaticalization of new evidential and modal marking and new relative tenses. Another logical possibility is for morphophonologically conditioned alternations to become syntacticized, and this is exactly what Ledgeway argues has happened in the dialects of the Salento region in Italy. In these dialects, the complementizer cu can be deleted under certain conditions. This deletion appears to be ordered after raddoppiamento fonosintattico, a process which lengthens the first consonant of a following word, suggesting that deletion is a purely phonological process. Ledgeway argues that this is correct for southern Salentino, but not for the northern dialects, for which he adduces evidence to show that raddoppiamento fonosintattico has been reanalysed as an irrealis marker. This sort of syntacticization of an originally phonological process is 15 consistent with what has been reported in the literature on the life cycle of phonological processes (see, for example, Bermúdez-Otero & Trousdale 2012). Julien looks at a case in which a morphological form takes on a new function. In some Uralic languages, a past participle can be used in the negated past without giving rise to a perfective reading. Julien argues that the introduction of a new copula was central to this development as it led to surface forms which a new generation of acquirers could analyse as including an overt realisation requirement on finiteness. Confronted with conservative, copula-lacking present perfect negative structures featuring just a participle alongside the negation element, the new generation thus concluded that these forms in fact contrast with the likewise copula-lacking, but crucially finite-marked negative presents. This resulted in the original non-finite participial form being reanalysed as a finite past form. This reanalysis of the verbal inflection marking, in turn, triggered a reordering of the heads in the clausal hierarchy, with Tense and Polarity being inverted in the innovative grammar. Morphological reanalysis, then, drives syntactic reanalysis, with a domain that has been shown to vary crosslinguistically, rather than following a universally fixed hierarchical ordering – negation and polarity (see Cinque 1999, and Laka 1994) – being affected by the change. 16 Migdalski’s chapter provides another illustration of how morphology may drive syntax. He shows that in some Slavic languages, such as Serbo-Croatian, clitics changed from being verb-adjacent to occupying second position. Migdalski relates this to the loss of tense distinctions in verbal morphology, which was contemporaneous with the shift to second-position clitics in various Slavic dialects. These changes he interprets as signalling the loss of T, a category which has been argued not to be universally present (see, for example, Bošković 2012, and also Ritter & Wiltschko in press). Strikingly, T-less languages are predicted to lack expletives, subject-object asymmetries, and, Migdalski argues, sequence of tense effects, i.e. a diverse cluster of properties of the type that one might have expected of a classical parameter. If Migdalski’s analysis is on the right track, it is clear that morphological change may have farreaching and diverse consequences. Like Allen (2001), Michelioudakis deals with a change in which the supposed syntactic effects of a morphological change seem to lag behind. While morphological dative case is lost during the Hellenistic Greek period, Michelioudakis traces the effects of abstract, inherent, interpretable dative Case into Medieval Cypriot Greek; a later reanalysis as uninterpretable is merely facilitated by, and not deterministically triggered by, the earlier morphological change. 17 What the contributions in this section signal, then, is, that despite the proliferation of work on morphology-syntax interactions over the last thirty years, much more still remains to be done. Diachronic data are of particular interest for the study of these interactions, as they allow the investigation of temporally adjacent varieties that differ only minimally. The contributions in this section represent a small step towards this goal. 4. SYNTAX, PROSODY AND INFORMATION STRUCTURE One of the most striking features of this volume is the number of contributions that go beyond the ‘core functional categories’ C, T and v to assume a more articulated functional domain. This type of work is typical of the ‘cartographic’ research programme, which has its roots in papers like Pollock (1989) but which crystallized in the late 1990s with the decomposition of the IP-domain by Cinque (1999) and of the CP-domain by Rizzi (1997); see Cinque & Rizzi (2010) for an overview. Most of the chapters in this volume exhibit the influence of the cartographic enterprise: as we have already noted, Hegedűs is inspired by cartographic analysis of the PP, while Julien makes use of an articulated IP. Most influential, however, has been the work of Rizzi (1997) on the clausal left periphery: the chapters by Reintges, Ledgeway, Hill, Walkden, Cormany, Danckaert, 18 Hinterhölzl, and Wallenberg all make explicit reference to a Rizzian left periphery in their analyses. This reflects a broader concern with information structure that is showcased in the third and longest part of this book, and also a recognition that the facts of historically attested languages – within and outside the Indo-European family – are not easily accounted for using only the limited palette of functional categories provided in Chomsky (1995a).5 Information structure as object of study sits uneasily between pragmatics, semantics and syntax: Vallduví (1992) and Lambrecht (1994) are key early works, and Krifka (2008) provides a readable introduction. That the field is still in its infancy is attested by the fact that there is still substantial disagreement over the crucial theoretical primitives involved. Some notion of topic and focus is assumed by just about everyone, though definitions differ: while the chapters in this section by Hill, Walkden and Cormany largely focus on topics, focus is the topic of the chapters by Danckaert, Taylor & Pintzuk, Hinterhölzl, and Wallenberg. In addition, contrast deserves some mention here: while Rizzi (1997) makes no reference to it, it has been argued to exist as a primitive information-structural feature by Molnár (2001) and others (e.g. É. Kiss 1998; see also the contributions to the (2010) Lingua special edition on ‘Contrast as an information-structural notion in grammar’), and Neeleman, Titov, van de Koot & 19 Vermeulen (2009) assume that [contrast] is a privative feature which can be present or absent on topics and foci. The study of information-structural interactions in formal diachronic syntax is a recent development, but one that has proven extremely productive, as the collections of papers in Hinterhölzl & Petrova (2009a), Ferraresi & Lühr (2010) and Meurman-Solin, López-Couso & Los (2012) illustrate. The crucial notions, however, are not new at all. Work by the Prague School (e.g. Firbas 1966; Halliday 1967) assumes a principle of given-before-new. A similar principle is present in Behaghel’s (1932: 4) Second Law: ‘Es stehen die alten Begriffe vor den neuen’ (old expressions come before new ones). The broadening of generative studies to include notions of information status as possible determinants of syntactic structure thus represents something of a unification of two traditions. If one assumes the existence of purely informationstructural functional projections, a further important question is: where are they located? Rizzi’s (1997) original proposal locates them at the left periphery of the clause. Belletti (2001, 2004) has argued that there also exists a ‘low left periphery’ above vP, with Poletto (2006, 2014) motivating a clauseinternal left periphery of this kind for Old Italian; Danckaert and Wallenberg further explore this possibility. Hill, in turn, proposes a further left periphery: one that lies at the edge of the 20 DP. Focusing on the development of the Romanian object marker pe, Hill shows that it is no longer a preposition, as it was in Old Romanian, but now serves to mark contrastive topics. If CP, vP and DP may each have a separate informationstructurally active left periphery, it is tempting to suggest that this possibility is definitive of non-defective phasehood. Another indication of the immaturity of studies on syntax-information structure interactions is the fact that Rizzi’s (1997) model of the left periphery has been adapted in two different, and fundamentally incompatible, ways, by Benincà & Poletto (2004) on the one hand and Frascarelli & Hinterhölzl (2007) on the other, based largely on exactly the same language (Italian). The information-structural part of Rizzi’s original split CP is given in (1), that of Benincà & Poletto in (2), and that of Frascarelli & Hinterhölzl in (3). (1) Top* > Foc > Top* (Rizzi 1997: 297) (2) Hanging Topic > Scene Setting > Left Dislocated > List Interpretation > Contrastive Focus (adverbs/objects) > Contrastive Focus (circumstantial adverbs) > Information Focus (Benincà & Poletto 2004: 71) (3) Aboutness Topic > Contrastive Topic > Focus > Familiar Topic* 21 (Frascarelli & Hinterhölzl 2007: 112–113) As can be seen, Benincà & Poletto (2004) and Frascarelli & Hinterhölzl (2007) decompose the upper Topic field in different ways. Furthermore, while Benincà & Poletto (2004) do away with the lower Topic field entirely, Frascarelli & Hinterhölzl (2007) argue that it is reserved for ‘familiar’ topics, i.e. those that are simply given information, as opposed to those which define ‘what the sentence is about’ (Reinhart 1981). If progress is to be made, tensions like these must be resolved. Walkden argues that the preverbal element in early West Germanic V3 constructions meets the description of Frascarelli & Hinterhölzl’s (2007) Familiar Topic, thus providing some small support for the notion of a Topic field in the lowest part of the CP-domain. Walkden argues that the V3 pattern is a retention from Proto-West Germanic, and that generalized V2 was most likely a later development. If a left-peripheral ‘template’ such as that proposed by Rizzi (1997), Benincà & Poletto (2004), and Frascarelli & Hinterhölzl (2007) turns out to be the right way to analyse interactions between syntax and information structure at the left periphery, a further question arises: are all of the functional projections present in all languages, or can they be syncretized in some languages, as proposed by Giorgi & Pianesi (1997) for the IP-domain? Cormany suggests that syncretism is possible 22 through a process of conflation of left-peripheral projections diachronically, drawing his evidence from a corpus study of historical Friulano. Specifically, doubling of various types of subjects becomes increasingly possible in the history of Friulano, and the emergence of the possibility of doubling each kind of subject occurs in just the order that an account based on conflation of left-peripheral projections would suggest. Little has been said about focus so far in this section, though this is probably the information-structural category on which the most has been written. The remaining chapters in part three of this volume all address accounts based on focus. Like topic, focus is likely not an irreducible notion: É. Kiss (1998), for example, distinguishes identificational focus – representing the exhaustive subset of the set of contextually or situationally given elements for which the predicate phrase holds – from information focus (or presentational focus), which simply represents new or non-presupposed information (see also Cruschina 2012 for illuminating overview discussion). Belletti (2001, 2004) then argues that identificational focus is associated with the FocP in the clausal left periphery, while information focus is associated with the ‘low’ left-peripheral FocP above vP. However, as Danckaert observes, Latin seems to counterexemplify this insofar as information foci may occur in the high left periphery of subordinate clauses in the Archaic and early Classical periods of this language. Rather than 23 abandon Belletti’s generalization, Danckaert suggests that the apparent violation in Latin occurs because the lower FocP is made unavailable by independent movement operations; movement to the higher FocP, which is normally a position for identificational foci, then represents the syntax – or, recalling Berwick & Chomsky’s (2011) take on the actual locus of what is typically described as ‘syntactic variation’, PF – making the best out of a bad situation. For Batllori & Hernanz, too, the opposition between two types of focus is crucial. Their empirical starting point is the fact that, while both Old Spanish and Old Catalan allow weak (=information) focus fronting, modern Catalan only allows it with quantifier phrases. Batllori & Hernanz analyse weak focus fronting as movement to a dedicated specifier position in the lower part of the clausal left periphery, following Benincà & Poletto’s (2004) cartography. They hypothesize that in Catalan this position was lost, and that weak focus fronting was consequently reanalysed as targeting the specifier of a Polarity Phrase just below it, accounting for the restriction of fronting to quantifier phrases in this language. As well as information structure, Hinterhölzl addresses a further important interaction: the interplay between syntax and prosody. The ambitious goal of his chapter is to reduce all word-order variation to the requirements of prosodic and information-structural conditions. Hinterhölzl argues that Old High German, like the other early Germanic languages, 24 exhibited mixed OV/VO properties, and analyses this in terms of a Kaynean universal base combined with the possibility to spell out lower copies in movement chains. The choice of which copy to spell out is then resolved with reference to prosodic requirements, e.g. whether the constituent is branching or non-branching, and information-structural requirements, e.g. given vs. new. It is clear that investigating information-structural and prosodic interactions in diachrony is a productive line of inquiry. However, it is only intellectually responsible to do so if reliable, replicable methods for identifying informationstructural categories can be developed. Taylor & Pintzuk outline such a method, annotating objects in Old English for their prosodic weight and for whether they represent given or new information. Their findings, they argue, provide evidence that a one-to-one correspondence between VO order and new information status cannot be maintained. Instead two postverbal object positions must be postulated for Old English: one associated exclusively with new information, and another that is more information-structurally promiscuous. Taylor & Pintzuk do not appeal to a cartographic template, instead representing the position for new objects as right-adjunction to TP; however, the data they present is consistent with an account in which information-structural movement is to 25 specifiers of dedicated functional projections rather than adjunction, and this is the topic of Wallenberg’s chapter. Heavy NP Shift across Germanic is the empirical domain of this chapter, which lays out a Kaynean cartographic analysis in which ‘rightward’ focus movement is to FocP at the clausal left periphery followed by remnant movement of the rest of the clause to TopP. Using a quantitative approach, Wallenberg compares this account to a traditional analysis in terms of right-adjunction to either TP or vP, and finds that the frequency of Heavy NP Shift is higher in Early Middle English than in either Old or Early Modern English – a result predicted under the traditional analysis, but not under the new one. The closing chapter in this section is something of an outlier, in that it explicitly argues against an informationstructural or prosodic analysis. Aldridge looks at pronominal object shift in Archaic Chinese, which previous authors have attempted to account for in terms of prosodically-motivated cliticization (Feng 1996) or focus movement (Djamouri 1991, 2000). Aldridge shows that these analyses do not account for the particular distribution of pronominal object shift, and presents an alternative in which the pronoun is fronted in order to value structural accusative Case. The analysis receives support from diachronic facts: pronominal object shift declined at the same time as morphological case was being lost on pronouns. The chapter thus serves as a salutary reminder that 26 appealing to information-structural or prosodic motivations is not the solution to all our problems, and that sometimes a more traditional morphosyntactic account better fits the facts. 5. SUMMARY The chapters collected in this volume show that historical work within generative frameworks is branching out. Syntax within a Minimalist framework may remain an autonomous module, but it is ultimately responsible to its interfaces. This volume showcases interactions between the syntactic component and the lexicon, morphology, prosody and information structure. Working on these interactions is not a simple task, and requires interdisciplinary knowledge as well as new methods for investigating change. The results of such work, however, can shed new light on previously poorly-understood syntactic alternations both synchronically and diachronically. Historical work of this kind can also contribute to the ongoing investigation of the Strong Minimalist Thesis: the claim that language is an optimal solution to legibility conditions (Chomsky 2000: 96). Studying lexical, morphological and information-structural interactions will help us establish what these legibility conditions are, and how individual languages are configured to meet them in different places and at different times – whether or not syntactic change, in the narrow sense, really exists. 27 Acknowledgements This volume has its origins in the 12th Diachronic Generative Syntax Conference, held in Cambridge, 14th–16th July 2010. The editors would like to express their gratitude to the staff of Queens’ College, Cambridge, to the conference helper, Tim Bazalgette, and particularly to our co-organizers, Alastair Appleton, Elliott Lash, Dimitris Michelioudakis, and Ioanna Sitaridou. Theresa Biberauer also gratefully acknowledges the financial support of European Research Council Advanced Grant No. 269752 “Rethinking Comparative Syntax” (ReCoS). Thanks, too, to Anna Hollingsworth for her careful and efficient creation of the index to this volume. 1 The first line of this introduction is a homage to Shapin (1996: 1). 2 Cf. also, from a different perspective, Coseriu (1985). The Diachronic Generative Syntax (DiGS) conference series, of which the Cambridge instalment marked the 20th anniversary, shares a set of core principles: an emphasis on rigorous synchronic description, an emphasis on reliable and wellunderstood data, and scepticism towards independent diachronic processes. A clear introduction to, and manifesto for, the “DiGS approach” can be found in Whitman, Jonas & Garrett (2012). In this introduction we focus only on those trends in the field that have made themselves apparent over the 28 last few years and that seem to us to be the most exciting directions of research. 3 Though it is important to recognize that, from an I-language perspective, such relations are just as fictional and inchoate as community grammars, as they depend crucially on the primary corpus available to the learner, itself not easily definable or delimitable. See Walkden (2014: 31–3) for discussion. 4 Interestingly, Stockwell (1976) makes exactly the same claim a quarter of a century earlier: ‘the notion of ‘pure syntactic change’ ... implies that there exists a class of formal changes which are not motivated by semantic change or phonetic change. I do not believe that is possible.’ 5 Chomsky (2008: 143) explicitly states that CP in his work is shorthand for the clausal region that Rizzi (1997) refers to as the left periphery. Nevertheless, approaches to information structure have been developed within Minimalism that do not rely on an expanded cartographic left periphery: see Neeleman, Titov, van de Koot & Vermeulen (2009), Fanselow & Lenertova (2011), and Abels (2012), among others. Mensching (2012) is a recent attempt to capture information-structural generalizations in diachrony without additional functional projections. 29