Chapter 27: Principles and Parameters Adam Ledgeway and Ian Roberts 27.1 Introduction: Principles and parameters The Principles-and-Parameters (P&P) approach to cross-linguistic variation was first developed by Chomsky and his associates in the early 1980s (see in particular Chomsky (1981), and, for more general introductions, Roberts (1996), Baker (2001); see also discussions in §7.2, §13.5, §16.4.1, §28.2). The leading idea is that Universal Grammar (UG) contains an invariant set of principles associated with parameters which define the space of possible variation among actual languages. Taking the principles to be innately given, and the parameters to be triggered by salient parts of the primary linguistic data (PLD) for language acquisition, this approach was held to be a major step in the direction of explanatory adequacy (in the sense of Chomsky 1964), since language acquisition could be seen as setting the parameters of the native language on the combined basis of the innate UG and the triggering aspects of the PLD. To give a concrete, if rather simplified, example: we know that languages can be divided into those which have unmarked VO order, e.g. English, and those which have OV order, e.g. Japanese (cf. also the discussion of Romance and Latin in §27.3 below). On the classical P&P view, the notion of ‘verb’ is given by the universal theory of syntactic categories, the notion of ‘object’ is given by the universal theory of grammatical functions, and the idea that the two combine to form a VP is given by the universal theory of phrase structure. These are all taken to be reflexes of UG principles. But experience tells the child which order of O and V inside VP is the appropriate one, and so a child hearing Japanese sets the parameter to OV, while the child hearing English sets it to VO. Parameters describe what is variant in natural-language syntax, and as such they predict the dimensions of language typology, predict aspects of language acquisition and predict what can change in the diachronic dimension. A consequence of this setting of a parameter for OV or VO on the basis of experience is a more general and abstract setting of a directionality parameter determining the orders of all heads and complements (with, again, the notion of ‘head’ and the notion of ‘complement’ being UG-defined); see Dryer (1992) for typological support for the idea that head-complement order is predicted by the order of V and O. The P&P approach was seen as a significant step forward for generative grammar, since earlier approaches (Chomsky 1973; 1975; 1977) had defined UG as a grammatical metatheory specifying a broad format for rules and some general principles on rule application (island constraints, etc.), a particular grammar as a system of language-specific, construction-specific rules, and language acquisition as rule induction. This theory offered little hope for insights into either language typology (see chapter 30, this volume) or language acquisition (see chapter 18, this volume), and the P&P approach stood in stark contrast to this from its inception. The immediate consequence of P&P theory in the 1980s was an explosion of formal syntactic work on a wide range of the world’s languages. Comparative work became the norm. The stimulus to comparative syntactic research led to the postulation of a number of parameters, without much attention being paid to the format for parameters. This initial conception gave rise to a rather arbitrary-looking collection of parameters: the Null Subject Parameter (Taraldsen 1978; Rizzi 1982), a parameter determining the Case properties of Prepositions (Kayne 1984), the head-directionality parameter (Hawkins 1983; Koopman 1984; Travis 1984), V-movement parameters (Emonds 1978; Pollock 1989; den Besten 1983), the overt vs covert nature of wh-movement (English vs Chinese: Huang 1982) and non-configurationality (Hale 1981; 1982; 1983). Partly inspired by the first postulation of the minimalist programme in the early 1990s, with its emphasis on formal features as the driving force behind derivations, a significant shift took place in the conception of the locus for parameters: parameters were thought to be specified in the (functional) lexicon, rather than directly on UG principles (a variant of this proposal had earlier been put forward by Borer 1984:29). More precisely, Chomsky (1995) proposed that parameters be viewed as being specified by the formal features of functional heads. This view of parameters naturally leads to the concept of microparametric syntax (Kayne 2005b), according to which there is a rather large number of parameters, each responsible for a fairly small point of difference between grammars (e.g. that past participles agree with fronted wh-words in French, but not in Spanish; see §27.4.1.1). The microparametric view, although not uncontroversial (see Baker 2008b) has become the dominant one in current formal comparative syntax. This view was enshrined in what Baker (2008b:156), acknowledging the twin sources of the approach, called the Borer-Chomsky Conjecture: 1 All parameters of variation are attributable to differences in the formal features of particular items (e.g. the functional heads) in the Lexicon. The precise set of formal features of functional heads remains undetermined and in particular no satisfactory intensional definition of this set of elements has been given (but see Biberauer (2011) for a possible intensional characterization of formal features). A partial extensional characterization, however, includes at least categorial features (N, V, etc.), φ-features (person, number, gender, etc.), abstract Case features (nom, acc, etc., ) and movement-triggering features (EPP, EF, etc.). 27.2 Macro- and micro-parameters As mentioned above, the Borer-Chomsky Conjecture was taken as defining the microparametric approach: each functional head in each language can be characterized as having its own set of formal features; this approach does not necessarily imply the clustering of surface phenomena that was characteristic of the Government-Binding view of parameters (for very illuminating discussion of this point, see Kayne 2005b). A contrasting approach was pursued by Baker (1996; 2008a) (see also Bošković 2008; Huang 2015), who argued for the existence of macroparameters, parameters which may be directly associated with UG principles (unlike the formal features of the Borer-Chomsky Conjecture) and which profoundly impact on the overall nature of a grammatical system. The Polysynthesis Parameter of Baker (1996) was of this kind, being formulated in relation to a general notion of ‘argument visibility’, a very general requirement on the formal realization of the semantic arguments of a predicate, and positing two quite distinct ways of satisfying this requirement: one in terms of syntactic configurations, the other in terms of the formation of complex words. As such, the setting of this parameter had deep and ramified consequences for the grammatical system; indeed, Baker (1996:3) connects this concept of macroparameter with Sapir’s (1921) notion of the ‘genius’ of a language (we will come back to this last idea in §27.5). Although much recent work in comparative syntax has been largely microparametric in character, with the Borer-Chomsky Conjecture often implicitly or explicitly assumed, the notion of macroparameter has not been conclusively shown to be unfounded (Newmeyer (2005) criticizes the GB notion of parameter in general, and while this could be construed as a criticism of Baker-style macroparameters it does not seem to have been explicitly intended as such). Moreover, in his general discussion and defence of the microparametric approach, Kayne (2005b) explicitly asserts the value of work on macroparameters. In this chapter we follow the lead taken by Kayne (2005b:10) and developed in particular by the Rethinking Comparative Syntax (ReCoS) project (see note 2 for details and references; see also in particular §§7.3-4) in combining the two approaches (see also Huang and Roberts to appear). We hope to show that this can be valuable in understanding aspects of syntactic change. 27.2.1 The microparametric approach The microparametric approach, as articulated by the Borer-Chomsky Conjecture, has various advantages, which we now briefly review (for further discussion, see Roberts 2012; 2014a). First, the microparametric approach imposes a strong limit on what can vary. Limiting possible variation to the formal features of functional heads has the consequence that various imaginable parameters, which could have been countenanced under a GB approach, cannot be formulated. One important case is the ‘arity’ of Merge, i.e. the number of syntactic objects this operation can combine at a time. Merge is standardly taken to be binary, and indeed this may follow from its very nature as an optimal formal operation (Watumull 2015). Most importantly, External Merge at least does not seem to be regulated by formal features of functional heads. Interestingly, early versions of the GB ‘(non)configurationality parameter’ countenanced exactly the possibility of what we would now call n-ary Merge, in treating non-configurational languages as having ‘flat’, i.e. n-ary branching, structures (Hale 1981; see §§27.3-4). The nature of (External) Merge is such that the Borer-Chomsky Conjecture could not allow such an option; note that this implies that whatever the intensional characterization of the formal features of functional heads turns out to be, it cannot include a feature such as [binary Merge]. A second advantage, pointed out Borer (1984:29), is that associating parameter values with lexical entries (of functional heads) reduces them to the one part of a language which clearly must be learned anyway: the lexicon. Note that this is true even if, perhaps especially if, the domain-specific innate component of language in first-language acquisition is radically reduced, as frequently suggested in the context of the minimalist programme (cf. Chomsky 2005; 2007). Third, the Borer-Chomsky Conjecture allows us to formulate parameters in a very simple and appealing way, along the following general lines: 2 For some formal feature F and some parameter P, P = F. Note that (2) effectively states the identity of formal features and parameters. We will see various examples of the general schema in (2) as we proceed. This simplicity of formulation in turn makes possible a statement of parametric variation at the UG level which relies on the logic of underspecification: 3 a For some formal feature F, -F is the default value of P. b P has +F when triggered (i.e. under specified conditions), -F elsewhere. c +F is the marked value of P. By effectively treating formal features/parameters as privative in this way, we are able to derive a very simple formal approach to markedness. Before looking at examples of microparametric change we need to give a definition of a microparameter. Biberauer and Roberts (2012b) give the following definition: 4 Definition of a microparameter: For a given value vi of a parametrically variant feature F: a small subclass of functional heads (e.g. modal auxiliaries, pronouns) shows vi. An example of microparametric change is the development of the class of English modals in the 16th century. It is well-known that the class of English modals emerged through grammaticalisation at around this time (for discussion and analysis, see Lightfoot 1979; Warner 1993; 1997; Roberts 1985; Roberts and Roussou 2003; and also §15.3, §18.3, §23.3.1). In general, the definition of microparameter in (4), combined with the general characterization of grammaticalization given in Roberts and Roussou (2003) as categorial reanalysis of a member of a lexical category as a member of a functional category (or of one functional category as another; see also van Gelderen 2004; 2011; and discussion in §1.3.1, §4.4 of this volume, and note 1 below) implies that grammaticalization is typically microparametric change. As observed in Traugott (1972), there is some variation among the modals, but the basic line of development can be summarized as follows: in Middle English modals were lexical verbs taking infinitive clausal complements (in fact, they were probably ‘restructuring verbs’ since they triggered verb (projection) raising in Old and Middle English; see Biberauer and Roberts 2008). As one would expect of lexical verbs, the Middle English premodals were able to appear in non-finite forms: 5 I shall not konne answere (1386, Chaucer, Roberts 1985:22) ‘I shall not can [be able to] answer.’ By around 1550, modals had become restricted to finite contexts and, for the most part, only appear with VP complements. Roberts and Roussou (2003:40f.) propose that the following structural reanalysis took place at roughly this time: 6 [TP it [T may [VP (may) [TP (it) happen ]]]] > [TP it [T may [VP happen ]]] Here we see a categorial change: the modal was a V in the earlier grammar, but it is a realisation of T (or of some relatively high functional head) in the later grammar.[footnoteRef:1] [1: Cinque (2006) argues that restructuring verbs in Italian are functional heads. If West Germanic verb (projection) raising triggers are assimilated to this class (see Wurmbrand 2015), then the OE and ME premodals were already functional heads. In these terms, the 16th-century reanalysis can be viewed as reanalysis from one (class of) functional heads to another, placing them higher in the inflectional field than formerly. This higher position is one which is required to be finite, as is generally the case for epistemic modals across languages. On this approach, these finiteness requirements are the consequence of ‘high’ merger rather than any semantic property. For a cartographic analysis of Modern English modals, see Biberauer and Roberts (2015a). Note that an analysis of the development of the modals of this kind is still consistent both with Roberts and Roussou’s approach to grammaticalization and the idea that grammaticalzsation is typically microparametric change.] Roberts and Roussou propose that the change was caused by the loss of the infinitive ending on verbs (formerly -e(n)). This took place around 1500. Prior to that time, we find forms like (7) (although they were somewhat rare in the 15th century): 7 nat can we seen … Not can we see “we cannot see” (c1400: Hoccleve The Letter of Cupid 299; Gray 1985:49; Roberts 1993:261) Roberts and Roussou propose that the presence of the infinitival ending triggered the postulation of a non-finite T in the complement to the modal. When this ending was lost, there was no evidence for the non-finite T and hence no bar to the reanalysis in (6), which only has a single, main-clause T node. The reanalysis of the modals seems to have taken place within around 50 years of the loss of the infinitival ending (see Roberts and Roussou (2003:36-48) for more detailed discussion, and Warner (1997) for a careful discussion of the chronology of this change, a matter we have simplified here for the purposes of exposition). So we see that the class of modals was introduced by a microparametric change. Biberauer and Roberts (2012b) point out that the modals seem to have started to undergo further changes (involving, among other things, conditional inversion; see also Biberauer and Roberts in press b) in the 18th century, just 200 years after their creation as a separate class. This kind of diachronic instability, they suggest, is typical of microparametric settings. 27.2.2 The macroparametric approach Baker (2008b) offers a very interesting and, in our view, convincing defence of macroparameters. As he points out, on the microparametric view ‘there should be many mixed languages of different kinds, and relatively few pure languages of one kind or the other’ (p. 350). On the other hand, the macroparametric view predicts, falsely, rigid division of all languages into clear types (OV vs VO, etc.): every category in every language should pattern in one way or the other (we glossed over this obvious point in our presentation of the directionality parameter in the Introduction). But if we combine the two approaches, as he advocates, then we expect to find a bimodal distribution: languages should tend to cluster around one type or another, with a certain amount of noise and a few outliers from either one of the principal patterns. This is what we find in the case of word order, as the evidence from the World Atlas of Language Structures (Dryer 2013a,b) shows. Biberauer and Roberts (2012b) give the following definition of a macroparameter: 8 For a given value vi of a parametrically variant feature F, all functional heads of the relevant type share vi. They suggest that for head-directionality, the ‘relevant type’ of functional heads is all heads (in fact, assuming the approach to linearization in Kayne (1994), this might extend to all lexical heads too, a complication we leave aside here); for radical pro-drop, the relevant set is all potentially φ-feature bearing probes; for polysynthesis, all potential incorporation triggers (which may amount to all potential probes in terms of the approach to incorporation in Roberts 2010c). · Macroparametric changes, then, may affect properties such as head-directionality, radical pro-drop and polysynthesis. Strikingly, these properties strongly tend to be diachronically quite stable. Harmonic head-final order has been extremely stable throughout the history of the Dravidian languages (Steever 1998:31), and in both Japanese and Korean. The oldest texts in Japanese date from around 700-800AD (Frellesvig 2010), and so are over 1000 years old; these texts show both consistent head-finality and radical pro-drop (see also Yanagida 2005; Yanagida and Whitman 2009). The same is true for the oldest texts in Korean. Lee and Ramsey (2011:55) in their discussion of the text inscribed on a Silla-period stele, Imsin sŏgi sŏk, probably dating from the 6th century, say: · · In this text, all the Chinese characters are used in their original, Chinese meanings, but the order in which they are put together is completely different from that of Classical Chinese. The syntax is almost purely Korean. For example, instead of the Chinese construction ‘from now’, the order of the two characters is reversed, Korean-style […] Sentences end in verbs. Korean, then, also appears to have been rigidly head-final throughout its recorded history. Lee and Ramsey (2011) also indicate that the same is true for radical pro-drop. Concerning polysynthesis, Branigan (2014) makes the twin observations that (a) polysynthesis (which he analyses as multiple incorporation) is a ‘signature’ property of the Algonquian family, and (b) this family is very old and geographically widespread. Regarding (a), Branigan (2014:22) points out that ‘all Algonquian languages appear to make use of multiple head-movement in essentially identical ways’; similarly, Mithun (1991:338) observes that Algonquian languages are polysynthetic. This property seems to have been strongly conserved over millennia and across a vast geographical area (in which the Algonquian languages were in contact with many other Native American language families). Of course, it is well known that both Latin/Romance and English have undergone changes in head-directionality in their recorded history (we will say more about the Latin/Romance change in §27.3). In fact, it seems clear from recent work on the older Indo-European languages that these systems conform to a general type in showing non-rigid head-final order, second-position effects, a very active left periphery, sub-extraction from DP, null subjects and objects, synthetic verbal morphology and case inflections (on Latin, see Devine and Stephens 2006; Salvi 2011; Ledgeway 2012a,b; 2014b; in press b; Dankaert 2012; on Greek, Taylor 1990; on Sanskrit, Hale 1995; Kiparsky 1995; on Old Church Slavonic, Pancheva 2008; on Celtic, Watkins 1963; 1964; Russell 1995:300-04; Newton 2006; on Germanic, Walkden 2014:106-12; Ringe 2006:295; on Old Iranian, Skjærvø 2009:94f.; and on Anatolian, Garrett 1990). Many of these properties have been lost in the more recent history of the respective branches: on Romance, see below; Greek shows a similar overall development to Romance (although morphological case is retained as an impoverished Nominative-Accusative-Genitive system alongside a very rich article system); West Germanic (aside from the recent history of English) is somewhat conservative although North Germanic has undergone the OV>VO change (see in particular on Old Icelandic Hroársdottir 2000) but relatively innovative in DP; Slavic appears to have undergone the same change (Pancheva 2008); Celtic has innovated VS order but is otherwise somewhat similar to Romance while Indic, presumably as a consequence of long-standing contact with Dravidian (in Matisoff’s (1990) terms, these languages belong to the ‘Indosphere’), has developed rigid OV order. (The situation in the Iranian languages is more complicated owing in part to contact with Turkic, see Harris and Campbell 1995:139-41). But we can observe that (a) evidence from Anatolian in particular suggests that the parent language was head-final and (b) head-final systems tend to be stable. This leads us to ask why several branches of European Indo-European have developed in these rather similar, but, for head-final languages, atypical ways. An important factor may have been the widespread second-position phenomena in these branches of Indo-European. Following the general line of research into these phenomena instigated by den Besten’s (1983) account of Germanic verb second, we take these phenomena to involve a combination of head- and XP-movement into the left-periphery of the clause; this general activation of the left-periphery. This, combined with the development of initial complementizers (which may have been connected, see Kiparsky 1995) may have ‘destabilized’ the earlier head-final order. Hence we see what from a wider cross-linguistic perspective may be a rather unusual pattern of OV>VO change in these families. We thus concur with Baker’s conclusion that both microparametric and macroparametric variation must be countenanced. We add to his conclusion that this must hold for diachronic variation (i.e. change) as well as for synchronic variation. Furthermore, there is evidence that microparameters are less diachronically stable than macroparameters. In the next sections, we will provide further support for this view from the diachronic developments in the passage from Latin to Romance. 27.3 Combining macro- and microparameters: the Latin-Romance transition Even the most cursory of comparisons of Latin and Romance syntax reveals some fundamental changes in the Latin-Romance transition which, in typological terms, can be interpreted as involving some ‘large steps’ and, at the same time, a series of ‘smaller steps’ both in the passage from Latin to Romance and in the subsequent developments that have produced considerable differentiation across the many Romance languages and dialects. Changes of the former type have traditionally been modelled in terms of macroparameters (see §27.2.2) which, on most accounts, would include at least the following major dimensions of linguistic variation: 9 a Head directionality (Tesnière 1959; Chomsky 1981; Hawkins 1983; Travis 1984) b Configurationality (Hale 1981; 1982; 1983; Ledgeway 2012a:chs 3,5; 2012b) c Nominative/ergative alignment (Comrie 1978; Dixon 1994; Sheehan 2014) d Polysynthesis (Baker 1996) e Topic/Subject prominence (Li and Thompson 1976; Huang 1982) Of these macroparameters, only the first two are relevant to Latin-Romance developments. In terms of the head parameter, at least in its earliest attestations, Latin was predominantly head-final (10a) whereas modern Romance is head-initial (10b), with Classical Latin representing a transitional stage in which both conservative head-final (11a) and innovative head-initial (11b) orders are found (Adams 1976; Ledgeway 2012a:ch. 5). 10 a quoius forma uirtutei parisuma fuit (archaic Lat., CIL 12.7) whose beauty.nom valour.dat most.equal.nom was b la cui bellezza fu pari al valore (It.) the whose beauty was equal to.the valour ‘whose beauty was fully equal to his valour’ 11 a constantibus hominibus par erat (Lat., Cic. Diu. 2.113) resolute.abl.pl men.abl equal.nom it.was ‘[our apprehension] was equal to that of men of strong character’ b illa erat uita […] libertate esse parem ceteris (Lat., Cic. Phil. 1.34) that.nom was life.nom freedom.abl be.inf equal.acc rest.dat.pl ‘What he considered life […] was the being equal to the rest of the citizens in freedom’ In terms of structural organization, Latin has also been argued to exhibit a non-configurational syntax in which relationships between individual linguistic items are signalled lexocentrically through the forms of the items themselves (case inflections, agreement), whereas in Romance relationships between related linguistic items are encoded by their fixed positions relative to each other (Vincent 1988:53f., 62f.; 1997:149, 163; 1998a:423f.; Ledgeway 2011:§3; 2012a:ch. 3). Consequently, in Latin not only is it difficult to establish fixed orders for individual heads and their associated complements or modifiers within their given phrase (12a-b), even adjacency between semantically-related items is not a requirement (Marouzeau 1949:42; 1953:62; Ernout and Thomas 1953:162; Pinkster 1990:184-86; Oniga 2004:101-02; Powell 2010). As a consequence, we frequently find discontinuous structures such as (13a), where the adjectival modifier celeris has been fronted under focus to the left edge of the containing DP separating it from its associated nominal subsidii. In Romance, by contrast, all elements appear to have pre-established positions (12c) and the languages do not readily license hyperbaton under edge-fronting (13b). 12 a Caesar suas copias in proximum collem subducit (Lat., Caes. B.G.1.22.3) Caesar.nom his.acc troops.acc in next.acc hill.acc withdraws ‘Caesar leads off his forces to the next hill’ b copias suas Caesar in proximum collem troops.acc his.acc Caesar.nom in next.acc hill.acc subduxit (Lat., Caes. B.G. 1.24.1) withdrew ‘Caesar drew off his forces to the next hill’ c César retire ses troupes (*ses) (*César) (Fr.) Caesar withdraws his troops his Caesar 13 a legio pompeiana, celeris spe subsidii legion.nom Pompeian.nom quick.gen hope.abl help.gen confirmata (Lat., Caes., B.C. 3.69.2) assured b legiunea pomepeiană, întărită de (*rapid) nădejdea unui ajutor rapid (Ro.) legion=the Pompeian strenghtened by quick hope=the of.a help quick ‘the Pompeian legion, encouraged by the hope of speedy assistance’ However, as observed above (cf. §27.2.1), over recent decades much work has radically departed from this macroparametric view with a shift of focus on predominantly surface-oriented variation (cf. Kayne 1996; 2000; 2005a,b; Manzini and Savoia 2005), an approach well suited to modelling the ‘smaller steps’ in diachronic change. This has led to the proliferation of a remarkable number of local, low-level microparameters interpreted as the (PF)lexicalization of specific formal feature values of individual functional heads (Borer 1984; Chomsky 1995) in accordance with the Borer-Chomsky Conjecture (Baker 2008b:353). By way of illustration, consider (14a-d) where we see that, in contrast to Romance, Latin lacks functional categories in that none of the functional heads are overtly lexicalized in accordance with the traditional Latin-Romance synthetic-analytic dichotomy (Ledgeway 2012a:ch.s 2, 4; in press e). At the same time, we also observe how across Romance there is significant variation in which of the functional heads are realized and the overt distinctions they mark. For instance, only French lexicalizes all functional heads in (14), including an overt transitive/causative light verb fait, whereas Italian only optionally encodes the partitive distinction on D through the partitive article del ‘of.the’. By contrast, Romanian fails to overtly mark either of these head positions but uniquely displays robust marking on C for the realis/irrealis opposition (că vs să), otherwise paralleled in the indicative/subjunctive distinction realized on T in the Romance perfective auxiliary, in turn further distinguished by way of the have/be split (a vs fi) in Romanian (Ledgeway 2014a). In short, what we see here are minimal differences across otherwise highly homogenous systems which can be read both horizontally and vertically as cases of synchronic and diachronic microvariation, respectively. C T v D 14 a Dico/Uolo Ø eum Ø Ø coxisse Ø panem (Lat.) b Je dis/veux qu’ il a/ait fait cuire du pain (Fr.) c Dico/Voglio che ha/abbia Ø cotto (del) pane (It.) d Spun/Vreau că/să a/fi Ø copt Ø pâine (Ro.) I.want/say that(Realis/Irrealis) him/he hasind/(be)sbjv made bake(d) some bread ‘I want him to have/I say that he has baked some bread’ Arguably, then, any account of the Latin-Romance transition must make reference to changes of both a macro- and microparametric kind (cf. §27.2). Approaches couched narrowly in terms of macroparameters would lead us to expect successive stages of languages to rigidly fall into one of a few ‘pure’ types, while microparametric approaches would lead us a priori to expect wildly ‘mixed’ types. As observed by Roberts (2010b:24f.), neither scenario correctly captures the relevant facts about the Latin-Romance transition. For example, a purely macroparametric view would incorrectly lead us to expect Romance varieties to present properties like those in (15), in which among the macro-dimensions of variation in (15a-e) some (viz. 15a-b) might show change with respect to Latin, whereas low-level micro-properties such as those in (15f-j) are not expected to diverge at all from Latin, contrary to fact: 15 a Head-initial: (S)VO, postnominal genitives b Configurational: grammatically fixed word order c Nominative-accusative d Non-polysynthetic e Subject prominent f Absence of functional categories: articles, clitic pronouns, auxiliaries, few complementizers (cf. Ledgeway 2012a:ch.4; in press b) g Rich inflectional agreement, null arguments (including objects, cf. Vincent 2000) h Predominant infinitival complementation, notably accusative with infinitive (cf. Herman 1989; Greco 2012) i (Imperfective) synthetic passive/middle voice (cf. Cennamo in press) j Simple preverbal negation (cf. Molinelli 1988; Willis, Lucas and Breitbarth 2013) By contrast, under a purely microparametric view not only do we expect Romance varieties to present those properties which have actually changed with respect to Latin combining small-scale changes (16f-j) with more far-reaching large-scale developments (cf. 16a-b), but we should also expect the relevant variation to be greater and less constrained, with some Romance varieties displaying unattested clusters of properties which freely mix features of Latin and attested Romance syntax. Yet, we do not find fictitious varieties such as *Latinalabrese ( all F(p) > some F(p) (for F a feature and p some grammatical behaviour). More specifically, functional heads increasingly display a disparate behaviour in relation to particular feature values which may, for example, characterize: (i) a naturally definable class of functional heads (e.g. [+N], [+finite]), a case of mesoparametric variation; (ii) a small, lexically definable subclass of functional heads (e.g. pronominals, proper nouns, auxiliaries, unaccusatives), a case of microparametric variation proper; and (iii) one or more individual lexical items, a case of nanoparametric variation. [2: Recent publications of the ReCoS project (http://recos-dtal.mml.cam.ac.uk/) include Biberauer, Holmberg, Roberts and Sheehan (2012), Biberauer and Roberts (2012a,b; 2015a,b; in press a,b), Roberts (2012). Biberauer, Roberts and Sheehan (2014), Sheehan (2014). See also Ledgeway (2013; 2015; in press d,f).] [Insert example (20) HERE] In light of these assumptions, consider again the head parameter. In §27.3 above we saw how at the macroparametric level the passage from Latin to Romance is marked by a reversal in the head parameter (cf. 10-11), from which, following Ledgeway (2012a:ch. 5; 2012b; 2014b; in press a), the perceived effects of configurationality can also be ultimately derived. This development can therefore be modelled by way of the parameter hierarchy in (21). [Insert example (21) HERE] The hierarchy in (21) based on Roberts (2012:320-23) makes the plausible assumption that head-initiality represents the unmarked and least costly option, as made explicit in many recent structural analyses following Kayne (1994) where head-initiality instantiates the basic underlying order, whereas the derivationally more complex nature of head-finality obtained through roll-up movement of the complement across the head to a derived (inner) specifier represents the more marked option (cf. Ledgeway 2012a:ch. 5; Biberauer, Holmberg, and Roberts 2014). On this view, in Romance the effect of the parameter is unmistakably ‘macro’ in that all functional heads in Romance are unequivocally aligned with the head-initial setting, whereas in Latin the parameter fluctuates between both settings as a result of its occupying an intermediate position in the gradual shift from head-finality to head-initiality. This oscillation between head-last and head-first structures in the history of Latin can be captured along two axes of variation (cf. Ledgeway 2012a:236), the first in terms of diachronic variation (head-last (archaic Latin) head-initial (early/late Latin)) and the second in terms of diaphasic (and no doubt diastratic and diamesic) variation (head-final (formal, literary) vs head-initial (subliterary, colloquial)). The facts can therefore be interpreted in terms of a progressive reversal of the head parameter from a regular head-final setting (as in languages like Japanese) towards a head-initial setting, with Classical Latin displaying an ambivalent behaviour on account of its non-uniform characterization in relation to these two dimensions of variation (cf. 11a-b), namely, non-archaic (predominantly head-first), but formal and literary (predominantly head-final). As we move down the hierarchy in (21) we come across increasingly more marked and restricted linguistic options of the microparametric type, including those whose effects have been described above as involving meso- and nanoparametric distributions. For example, we find languages like modern German which have been argued to present a ‘mixed’ (viz. mesoparametric) setting for the head parameter (Biberauer, Holmberg, and Roberts 2014), with head-final roll-up orders in the verbal domain (cf. 22a) but head-initial orders in the nominal domain (cf. 23b) ultimately statable in terms of a distinction between [±V] heads (we take C to be excluded from this class of [+V] heads). This sort of behaviour also finds a parallel in early Romance (e.g. Old French; cf. Bauer 1995:107-11) where it has been observed that (S)OV order exceptionally survives in subordinate clauses (cf. 23a-b). One natural way to frame this generalization is in terms of a strictly microparametric representation whereby roll-up movement is limited to a particular subclass of V-heads, namely those marked by the lexical specification [+subordinate]. Diachronically, we thus witness a movement down the hierarchy whereby the distribution of roll-up movement in the verbal domain, still systematic in Latin in embedded contexts (Charpin 1989; Ledgeway 2012a:177-79), becomes increasingly constrained and infrequent before eventually falling out of the hierarchy entirely by the modern Romance period. 22 a daß Paul [[das Buch gelesen] hat] (Ger.) that Paul the book read has b Das Buch von Paul (Ger.) the book from Paul 23 a in quant Deus sauir et podir me dunat (OFr., Strasbourg Oaths) as God wisdom and power me grants b cum om per dreit [[son fradra saluar] dift] (OFr., Strasbourg Oaths) as one by right one’s brother protect.inf must Of course, before dropping out of the hierarchy, it is not unusual for once-productive options to persist as isolated or sporadic lexical archaisms on the margins of the system. Arguably, such nanoparametric variation is visible in a small number of individual Romance lexical items which may residually present conservative roll-up orders, including specific (perhaps lexicalized) uses of the Italian prepositions malgrado ‘despite’ (cf. head-final mio malgrado lit. ‘1sg.gen despite (= against my will)’ vs head-initial malgrado Gianni ‘despite Gianni (= against Gianni’s will)’) and the French adposition durant ‘during’ (cf. head-final/-initial sa vie durant/durant sa vie ‘throughout/during his life’). 27.4.1 Diachronic microvariation: Some Romance and Germanic examples In what follows we examine the development of several phenomena across a number of different Romance and Germanic varieties which show how minimal differences among otherwise highly homogenous ‘systems’ can be used to investigate microvariation along the diachronic axis in order to better understand what precisely may vary and how such variation may be implicationally structured in relation to the predictions of parametric hierarchies like (20). The overall picture highlights an unmistakable tension between the demands of detailed empirical description on the one hand, which forces us to assume many distinct featural (viz. microparametric) instantiations of different functional heads, and the desire to provide a principled explanation within the limits of a maximally constrained theory of UG on the other. 27.4.1.1 Romance past participle agreement The distribution of agreement of the active past participle displays a number of patterns across Romance (cf. Smith 1999; Loporcaro 1998; in press; Manzini and Savoia 2005, II:553-96; Ledgeway 2012a:317f.), a representative sample of which is exemplified in (24a-g): 24 a La manzana, la había [vP [Spec la] comido] la (Sp.) the.fsg apple.fsg it.f= I.had eaten.msg ‘I had eaten the apple.’ b proi seme [AgrOP [Spec proi] magnite] lu biscotte / proi so [vP [Spec __] pro are.1pl eaten.mpl the.msg biscuit.msg / pro am magnite] li biscutte (Ariellese) eaten.mpl the.mpl biscuits.mpl ‘We have eaten the biscuit. / I have eaten the biscuits.’ c Avètz [vP [Spec __] presas] de fotòs? (Occ.) you.have taken.fpl of photos.fpl ‘Did you take any photos?’ d La clé que j’ai [vP [Spec la clé] prise] la clé (Fr.) the.fsg key.fsg that I.have taken.fsg ‘The key which I took’ e Li/Ci hanno [vP [Spec li/ci] visti] li/ci (It.) them.m/us= they.have seen.mpl ‘They saw us.’ f Los/Nos as [vP [Spec los/nos] vistos/vistu] los/nos (Lula, Sardinia) them.m/us= you.have seen.mpl/msg ‘You have seen them/us.’ g Els/Les he [vP [Spec els/les] llegit/llegides] els/les (Barcelona Cat.) them.m/f I.have read.msg/fpl ‘I’ve read them.’ Assuming active participle agreement to be the surface reflex of an underlying Agree relation for φ-features between, say, the functional head vPtP and a given nominal, we are forced to recognize at least seven different microparametric specifications for vPtP. The simplest and least constrained system is exemplified by Ibero-Romance varieties such as Spanish (24a), where vptcp quite simply never displays any agreement, failing to enter into an Agree relation with any DP. Its mirror image is the pattern of participial agreement found in the dialect of Arielli (24b) spoken in eastern Abruzzo, Italy, where the participle, and hence vptcp, simply agrees with any plural DP, be it the internal or external argument (D’Alessandro and Roberts 2010). Slightly more constrained, though still liberal by general Romance standards, is the conservative pattern found in Occitan varieties (24c) where the participle agrees with all types of DP object, a pattern further constrained in modern standard French (24d) by the additional requirement that the object DP be overtly fronted (either under object-to-subject fronting as with unaccusative structures, or under relativization and wh-fronting). In this respect, modern Italian (24e) proves even more restrictive in that, in addition to A-moved superficial subjects of unaccusatives and passives, vptcp only agrees with fronted nominals when they are represented by pronominal clitics, an option taken a stage further in Sardinian dialects (24f) where there is a further requirement that the pronominal clitic also be third person. Finally, there are varieties such as standard Barcelona Catalan (24g), where vptcp is further restricted to agreeing only with feminine third person pronominal clitics. Empirically, then, we are forced to assume as many as seven distinct featural (viz. microparametric) instantiations of vptcp across Romance, the distribution of which can be modelled in terms of a small-scale parametric hierarchy along the lines of (25), ultimately part of a larger hierarchy related to agreement and argument marking (cf. also Ledgeway 2013:189-92; Sheehan 2014): [Insert example (25) HERE] Starting at the top of the hierarchy in (25), the first question allows us to draw a simplex – arguably macroparametric – distinction between languages such as Chinese, where all functional heads systematically fail to license any form of overt agreement, and varieties like the central Italian dialect of Ripatransone where, by contrast, gender and number agreement is ubiquitous, surfacing on all categories.[footnoteRef:3] Moving down the tree, we can then ask more restrictive questions to identify more marked options regarding the ability of smaller and smaller subsets of functional heads (e.g. just those marked [+N] or [+V]) to probe for -features,[footnoteRef:4] until we eventually come to isolate the functional head responsible for active participle agreement, viz. vptcp. This is the portion of the hierarchy with which we are most interested here and where the gradual cascading effect produced by the options presented in (25) not only mirrors the gradual diachronic contraction of Romance participle agreement (for example, as late as the 19th century the distribution of Italian past participle agreement largely mirrored that of Modern French), but also highlights how variation in relation to the ability of vptcp to probe the -features of specific nominals is not uniform but, rather, licenses differing degrees of surface variation in accordance with the growing markedness conditions that accompany the available parametric options as one moves down the hierarchy. [3: For discussion of the dialect of Ripatransone, see Parrino (1967), Lüdtke (1974; 1976), Mancini (1993), Harder (1998), Ledgeway (2012a:277-86).] [4: For a recent analysis of parametric variation in relation to the distribution of subject clitics in Tuscan and northern Italian dialects, arguably instances of Agr(ement) markers, see Roberts (2014a).] In this respect, we can note that Spanish and Ariellese represent rather simple and relatively unmarked options, in that vptcp in these varieties either indiscriminately fails to probe all DP arguments or, on the contrary, systematically probes all (plural) DP arguments. Occitan varieties, on the other hand, are slightly more constrained in that vptcp only probes a subset of DP arguments, namely those marked [+acc], whereas in French there is the further proviso that the DPACC must have also undergone A- or A'-movement. In all four cases, however, we are dealing with a case of mesoparametric variation, in that the four options can be subsumed within a naturally definable class insofar as they exclusively make reference to a single functional head [D], in turn further specified for the feature [+acc] in Occitan and French (presumably un(der)specified in the case of Spanish and Ariellese) and the relevant A/A'-movement feature in French. We observe however a shift from meso- to microparametric variation as we move down the hierarchy to Italian, insofar as the relevant class of triggers for participial agreement is no longer represented tout court by a naturally definable class of functional heads (viz. [D]), but now also makes reference to a small and lexically definable subclass of Ds, namely pronominals. Arguably, in the case of Sardinian and Barcelona Catalan where this lexically definable subclass is today further broken down into the ever more marked pronominal categories of third person and, in turn, feminine, we are now entering nanoparametric territory where the relevant generalizations hold of just a handful of individual lexical items, namely Sardinian lu (msg), la (fsg), los (mpl) and las (msg) and Barcelona Catalan la (fsg) and les (fpl). 27.4.1.2 Romance auxiliary selection An area of spectacular diachronic and synchronic microvariation in Romance regards the numerous dimensions of variation characterizing the choice of auxiliary in the formation of various perfective periphrases in conjunction with the past participle. Indeed, work over recent decades in particular has brought to light an unrivalled degree of variation (for relevant bibliography, see Ledgeway 2012a:292-99, 311-17), the precise empirical limits of which still remain to be defined. While making no claims to exhaustiveness, in what follows we review how some of the major patterns of meso- and microvariation in this area and their interrelationships can be mapped in terms of parameter hierarchies (for fuller discussion, see Ledgeway in press f). We begin by considering the hierarchy in (26). [Insert example (26) HERE] (26) reveals five broad dimensions of mesoparametric variation in auxiliary distribution, the markedness and complexity of which grows as we move down the hierarchy. Question (1) serves to draw the simplest and broadest distinction between those varieties which do not show any alternation in the perfective auxiliary on the one hand and all other varieties (the vast majority) that display varying patterns of alternation between be (B) and have (H) on the other. Clearly, the simplest option is represented by those varieties which generalize one auxiliary to all perfective contexts without further differentiation (cf. Tuttle 1986:267-76; Manzini and Savoia 2005,II:759-809; Legendre 2010:188-89; Ledgeway 2012a:341-42), be it be as in many central-southern dialects of Italy (cf. Pescolanciano in 27a-c) or have as in many Ibero-Romance and (extreme) southern Italian varieties (cf. Portuguese in 28a-c). Although on cross-linguistic and theoretical grounds have has been argued, following Benveniste (1960), Freeze (1992), Kayne (1993), to be derivationally more marked than be qua the surface spell-out of an underlying loc+be structure (for recent implementations, see Roberts 2013:20-23; Ledgeway 2014a), the generalization of one or the other auxiliary is non-contrastive in these varieties and ultimately has no bearing on the mesoparametric choice in question. 27 a (mə) sɔŋgə / (tə) si / (ʦ) ɛ… məˈnuːtə/maɲˈɲɛɐtə (/laˈvaːtə) (me=) I.am (you.sg=) you.are.sg (self=) is come/eaten washed ‘I/you/(s)he have/has come/eaten/washed’ b (mə) jiva / (tə) jivə / (ʦə) jiva… məˈnuːtə/maɲˈɲɛɐtə (/laˈvaːtə) (me=) I.was (you.sg=) you.were.sg (self=) (s)he.was come/eaten washed ‘I/you/(s)he had come/eaten/washed’ c (mə / tə / ʦə) fussə… məˈnuːtə/maɲˈɲɛɐtə (/laˈvaːtə) (me= you.sg= self=) I/you.sg/he.be.ipfv.sbjv come/eaten washed ‘I/you/(s)he would have come/eaten/washed’ 28 a tenho(-me) / tens(-te) / tem(-se)… vindo/comido (/lavado) I.have(=me) you.sg.have(=you.sg) (s)he.has(=self) come/eaten washed ‘I/you/(s)he have/has come/eaten/washed’ b tinha(-me) / tinhas(-te) / tinha(-se) … vindo/comido (/lavado) I.had(=me) you.sg.had(=you.sg) (s)he.had(=self) come/eaten washed ‘I/you/(s)he had come/eaten/washed’ c teria(-me) / terias(-te) / teria(-se) … I.have.cond(=me) you.sg.have.cond(=you.sg) (s)he.have.cond(=self) vindo/comido (/lavado) come/eaten washed ‘I/you/(s)he would have come/eaten/washed’ 27.4.1.2.1 Mood and tense as determinants of auxiliary selection If a variety does present auxiliary alternation,[footnoteRef:5] then as indicated in (26) this variation can, in order of complexity, be determined by mood, tense, person, and argument structure. Beginning with mood and tense, we thus find varieties such as: (i) Romanian where auxiliary choice is entirely dictated by the realis ( have) vs irrealis ( be) mood distinction (Avram and Hill 2007),[footnoteRef:6] in that (inflected) have (viz. am, ai,…) is uniquely licensed in the present indicative (29a) and (invariable) be (viz. fi) in the present subjunctive (29b), future and conditional perfect (29c), and the perfect infinitive (29d); and (ii) the Campanian dialect of San Leucio del Sannio (Iannace 1983:72-80, 88f.; Ledgeway 2012a:342f.) where auxiliary distribution proves sensitive to tense distinctions, inasmuch as the present perfect (30a) and the future-oriented conditional perfect/pluperfect subjunctive (30b) – henceforth ‘counterfactual perfect’ – align with have and the pluperfect indicative with be (30c). [5: For reasons of expository simplicity, in (26) we informally talk of a language presenting a rule of auxiliary alternation. Following Benveniste’s (1960) seminal derivational analysis of the be vs have alternation, according to which forms of copula/auxiliary have are to be interpreted as the superficial manifestation of the incorporation of a(n abstract) locative preposition into an underlying copula/auxiliary be (cf. also Freeze 1992; Kayne 1993), this can be formalized by asking whether be is probed by the locative prepositional head (viz. [TP have (= be+P°) …be [v-VP ptcp]]).] [6: Ledgeway (2014a) argues that the relevant distinction is one of finiteness, with finite forms aligning with have and non-finite forms with be. However, the differences between the two approaches are irrelevant to the current discussion – and in any case probably amenable to conflation if, following Vincent (1998b:151-52), mood and finiteness represent sub-parts of the same overall grammatical category (cf. also Miller 2002:1, 68f.) – in that what is crucial here is that both approaches serve to draw the relevant binary split (be it [±realis] or [±finite]) across all verb forms.] 29 a Am / Ai / A / Am / Aţi / Au mâncat / plecat I.have you.have (s)he.has we.have you.have they.have eaten left ‘I/you/(s)he/we/you/they have(/has) eaten/left (…ate/left)’ b Vor / Ar fi mâncat / plecat they.will they.would be.inf eaten left ‘They will/would have eaten/left’ c Nu cred să fi mâncat / plecat not they.believe that be.sbjv eaten left ‘They don’t believe that I/you/(s)he/we/you/they have(/has) eaten/left’ d Înainte de a fi mâncat / plecat citeam ziarul before of to be.inf eaten left I.read newspaper.def ‘Before having eaten/left, I was reading the newspaper’ 30 a Èggio fatto tutto / Èggio muorto / M’ èggio lavato I.have done all I.have died me= I.have washed ‘I have done everything/died/washed’ b Si nun’ èsse muórt’ u marito nun s’ èsse mòssa da llà / if not had.sbjv died the husband not self= she.had.sbjv moved from there Chi l’ èsse mai ditto? who it= had.sbjv ever said ‘If her husband hadn’t died she wouldn’t have ever moved from there / Who would have ever thought it?’ c Èrem’ auta dice quéllo / Èra venutu / S’ era truatu nu bèllu pòstu we.were had.to say.inf that he.was come self= he.was found a nice job ‘We had had to say that / He had come / He had found himself a nice job’ Auxiliary distribution in both varieties represents the surface reflex of a relatively simple distinction between naturally definable instantiations of TAux (or maybe vAux): in the case of Romanian the auxiliary system differentially marks a binary [±realis] distinction, whereas in Sanleuciano the alternation arguably spells out a binary [±past] temporal distinction in that [+past] licenses be in the pluperfect while the [-past] specification on TAux, which unites the present and counterfactual,[footnoteRef:7] licenses have in the present and counterfactual perfects. Although both mesoparametric auxiliary options apparently make reference to a binary featural opposition, sensitivity to mood has been placed higher than tense in (26) in accordance with the assumption that, while the most primitive modal distinction involves a simple binary contrast between realis and irrealis, tense involves, following Vikner’s (1985) neo-Reichenbachian analysis, three binary temporal relations as formalized, for example, in Cinque’s (1999:81-83) functional representation TPast (= R1…S) > TFuture (R2…R1) > TAnterior (E…R2), according to which the featural specification [+past] arises from the combination of the values R1_S; R1,R2; E,R2. Furthermore, more general considerations such as the observation that all (Romance) verbs have to be specified at the very least for mood (/finiteness), while not all verbs are necessarily specified for tense, lead us to assign a more basic, and hence less marked, status to mood over tense in (26). [7: We follow here, among others Iatridou (2000) and Ritter and Wiltschko (in press), in taking the superficial past tense morphology found in counterfactuals to be ‘fake’, insofar as it does not receive a past tense interpretation, but, rather, is responsible for the licensing the counterfactual reading. See also the discussion of microparametric variation in counterfactual inversion in Biberauer and Roberts (in press b).] 27.4.1.2.2 Person and argument structure as determinants of auxiliary selection Below the modal and temporal dimensions in (26) follow those relating to person and argument structure. The former accounts for patterns such as that in (31) for the eastern Abruzzese dialect of Arielli (D’Alessandro and Ledgeway 2010; D’Alessandro and Roberts 2010) where, in the present at least, auxiliaries are distributed along the lines of a simple binary person split in accordance with a [±discourse participant] distinction (cf. Benveniste [1950]1966:228; Harley and Ritter 2002), which yields be in 1/2 persons and have in 3 persons. 31 a So / Si / A fatecate / ’rrevate. I.am you.sg.are have.3 worked.sg arrived.sg b Seme / Sete / A fatichite / ’rrivite. we.are you.pl.are have.3 worked.pl arrived.pl ‘I/you/(s)he/we/you/they have/has worked/arrived.’ In other varieties such Lengadocien Occitan we find and a conservative binary active-stative split (Ledgeway 2012a:319-23), where have surfaces in conjunction with A/SA (transitives/unergatives) subjects (32a) and be with SO (unaccusatives) subjects (32b-c). 32 a Avètz fach bon viatge? you.have made good trip ‘Did you have a good journey?’ b Soi vengut amb los amics / L’ aiga s’ èra poirida. I.am come with the friends the water self= was rotten ‘I cam with friend / The water has gone off.’ As shown in some detail in Ledgeway (in press f), auxiliary systems that operate fundamentally in terms of person and argument structure distinctions frequently blend these with modal and temporal restrictions to produce increasingly marked and complex proper subsets of person and verb class combinations. For example, in the southern Lazio dialect of Pontecorvo (Manzini and Savoia 2005,II:701f.) a [±present] temporal restriction limits the person split to the present perfect (33a), with generalization of be in the pluperfect (33b) and counterfactual (33c). Similarly, the active-stative split found in many early Romance varieties is frequently suspended in [-realis] contexts (Nordahl 1977; Ledgeway 2003; Stolova 2006), where all instances of SO may exceptionally align with have on a par with A/SA, witness the old Sicilian and Spanish contrasts in (34a-b), respectively: 33 a su / si[footnoteRef:8] / a / semə / setə / avə parˈlacə/vəˈnucə [8: 2sg. si causes initial consonantal lengthening of the following participle not indicated in (33a).] I.am you.sg.are (s)he.has we.are you.are they.have spoken/come b ɛrə / irə / ɛra / ɛraˈvamə / ɛraˈvatə / ˈɛrəɲə parˈlacə/vəˈnucə I.was you.sg.were (s)he.was we.were you.were they.were spoken/come c saˈria / sarˈrissə / saˈria / saˈrissəmə / sarˈitə / saˈriəɲə parˈlacə/vəˈnucə I.was you.sg.were (s)he.was we.were you.were they.were spoken/come 34 a li pili ià li eranu caduti / si killa dirrupa avissi caduta the hairs already to.him= were fallen if that rock had.sbjv fallen ‘his hair had already droppd out / if that rock had fallen’ b Si el sieruo que es fuydo mora mucho en casa / si ladrones que furtan if the servant that is fled stays much in house if thieves that steal de dia & de noche ouissen entrado of day and of night had.sbjv entered ‘If the servant who has fled remains a long time at home / if thieves who steal by day and night had entered’ It is crucial to note in this respect that the relevant lower-level modal and temporal contrasts introduced into such systems do not override or efface the fundamental person or verb class distinctions – hence we cannot speak of mood- or tense-based systems as in (29)-(30) above – but, rather, are embedded within the categories of person and argument structure to introduce more fine-grained person and verb class combinations. It is for this reason that person and argument structure are positioned lower than mood and tense in (26), in that mood and tense as independent determinants of auxiliary variation are not constrained, at least in Romance, by person and argument structure, whereas person and argument structure as independent dimensions of auxiliary selection are frequently augmented by the incorporation of restrictions relating to mood and tense. By the same token, auxiliary systems driven by argument structure may, in turn, also incorporate restrictions on person in addition to those on tense and mood. In this way, the hierarchy in (26) correctly models the subset and inclusiveness relations implicit in Romance auxiliary systems, including so-called cases of triple auxiliation (Loporcaro 2007), according to which, for example, person-based systems may embed modal and temporal restrictions (cf. 33a-c) but not those relating to argument structure, whereas auxiliary systems based on active-stative splits may variously overlay modal, temporal, and personal restrictions. As an example of this latter option, consider the Pugliese dialect of Minervino Murge (Manzini and Savoia 2005,III:27-28): in the present all verb classes show free variation in all persons except the third person where be is only an option (alongside have) with unaccusatives (35a), whereas transitives/unergatives only license have (35b): 35 a sɔ(ndə)~jaɟɟə / si~a / jɛ~ɔ / simmə~aˈvimmə / sɛitə~aˈvɛitə / I.am~have you.sg.are~have (s)he.is~has we.are~have you.are~have sɔndə~jɔnnə məˈnɛutə they.are~have come b sɔ(ndə)~jaɟɟə / si~a / ɔ / simmə~aˈvimmə / sɛitə~aˈvɛitə / I.am~have you.sg.are~have (s)he.has we.are~have you.are~have jɔnnə dərˈmɛutə they.have slept 27.4.1.2.3 Diachronic considerations To conclude this discussion of Romance auxiliary selection, some general comments about the implications of (26) for syntactic change are in order. We begin by observing that the most conservative pattern of auxiliary distribution of the five mesoparametric options presented in (26), that determined by argument structure (cf. Bentley 2006) with strong precedents already in Latin (Vincent 1982; Ledgeway 2012a:130-34; Adams 2013; Roberts 2013:17-20), is situated at the bottom of the hierarchy. This implies that all deviations from this mesoparametric pattern in the history of Romance involve a movement up the hierarchy towards one of the other four less marked and conceptually simpler options but,[footnoteRef:9] significantly, no movements downwards from, say, mood-driven auxiliation to person-auxiliation. There is also no a priori reason to assume that movement up the hierarchy must proceed stepwise, as witnessed by the development of Romanian where, following Dragomirescu and Nicolae (2009) and Ledgeway (2014a), the shift from an original active-stative split to a (finiteness-/)mood-driven system (cf. 29a-d) involves a saltational change, with no evidence of auxiliary variation having first passed through intermediate person- and tense-driven splits (for limited relics of the original active-stative split in modern Romanian, see discussion below). That said, movements up the hierarchy might be motivated by earlier downward microparametric shifts within a given mesoparametric network. This is the case with have generalization found in many modern Ibero-Romance and (extreme) southern Italo-Romance varieties where, as we saw above (cf. 34a-b), in (late) medieval texts the first extensions of have to unaccusative syntax are licensed uniquely in irrealis modal contexts, from where Ledgeway (2003) demonstrates that it gets a foothold in the system before progressively spreading to realis contexts yielding the generalized extension of have witnessed in these varieties today (cf. 28a-c). Arguably, in this case the rise of a modally-determined extension of have to unaccusative syntax represents a microparametric change, involving a downward movement within the mesoparametric network dedicated to argument structure which will ultimately provide the necessary impetus to trigger the mesoparametric change targeting the top of the hierarchy. Biberauer & Roberts (in press, b) describe a similar diachronic development in the history of English involving inversion; see §27.4.1.4. [9: This upward movement is reflected indirectly in many varieties which, although having abandoned the original active-stative split in the auxiliary system in favour of person-driven auxiliation or the generalization of a single auxiliary, preserve the split in the distribution of participle agreement witnessed in §27.4.1.1 (cf. Loporcaro 1998:8-12; in press:§49.2; Manzini and Savoia 2005,II:§5.6.2).] One final dimension of variation that we have not yet discussed in relation to (26) is the possibility that the core reflex of unaccusativity, viz be-selection, has become lexically fossilized and is today limited to a synchronically opaque, small number of intransitive predicates. This nanoparametric state of affairs, in which a once productive auxiliary distinction has all but fallen out of the system today precariously surviving in association with particular predicates as a lexical idiosyncrasy (though for potential semantic motivation, see discussion below of the Auxiliary Selection Hierarchy), accurately describes the situation in many langue d’oïl varieties. Even French – reflexives aside which today invariably license be – has witnessed a striking decline in the number of unaccusatives that today may still select be, ranging from somewhere between twenty and thirty predicates according to most counts (Benveniste 1965:181; Giancarli 2011:373f.) and representing a small subset of those which still systematically align with be in varieties such as Italian, Occitan, or Corsican (Maiden and Robustelli 2007:262; Giancarli 2011). In so-called popular varieties of French the retreat is even greater with have replacing be in conjunction with most if not all unaccusatives (Bauche 1946:105; Guiraud 1969:40f.), a situation replicated in many eastern langue d’oïl varieties (Flutre 1955:59; Remacle 1956:39-48; Descusses 1986:126; Hendschel 2012:§166 b p. 177). More striking are those varieties which residually show retention of be with just one or two unaccusatives. For example, in the Picard patois of Nibas (Vasseur 1996:52) and Valenciennes (Dauby 1979:35) be is today limited to mourir ‘die’ and aller ‘go’, respectively, with have having penetrated all other unaccusatives. Similarly, in the Lorrain variety of Ranrupt all unaccusatives today take have with the sole exception of ‘come’ which still licenses be (Aub-Büscher 1962:84§107). Here one must not forget various Canadian French varieties where, putting aside some complex sociolinguistic factors (Sankoff and Thibault 1977; King and Nadasdi 2005; Rea 2014), unaccusatives frequently show free alternation of have~be (including with reflexives), though often with a higher propensity of be in just a subset of core unaccusatives (e.g. aller ‘go’). In all cases we are clearly dealing with nanovariation, namely synchronically unpredictable cases of lexical exceptions which residually reflect formerly more widespread and regular patterns of variation. Arguably also relevant here are those limited cases of resultative be found in varieties which otherwise have generalized have with unaccusatives. For example, modern Romanian still retains a relic of auxiliary be with a subset of unaccusatives when interpreted with a resultative value (Avram 1994:494, 506-08; Motapanyane 2000:16; Avram and Hill 2007:49-52; Nevaci and Todi 2009:142; Dragomirescu 2010:210; Pană Dindelegan 2013:228), as illustrated by the relative acceptability of have (36a) and be (36b) with punctual and resultative temporal adverbials in the following examples taken from Dragomirescu and Nicolae (2009). 36 a Ion a / *e sosit ieri / de ieri în oraş. John has is arrived yesterday since yesterday in city ‘John arrived yesterday/since yesterday in the city.’ b Ion e /*a sosit de ieri în oraş. John is has arrived since yesterday in city ‘John has been here since yesterday in the city.’ Crucially, as Dragomirescu and Nicolae observe, the distribution of be in such cases is not indiscriminate, but is limited to a subclass of unaccusatives, namely verbs of directed motion and change of location and verbs of (dis)appearance situated at the top of Sorace’s (2000) Auxiliary Selection Hierarchy. Significantly, a similar, if not identical, phenomenon is reported by Manente (2008:42f.) to have occurred in the recent history of Québécois French, where auxiliary have has been extended to unaccusatives of change of location. Although in such cases have has now replaced original be to mark punctual events (37a), auxiliary be survives with these same verbs under the resultant state interpretation (37b). 37 a Jean a arrivé / parti / entré / tombé à huit heures / en deux minutes. Jean has arrived left entered fallen at eight hours in two minutes ‘Jean arrived / left / came in / fell at eight / in two minutes.’ b Jean est arrivé / parti / entré / tombé. Jean is arrived left entered fallen ‘Jean is here / away / in(side) / on the floor.’ The Québécois French facts thus replicate patterns found in Romanian, inasmuch as relics of auxiliary be are restricted in both varieties to resultative readings of a similar subclass of unaccusatives (viz. verbs of directed motion), the only difference being that we have documented evidence of the original have-be transitive-unaccusative split, of which resultative be is a residue, in the recent history of Québécois French, but not in Romanian. 27.4.1.3 Subject-clitic systems of North-Western Romance Across a range of varieties of North-Western Romance spoken in Northern Italy, France, Switzerland, and including Standard French, we can observe the development of systems of subject clitics interacting with the null-subject parameter synchronically and diachronically (§22.3.1, §25.3.3). An important aspect of these changes has again featured grammaticalization (again construable as defined by Roberts and Roussou 2003; see the discussion in §27.2.1) of pronouns (D-elements) as functional heads in T and C systems (see Poletto (2000), Roberts (2014a) and in particular Manzini & Savoia (2005) on the synchronic range of systems featuring extreme microparametric variation concerning which clitics have reanalysed from their earlier pronominal status and how). Partly owing to the complexity of the situation in Northern Italian dialects, we illustrate the main points here with French. The basic pattern of development is summarised in (38): 38 Pronouns: Stage I: Strong subject pronouns Stage II: Weak subject pronouns Stage III: syntactic clitics Null subjects: Stage I: consistent null subjects Stage II: restricted null subjects Stage III: no null subjects Stage IV: consistent null subjects Stage I is exemplified by Latin and most of the Romance languages outside the North-Western area under consideration here (with the notable exception of the recent history of Brazilian Portuguese; see Duarte 1995; Roberts 2014b). In all these varieties, we find strong subject pronouns and full-fledged null-subject systems. Old French (see below), Medieval Veneto (Poletto 1995) and probably other Medieval Northern Italian dialects (Vanelli, Renzi and Benincà 1985) show a combination of restrictions on null subjects along with strong subject pronouns. Standard Literary French and 16th-century Veneto are (to a close approximation) non-null-subject systems with weak subject pronouns (in the sense of Cardinaletti and Starke 1999). Finally, many Modern Northern Italian dialects and, arguably, ‘advanced’ vernacular varieties of contemporary French have grammaticalized the former subject pronouns as agreement markers (bundles of uninterpretable φ-features) in T (the functional head typically associated with finite agreement) and have thus ‘returned’ to null-subject status. Evidence for strong subject pronouns in Old French comes from examples like the following, showing that these pronouns, reflexes of Latin ego, etc., could appear in elliptical contexts, be coordinated and be modified by meïsmes (‘self’) (see for discussion and documentation, see Roberts 1993:112-14): 39 a Et je que sai? (OFr.) and I what know.1sg ‘What do I know?’  b e jo e vos i irum (OFr.) and I and you there= go.fut.1pl ‘and you and I will go there’ c se je meïsmes ne li di (OFr.) if I self neg cl.dat.3sg= tell.1sg ‘if I myself don’t tell him’ It is well-known that Old French allowed null subjects, but only in contexts of verb-second, i.e. (to a close approximation) only in main clauses. In (40), we see examples of null subjects in V2 contexts in Old French (taken from Roberts 1993:124ff.): 40 a Tresqu’ en la mer cunquist la tere altaigne. (OFr., Roland, 3) until in the sea conquered.3sg the land high ‘He conquered the high land all the way to the sea’ b Si chaï en grant povreté. (OFr., Perceval, 441) thus fell.1sg into great poverty ‘Thus I fell into great poverty’ c Si en orent moult grant merveille (OFr. Merlin, 1) thus of.it= had.3pl very great marvel ‘So they wondered very greatly at it.’ In non-V2 contexts, null subjects are not found. In (41a) we have an example of an overt pronoun in the subordinate clause coreferential with a null pronoun in the main clause, a situation strongly dispreferred in fully null-subject languages. In (41b) we see an embedded pronominal subject coreferential with an inanimate (la joie ‘the joy’) in the main clause, also strongly dispreferred in fully null-subject languages: 41 a Ainsi s’ acorderent que il prendront par nuit. thus self= they.agreed that they will.take by night ‘This they agreed that they would take by night …’ (OFr., Le Roman du Graal, B. Cerquiglini (ed.), Union Générale d’Editions, Paris, 1981, 26; Adams 1987:1; Roberts 1993:84) b dont la joye fut tant grant par la ville qu’ elle ne se pourroit of.which the joy was so great in the town that it neg self= could compter count.inf ‘the joy concerning which was so great around town that it could not be counted’ (Jehan de Saintré 160, 5; Sprouse and Vance 1999:263) In Modern French, the situation is quite complex owing in large part to the very strong normative tradition which arguably preserves an earlier diachronic stage in the literary language. Based on Roberts (2010a), who follows Zribi-Hertz (1994), we can distinguish at least four varieties. First, there are the ‘high’ registers, which Zribi-Hertz (1994:136) calls français standard moderne ‘modern standard French’. These varieties allow stylistic inversion, complex inversion and subject-clitic inversion (on the various kinds of inversion in French, see Kayne 1972; Rizzi and Roberts 1989); this is the ‘highest’ register of current literary Standard French. In stylistic inversion, expletive null subjects are allowed in a very narrow range of environments, depending on clause type, hence on the features of C (subjunctive, interrogative and relative, to a close approximation). So this variety represents a highly restrictive form of expletive null-subject language in TP. However, as both Pollock (2006) and Roberts (2010a) argue, subject-clitic and complex inversion constitute a form of conjugaison interrogative (‘interrogative inflection’), whereby φ-features in C license null subjects in SpecTP. So here we have a system of C-licensed expletive null subjects and, in inversion contexts, argumental null subjects. Second, there are registers which do not allow any form of stylistic inversion, but allow complex and subject-clitic inversion. This is also a fairly high register of spoken Standard French, certainly very much in use among educated Parisians and in the media. This is a non-null-subject variety in TP. Again, then, this is a restricted null-subject variety with argumental null subjects licensed by the φ-features of argumental C. Third, there are colloquial registers in which all forms of inversion and are lacking; Zribi-Hertz (1994:137) designates these as français parlé courant (‘everyday spoken French’). These are fully non-null-subject systems in which neither C nor T licenses a null argument under any conditions. This is probably the most widely-spoken form of Modern Standard French. Finally, there are vernacular varieties in which subject proclitics are to be analysed as realising φ-features of T, just as in many Northern Italian dialects (according to the analyses in Manzini and Savoia (2005), Roberts (2014a) and the references cited there). Zribi-Hertz (1994:137) refers to these varieties as français très évolué (‘very advanced French’). These varieties are fully null-subject in TP, thanks to the presence of the subject clitic in T realizing uninterpretable φ-features in a fashion analogous to the role played by ‘rich’ verbal agreement inflection in a canonical null-subject language such as Italian. This is shown by the fact that they co-occur with non-referentially quantified subjects (which as such cannot be in dislocated positions) as in the attested examples in (42): 42 a Tout le monde il est beau, tout le monde il est gentil. (film title) everyone he is handsome everyone he is nice ‘Everyone is handsome, everyone is nice.’ b Personne il fiche rien, à Toulon. No-one he does anything at Toulon ‘No-one does anything in Toulon.’ (Zribi-Hertz’ (19a, e), p. 137); (42b) from P. Mille Barnavuax et quelques femmes, 1908) It is hard to date the changes in the non-standard varieties of French due to the normative influence on written French, but the development of Northern Italian dialects, judging from Poletto’s (1995) treatment of Veneto, suggests that the various stages were fairly short-lived. Both Veneto and French seem to have gone through the stages listed in (38) in their recorded histories since about 1400. Again, then, we observe relatively rapid microparametric change. In terms of parameter hierarchies, we can posit a subpart of a larger hierarchy related to agreement and argument-marking dealing with microparametric variation in licensing null arguments along the lines of (43) adapted from Roberts (2012): [Insert Example (43) HERE] Further down this hierarchy there are options concerning partial null-subject languages (see Holmberg 2010) and expletive null-subject languages, along the following lines (where C[+F] designates a marked clause-typing feature): [Insert example (44) HERE] (Here we have glossed over the distinction between argumental and expletive null subjects, and hence the distinction between the first two varieties of Modern French described above). French and Veneto (and presumably other Northern Italian dialects) have moved progressively down this hierarchy, until they reach the stage of full grammaticalization of subject clitics as agreement markers in T, at which point they ‘jump’ back up to the position of Italian in (43), somewhat higher up the hierarchy. So again (cf. the discussion above of have-generalization from examples like (34a-b)) we see a pattern of incremental changes moving a system down a microparametric hierarchy followed by a relatively dramatic reanalysis leading to a ‘jump’ upwards in the hierarchy. 27.4.1.4 Inversion in the history of English As we have seen, there is evidence for the parametric taxonomy in (45): 45 For a given value vi of a parametrically variant feature [F]: a. Macroparameters: all functional heads of the relevant type share vi; b. Mesoparameters: all functional heads of a given naturally definable class, e.g. [+V], share vi; c. Microparameters: a small sub-class of functional heads (e.g. modals) shows vi; d. Nanoparameters: one/more individual lexical items has vi. If parametric change involves acquisition-mediated reanalysis of PLD, macroparameters will be set ‘easily’, hence resisting reanalysis and being strongly conserved; meso- and microparameters are correspondingly less salient in the PLD, hence less reanalysis-resistant and less strongly conserved. Nanoparameters are, in principle, still less reanalysis-resistant and thus more unstable, aside from the intervention of frequency effects. We have seen evidence for the stability of macroparameters and the relative instability of microparameters in the preceding sections, as well as cases of changes moving a system down a hierarchy followed in some cases by an abrupt ‘jump’ to a higher position in a hierarchy. In this section we summarize a case of change from meso to micro to nano involving Conditional Inversion (CI) in the history of English (a full exposition is given in Biberauer and Roberts in press b). We show that the central component of CI has remained unchanged since Old English, in that it involves T-to-C movement where C has a feature marking the clause Irr(ealis) (e.g. swelte ic, libbe ic ‘die I, live I’ = ‘if I live or die’). In Old English, CI was part of a family of operations raising inflected verbs into the C-system (V2). This feature is general to all root and some embedded Cs and holds across Germanic, making it a good candidate for a mesoparameter. What has changed since Old English is the range of elements undergoing CI, and how CI relates to other forms of inversion. The loss of V2 is usually dated to the 15th century (Fischer, van Kemenade, Loopman and van der Wurff 2000), but various forms of ‘residual V2’ in marked clause-types survived, e.g. interrogative inversion and CI. The shift from full to residual V2 is a shift from meso to micro: the class of T-attracting Cs contracts. In Early Modern English, lexical V-to-T movement was lost (this was probably connected to the grammaticalization of the modals discussed in §27.2.1; see Roberts (1985). Thereafter, only auxiliaries undergo CI, as in interrogative and other kinds of inversion. The shift from residual V2 to subject-aux inversion further restricts the items undergoing inversion, although the T-to-C trigger is unchanged. What changed here is a T-feature, from a meso – all verbs – to a micro – auxiliaries only – value. The most interesting change affecting CI is recent, though: between the 17th and the 19th century, CI was no different to other inversions, featuring with all auxiliaries, including ‘dummy’ do. At this period, then, we find examples of CI with a wider range of auxiliaries than is possible in contemporary English. Here are some early-19th-century examples: 46 a Would you be really … a man of honour .. you would … restore that parchment to Lord Evandale. (1816, Scott, Old Mortality (Tauchn.) 435; Visser 1963-73, III: 1730) b This was a very prudential resolution, could he have kept it. (1751, Smollett, Peregr. Pickle II, xix; Visser 1963-73, III:1748) c Might I kiss those eyes of fire, A million scarce would quench desire. (1807, Byron, Hours of Idleness: To Ellen; Visser 1963-73, III, 1778) From roughly 1850, CI became restricted to had, should and, more marginally, were. This looks like a nanoparameter, as it affects one modal, and specific forms of have and be. Meanwhile, interrogative inversion has remained fully productive for all auxiliaries. Optative inversion, however, was first limited to may, before becoming formulaic (May you rot! but *May you eat!). On the view of parameters advocated here, then, the parametric taxonomy in (45) allows us to understand how systems may become gradually more marked, requiring ever more specific triggers for operations, until a feature(class) ceases to act as a trigger, and the system radically simplifies. In the case of CI, the last stage is reached when in the varieties of vernacular English in which this construction has been entirely lost. 27.5 Conclusion: Speculations on a further kind of parameter In the foregoing, we have essentially tried to support the general case made by Baker (2008b) that there are different kinds of parameters, using diachronic evidence and at the same time showing how this kind of approach can give insights into syntactic change. We have tried to support our case in terms of a specific, emergentist approach to parameters (see also §7.3; §16.4.2) which is associated with parameter hierarchies and the related taxonomy in (45). Further support, and much detailed evidence and analysis, is given in the references provided, and more examples of parameter hierarchies can be found in the references given in note 2. There is a further issue that we would like to raise, in a somewhat more speculative vein. The principles-and-parameters approach is based on an analogy with genetics. The parametric genotype (e.g. the setting of a putative ‘directionality parameter’ as head-final vs head-initial) gives rise to the surface phenotype (OV/VO, Pre-/Post-positions, etc.). But an important concept in genetics has gone largely undiscussed and unexploited in work in this tradition: that of pleiotropy. In genetics, pleiotropy occurs when one gene influences multiple, seemingly unrelated phenotypic traits. In terms of the standard principles-and-parameters model, we could think of pleiotropic parameters as ‘deep’ parameters which profoundly influence the overall shape of a grammatical system. The notion of pleiotropic parameter does not correspond to the concept of ‘macroparameter’ as discussed above. From that point of view, as we have seen, the notion of macroparameter is epiphenomenal, arising when a class of microparameters, characterized as formal features of functional heads act ‘in concert’ to produce an emergent macroparametric effect. The notion of pleiotropic parameter, on the other hand, is a novel one: distinct from both the primitive GB concept and the derivative emergentist one. Taking seriously the idea that parameters are nothing more than optional, emergent formal features (i.e. not innately prespecified by Universal Grammar, and possibly not domain-specific; see Biberauer 2011; Biberauer, Roberts and Sheehan 2014), it is possible that a small subset of these features act as ‘master features’, determining the ways in which many other features manifest themselves (to use a further useful concept from genetics, the latter features are epistatic; epistasis is a phenomenon that consists of the effect of one gene being dependent on the presence of one or more ‘modifier genes’ in the genetic background). Pleiotropic parameters have a disproportionate and profound effect on the parametric phenotype as it can be observed in surface variation. We may be witnessing the effect of pleiotropic features/parameters in cases of convergent patterns of diachronic change of the kind documented above from the history of Latin/Romance. We leave this potentially important new research area open for further elaboration here, merely noting that it may prove as fruitful as the approach described here and, indeed, as the original notion of parameter of Universal Grammar itself. References Adams, J. 1976. ‘A typological approach to Latin word order’, Indogermanische Forschungen 81: 70-99. Adams J. 2013. ‘Past participle + habeo’, in Social variation and the Latin language. Cambridge: Cambridge University Press, pp. 615-51. Adams, M. 1987. ‘From Old French to the theory of pro-drop’, Natural Language and Linguistic Theory 5:1-32. Aub-Büscher, G. 1962. Le Parler rural de Ranrupt (Bas Rhin). Paris: Klincksieck. Avelar, J. 2009. ‘Inversão locativa e sintaxe de concordância no português brasileiro’, Matraga 16: 232-52. Avelar, J. and Galves, C. 2011. ‘Tópico e concordância em português brasileiro e português europeu’, in A. Costa, I. Falé, and P. Barbosa (eds), Textos seleccionados, XXVI Encontro nacional da associação portuguesa de linguística. Lisbon: APL, pp. 49-65. Avram, L. 1994. ‘Auxiliary configurations in English and Romanian’, Revue roumaine de linguistique 5-6: 493-510. Avram, L. and Hill, V. 2007. ‘An irrealis BE auxiliary in Romanian’, in R. Aranovich (ed.), Split auxiliary systems. A cross-linguistic perspective. Amsterdam/Philadelphia: Benjamins, pp. 47-64. Baker, M. 1996. The Polysynthesis parameter. Oxford: Oxford University Press. Baker, M. 2001. The Atoms of language: The mind’s hidden rules of grammar. Oxford: Oxford University Press. Baker, M. 2008a. The syntax of agreement and concord. Cambridge: Cambridge University Press. Baker, M. 2008b. ‘The macroparameter in a microparametric world’, in T. Biberauer (ed.), The limits of syntactic variation. Amsterdam: Benjamins, pp. 351-74. Battye, A. and Roberts, I. (eds) 1995. Clause structure and language change. Oxford: Oxford University Press. Bauche, H. 1946. Le langage populaire. Grammaire, syntaxe et dictionnaire du français tel qu’on le parle dans le peuple avec tous les termes d’argot usuel. Paris: Payot. Bauer, B. 1995. The emergence and development of SVO patterning in Latin and French. Oxford: Oxford University Press. Bentley, D. 2006. Split intransitivity in Italian. Berlin/New York: Mouton de Gruyter. Bentley, D. in press. ‘Split intransitivity’, in Ledgeway and Martin (eds). Benveniste, E. 1950. ‘Actif et moyen dans le verbe’, Journal de Psychologie 43: 119-27. Reprinted in Benveniste, E. 1966. Problèmes de linguistique générale I. Paris: Gallimard, pp. 168-75). Benveniste, E. 1960. ‘“Être” et “avoir” dans leurs fonctions linguistiques’, Bulletin de la Société linguistique de Paris, 55: 113-34 (Reprinted in Benveniste, E. 1966. Problèmes de linguistique générale I. Paris: Gallimard, pp. 187-207). Benveniste, E. 1965. ‘Structure des relations d’auxiliarité’, Acta Linguistica Hafniensia 9: 1-15. den Besten, H. 1983. ‘On the interaction of root transformations and lexical deletive rules’, in W. Abraham (ed.), On the formal syntax of the Westgermania. Amsterdam: John Benjamins, pp. 47-131. Biberauer, T. 2011. ‘In defence of lexico-centric parametric variation: Two 3rd factor-constrained case studies’, paper presented at the Workshop on Formal Grammar and Syntactic Variation: Rethinking Parameters (Madrid). Biberauer, T., Holmberg, A. and Roberts, I. 2014. ‘A syntactic universal and its consequences’, Linguistic Inquiry 45: 169-225. Biberauer, T., Holmberg, A., Roberts, I. and Sheehan, M. (eds) 2010. Parametric variation: Null subjects in minimalist theory. Cambridge: Cambridge University Press Biberauer, T., Holmberg, A., Roberts, I. and Sheehan, M. 2014. ‘Complexity in comparative syntax: The view form modern parametric theory’, in F. Newmeyer and L. Preston (eds), Measuring Linguistic Complexity. Oxford: Oxford University Press, pp. 103-27. Biberauer, T. and Roberts, I. 2008. ‘Cascading parameter changes: Internally-driven change in middle and early modern English’, in T. Eythórsson (ed), Grammatical change and linguistic theory: The rosendal papers. Amsterdam: Benjamins, pp. 79-114. Biberauer, T. and Roberts, I. 2012a. ‘Towards a parameter hierarchy for auxiliaries: Diachronic considerations’, in J. Chancharu, X. Hu and M. Mitrović (eds), Cambridge Occasional Papers in Linguistics 6: 209-36. Biberauer, T. and Roberts, I. 2012b. ‘The significance of what hasn’t happened’, paper presented at Diachronic Generative Syntax 14, Lisbon, 4 July. Biberauer, T. and Roberts, I. 2015a. ‘Rethinking formal hierarchies: A proposed unification’, in J. Chancharu, X. Hu and M. Mitrović (eds), Cambridge Occasional Papers in Linguistics 7: 1-31. Biberauer, T. and Roberts, I. 2015b. ‘Clausal hierarchies’, in U. Shlonsky (ed.), Beyond functional sequence. Oxford: Oxford University Press, pp. 295-313. Biberauer, T. and Roberts, I. (in press a). ‘Towards a parameter hierarchy for auxiliaries: Diachronic considerations’, in A. Kroch (ed.), Proceedings of the XII diachronic generative syntax meeting. Oxford: Oxford University Press. Biberauer, T. and Roberts, I. in press b. ‘Conditional inversion and types of parametric change’ in B. Los and P. de Haan (eds). Verb-Second languages: Essays in honour of Ans van Kemenade. Amsterdam: John Benjamins. Biberauer, T., Roberts, I., and Sheehan, M. 2014. ‘No-choice parameters and the limits of syntactic variation’, in R. Santana-LaBarge (ed.), Proceedings of WCCFL 31. Somerville, MA: Cascadilla Press, pp. 46-55. Borer, H. 1984. Parametric syntax. Dordrecht: Foris. Bošković, Ž. 2008. ‘What will you have, DP or NP?’, in Emily Elfner and Martin Walkow (eds), Proceedings of the thirty-ninth annual north east linguistic society. University of Massachusetts, Amherst: GLSA, pp. 101-14. Brandi, L. and Cordin, P. 1989. ‘Two Italian dialects and the null subject parameter’, in O. Jaeggli and K. Safir (eds), The null subject parameter. Dordrecht: Kluwer, pp. 111-42. Branigan, P. 2014. ‘Macroparameter learnability: An Algonquian case study’. Memorial University Newfoundland: unpublished ms. Calboli, G. (ed.) 1989. Subordination and other topics in Latin. Proceedings of the Third colloquium on Latin linguistics, Bologna, 1-5 April 1985. Amsterdam/Philadelphia: Benjamins. Cardinaletti, A. 2004. ‘Toward a cartography of subject positions’, in L. Rizzi (ed.), The structure of CP and IP. Oxford: Oxford University of Press, pp. 115-65. Cardinaletti, A. and Starke, M. 1999. ‘The typology of structural deficiency: A case study of the three classes of pronouns’, in H van Riemsdijk (ed.), Clitics in the languages of Europe. Berlin: de Gruyter, pp. 145-235. Cennamo, M. in press. ‘Voice’, in Ledgeway and Maiden (eds). Charpin, F. 1989. ‘Etude de syntaxe énonciative: L’ordre des mots et la phrase’, in Calboli (ed.), pp. 503-20. Chomsky, N. 1964. Current issues in linguistic theory. The Hague: Mouton. Chomsky, N. 1973. ‘Conditions on transformations’, in S. Anderson and P. Kiparsky (eds.), A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston, pp. 232-86. Chomsky, N. 1975. Reflections on language. New York: Pantheon. Chomsky, N. 1977. Essays on form and interpretation. New York, Amsterdam, London, North Holland. Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, N. 1995. The minimalist program. Cambridge MA: MIT Press. Chomsky, N. 2005. ‘Three factors in language design’, Linguistic Inquiry 36(1): 1-22. Chomsky, N. 2007. ‘Approaching UG from below’, in H.-M. Gärtner and U. Sauerland (eds), Interfaces + recursion = language? Berlin: Mouton de Gruyter, pp. 1-29. Cinque, G. 1999. Adverbs and functional heads. A cross-linguistic perspective. Oxford: Oxford University Press. Cinque, G. 2006. Restructuring and functional heads. The cartography of syntactic structures. Volume 4. Oxford: Oxford University Press. Clark, R. 1992. ‘The selection of syntactic knowledge’, Language Acquisition 2(2): 83-149. Clark, R. and Roberts, I. 1993. ‘A computational approach to language learnability and language change’, Linguistic Inquiry 24: 299-345. Comrie, B. 1978. ‘Ergativity’ in W.P. Lehmann (ed.), Syntactic typology. Studies in the phenomenology of language. Austin: University of Texas, pp. 329-94. D’Alessandro, R. and Ledgeway, A. 2010. ‘The Abruzzese T-v system: feature spreading and the double auxiliary construction’, in D’Alessandro, Ledgeway and Roberts (eds), pp. 201-209. D’Alessandro, R., Ledgeway, A. and Roberts, I. (eds) 2010. Syntactic variation: The dialects of Italy. Cambridge: Cambridge University Press. D’Alessandro, R. and Roberts, I. 2010. ‘Past participle agreement in Abruzzese: Split auxiliary selection and the null‐subject parameter’, Natural Language and Linguistic Theory 28: 41-72. Dankaert, L. 2012. Latin embedded clauses: The left periphery. Amsterdam: Benjamins Dauby, J. 1979. Le livre du ‘rouchi’. Parler picard de Valenciennes. Amiens: Musée de Picardie. de Andrade, A. and Galves, C. 2014. ‘A unified analysis of subject topics in Brazilian Portuguese’, Journal of Portuguese Linguistics 13: 117-47. Descusses, M. 1986. Le Patois ardennais de Gespunsart. Paris: Société d'études linguistiques et anthropologiques de France. Devine, A. M. and Stephens, L.D. 2006. Latin word order: Structured meaning and information. Oxford: Oxford University Press. Dixon, R. 1994. Ergativity. Cambridge: Cambridge Univeriy Press. Dragomirescu, A. 2010. Ergativitatea. Tipologie, sintaxă, semantică. Bucharest: Editura Universității din București. Dragomirescu, A. and Nicolae, A. 2009. ‘Relics of auxiliary selection in Romanian’. Iorgu Iordan − Al. Rosetti Institute of Linguistics, Buc