Theory and data in diachronic Construction Grammar

SAVE OFFLINE

Theory and data in diachronic Construction Grammar The case ofthe what with construction Graeme Trousdale University of Edinburgh

Constructionalization (the diachronic creation of conventional symbolic units at different levels of schematicity and complexity) is a process which involves a series of micro-changes at different linguistic levels. The development oí what with constructions in English is argued to be a case of grammatical constructionalization, whereby aspects of a construction become more general, productive, and less compositional. Equally, parts ofthe construction become more fixed (involving a reduction in variability), while other parts ofthe construction expand. The application of principles of construction grammar to aspects of diachronic change helps to clarify the relationship between theoretical principles of language change and the analysis of naturally occurring data; equally, the study ofthe what with construction reveals areas of potential convergence between formal and functional approaches to syntactic change, as well as areas of difference.

1.

Introduction

Studies which bring together the findings of research into grammaticalization with the principles of construction grammars of various kinds (see, for example, the various papers in Bergs and Diewald 2008, Hilpert 2008, Fried 2009, Traugott 2008b, and Patten 2010, among others) are typically concerned with understanding the nature of micro-changes, at various constructional levels.' Macro-changes may be represented abstractly as 'A > B', wbile micro-changes are "the tiny local steps between A and B tbat the arrow '>' encompasses" (Brinton & Traugott 2005:150). Such research is also connected with broader questions regarding the nature and locus of language change, for instance whether change is best understood as an abrupt reanalysis of Unguistic structures at acquisition, or as a gradual process over a lifetime of usage, tbis distinction sometimes said to cbaracterize tbe different approacbes taken by formal approaches to change on the one hand, and

Studies in Language 36:3 (2012), 576-602. DOI io.io75/sl.36.3.o5tro ISSN 0378-4177 / E-ISSN 1569-9978 © John Benjamins Publishing Company

Theory and data in diacbronic Construction Grammar 577

functional approaches to cbange on the other (on which see further Fischer 2007). Formal generativist approaches to grammaticalization have also recently focused on the nature of micro-changes, sometimes understood as parameter resetting (Roberts 2010) or feature economy (van Gelderen 2011). From the perspective of construction grammar, micro-changes associated with grammaticalization have been shown to take place in both the form and meaning poles of a construction (for instance, syntactic reanalysis and phonetic attrition in the former, semantic bleaching and pragmatic enrichment in the latter). Evidence for these micro-constructional changes has typicaUy been adduced from diachronic corpora; since most variants of construction grammar are usagebased frameworks, the use of quantified data from corpora of existing texts as the core material for investigating the development of linguistic change is very common (for instance, Hilpert 2008, Patten 2010). Formal approaches to language variation and change have also increasingly made use of quantified data from corpora of various kinds (for instance, Pintzuk & Haeberli 2008 on word order in Old Englisb). Some of this research has been particularly concerned with the intersection of formal and sociolinguistic accounts of synchronie variation in the speech community (for instance, Adger and Smith 2005, Rupp 2005, Adger 2006). Other examples have been concerned with register differences and formal accounts of variation in a corpus, such as the article which prompted the research presented here (Felser & Britain 2007). By comparing a formal (minimalist) and functional (constructional) approach to variation and change in a particular area of English grammar,^ the research presented in this article highlights some of the different predictions of the two approaches and tests these predictions using evidence from diachronic and synchronie corpora. In keeping with the aims and objectives of tbis special issue, this article attempts to investigate and compare insights from formal linguistics and cognitive linguistics, in order to illustrate similarities and differences between the two approaches, and — more crucially — to demonstrate that there are perhaps more areas of convergence than might otherwise appear to be the case, accepting that some sharp, fundamental differences between the two approaches to language structure wiU inevitably persist. In this article, I address the issues described above through an investigation of a particular construction in English, the what with construction,^ iUustrated in (1):^ (1) What with the gown, the limos, and all the rest, you're probably looking at about a hundred grand. (2009 Diane Mott Davidson, Fatally Flaky; COCA) The issues are explored by answers to the foUowing primary questions: a.

What synchronie analyses of the construction might be proposed?

578 Graeme Trousdale

b. Wbat do corpus data suggest about the recent development of the construction? c. How can the micro-changes uncovered in tbe corpora be modeUed? d. Wbat does this study suggest about the relationship between data and tbeory in cognitive linguistics? Tbe article is structured as foUows. In tbe next section, I provide a grammatical description of tbe synchronie pattern, foUowed by a summary of its analysis in a formal framework (Felser & Britain 2007), and an analysis in constructional terms. In Section 3, the recent historical evolution of the pattern is described; in tbis section, Felser & Britain's account of tbe bistorical material tbey consider is also discussed. Section 4 provides an analysis of tbe development of tbe what with construction from tbe perspective of grammatical constructionalization, focussing on tbe nature of the micro-steps involved at various stages, and some of the ways in which this gives rise to synchronie variability. Section 5 is tbe conclusion.

2

Free adjuncts, absolutes and tbe what with pattern in contemporary English

The grammatical properties of the what with pattern may be seen in tbe context of otber, more general types, of wbicb it is a particular instance. Tbe two relevant general types are free adjuncts and absolutes (on wbicb see furtber Stump 1985, Kortmann 1991), prototypical instances of wbicb are illustrated by (2) and (3) respectively: (2) Having got out of the carriage. Holmes shouted for Watson. (3) Moriarty having got out of the carriage. Holmes shouted for Watson. Tbe free adjunct (2) differs from tbe absolute (3) primarily because tbe former bas no overt subject NP in tbe participial clause, wbile tbe latter does: In absolutes, tbe overt subject of tbe non-finite verb is different from tbe subject of tbe finite verb in tbe main clause; in free adjuncts, tbe covert subject of tbe non-finite verb is co-referential with, and controlled by, tbe subject of tbe finite verb in tbe main clause. Tbis distinction also suggests tbat the bond between tbe non-finite and finite/matrix clause in (2) is stronger tban tbat in (3) — Ouirk, Greenbaum, Leecb & Svartvik (1985:1120) argue tbat the absence of a sbared subject in tbe two clauses of (3) are tbe grounds for describing sucb structures as absolutes. Tbey suggest tbat tbe non-finite clauses are "not explicitly bound to tbe matrix clause syntacticaUy"(Quirk, Greenbaum, Leecb & Svartvik 1985:1120), and as a result, free adjuncts may be considered more grammaticalized tban absolutes (cf Hopper & Traugott 2003:175ff.). Tbese differences between free adjuncts and absolutes bold for prototypical cases only; tbere are instances of free adjuncts wbose covert

Theory and data in diachronic Construction Grammar 579

subject is different from the subject of the main clause verb, and absolutes whose overt subject is coreferential with the subject of the main clause verb (Kortmann 1991:43, 91; Quirk etal. 1985: §15.58; Rio-Rey 2002:311). Kortmann (1991:7-8, 11-12) (see also Stump 1985) observes that the nonfinite clauses in (2) and (3) above may be augmented by a subordinator which further specifies the relationship between the finite and non-finite clauses. For instance, simultaneity of the two processes denoted by the verbs in each clause may be indicated by the subordinator while in free adjuncts: (4)

While coming out of the carriage. Holmes shouted for Watson.

Kortmann (1991:11) particularly observes that the items which augment absolutes and those which augment free adjuncts are in complementary distribution, as iUustrated by the following: (5)

With/*While Moriarty pounding on the carriage door. Holmes discovered he was in danger.

(6)

While/*With pounding on the carriage door. Holmes discovered he was in danger.

The items which typicaUy augment absolutes are with, Ulustrated in (3) above, and without, as in (7): (7)

Without Moriarity noticing. Holmes signalled to Inspector Lestrade.

In addition to with{out) augmentation, there is an additional pattern avaUable to speakers of English, namely the what with pattern, the focus of the present article.^ Examples of this pattern are iUustrated in (8): (8) a.

What with the gown, the limos, and all the rest, you're probably looking at about a hundred grand. (2009 Diane Mott Davidson, Fatally Flaky; COCA) b. In retrospect I realize I should have known that was a bad sign, what with the Raven Mockers being set loose and all. (2009 Kristin Cast, Hunted; COCA) c. But of course, to be fair to the girl, she wasn't herself at the Deanery, what with thinking of how Lord Hawtry's good eye had darkened when she refused his hand in marriage. (2009 Dorothy Cannell, She Shoots to Conquer; COCA) d. The bed was big and lonesome what with Dimmert gone. (2009 Jan Watson, Sweetwater Run; COCA) e.

The Deloche woman was going to have one heck of a time getting rid of the place, what with the economy the way it was in Florida.

58o Graeme Trousdale

(2009 Emifle Riehards, Happiness Key; COCA) Felser & Britain (2007) eonsider what with absolutes to be rare in eontemporary English (see also footnote 3). As illustrated by the examples in (8), what with patterns in Present Day English (PDE) ean eontain within them: a. b. c. d. e.

(eoordinated) NPs non-finite ¿n^-elauses with an overt subject non-finite ¿«^-clauses with no overt subject'' non-finite en-elauses (with an overt subjeet) verbless/smafl elauses

Kortmann (1991:202-4) observes the foflowing properties oí what with patterns: -

-

-

the 'eausality' funetion of what with is not restrieted to absolutes; what with also oeeurs in prepositional phrases (e.g. (8a) above) and gerundive elauses (e.g. (8e) above). what with patterns oeeur in a partieular pragmatie eontext, namely "if the matrix proposition denotes some non-event or negative state, or, more generally, some proposition whieh has eertain negative implieations (at least from the view of the speaker)". what with patterns typieally appear with eoordinated lists of 'reasons', or with general extenders sueh as and all in (8b) above.

Huddleston & Pullum (2002:626) eonsider the what with pattern to be idiosyncratic and fossilized. In their analysis, what with introduces reason adjuncts, and its idiomaticity "developed out of an otherwise almost obsolete use of what to introduce lists or coordinations, espeeially of PPs", noting that prepositions other than with are very rare in eontemporary English, and giving an example with what between to iflustrate the marginal variability. 2.1 A minimalist analysis In this sub-seetion, I summarise the aeeount of the what with pattern within the framework of Minimalism, proposed by Felser & Britain (2007). They argue that what with absolutes are not eonstruetions, whieh they take to be idiosyneraetie form-meaning pairings, along the lines of the definition provided by Goldberg (1995:4); their strueture is eompositional. An interesting feature of the aeeoimt provided by Felser & Britain (2007) is that they attempt to make sense of corpus data from a minimalist perspective, rather than rely on native speaker intuitions. Their search of the British National Corpus (BNC) aflows them to consider formal and functional variation: their specific interest concerns the distribution of what

Theory and data in diachronic Construction Grammar 581

with patterns in the spoken and written subcorpora of tbe BNC, supplemented vñtb some evidence taken from tbe internet. Since the focus here is preponderantly on the nature of the formal variation in the what with construction, properties associated with, for example, genre (spoken vs. written English) are not addressed in detail here. In their account, Eelser & Britain (2007) treat what with patterns^ as factive, 'strong-presuppositional' (Katz 1993) adjuncts. The function of what with is to introduce a reason adjunct linked to a superordinate or matrix clause (foUowing the definition of what with provided by the Oxford Fnglish Dictionary {Oxford Fnglish Dictionary, s.v. what II adv. or conj. 2. b)), a function that what with strings share witb because clauses. In discussing previous researcb on free adjunct and absolute structures in Englisb, Eelser & Britain propose a modification to tbe claim put forward by Kortmann (1991:202), wbicb is tbat what with patterns are more restricted tban with absolutes, because vñtb what with, tbe matrix proposition "denotes some non-event or a negative state, or ... bas certain negative implications (at least from the point of view ofthe speaker)". Rather, Eelser & Britain suggest that what with patterns function to justify a claim made in the matrix clause. The high proportion of what with patterns with 'negative state' matrix propositions follows from the fact that such claims typicaUy require (more) justification; but 'negative evaluation' is not inherent in the semantics of what with patterns. The analysis proposed by Eelser & Britain (2007) takes what with patterns as compositional strings, wbere botb what and with are located in pbrase structure as tbe heads of functional phrases. The specifics of this claim are elaborated immediately below, but here it should be pointed out tbat this compositionality is the central argument for Eelser & Britain to reject a constructional analysis for what with: The structure can be adequately analysed witbin the architecture of minimalist phrase structure. Felser & Britain acknowledge that what with patterns are uncommon, that they have a particular pragmatic function, and that they often appear before a coordinated string or list, but suggest tbat there are strong arguments not in favour of a constructional account. A crucial argument in this regard is that Eelser & Britain (2007) consider the 'core' meaning of what with patterns to be distributable over tbe parts of the sequence. That is, the crucial factor of non-compositionalityis missing, and noncompositionality was taken as a defining feature in earlier accounts of Construction Grammar, as exemplified in the definition of a construction provided by Goldberg (1995:4), namely: "C is a CONSTRUCTION iff^^fC is a form-meaning pair sucb tbat some aspect of E^. or some aspect of S. is not strictly predictable from C's component parts or from otber previously established constructions."* The grounds for the compositional analysis are as follows.

58a Graeme Trousdale

Felser & Britain (2007) take what with patterns which have ¿n^H-subject and en complements (for instance, in examples (8b) and (8d) above) as the 'basic' structure. In this analysis, with is a prepositional Complementizer, which takes a TP complement. Evidence in favour of this analysis comes from a range of syntactic tests, such as quantifier float, passivization, the possibility of expletive or pleonastic subjects, and negative scope effects. The other complements of what and with are also clausal in this analysis. In instances where the participial complement of {what) with has no subject (for instance, in example (8c) above), its status as a clause is evidenced by patterns associated with passivization and clausal negation. Even apparently verbless clauses, that is, those with what appear to be simply nominal complements such as (8a), should StiU be analysed as having a clausal structure, this time with a nuU T head and pro in spec-TP position. The argument proposed in this case relies on associations between what with patterns and with absolutes. Felser & Britain (2007) analyse instances of the latter, such as (9): (9) with a new baby, we have very little spare time. as having "an understood possessional have or existential there be interpretation", noting that what with constructions can involve coordination of NP and ¿«^-i-subj complements. An example of this from the COCA corpus is (10): (10) What with the boyfriend coming back and all the confusion of the paramedics and neighbours, they couldn'tfindanything. (2003, Geoffrey Becker, Great American; COCA) As a result, the formal structures of with the kids running around and with a new baby are (11) and (12) respectively: (11) [ç, with [.j.p the kids [.j., 0 [yp [y, running around] ] ] ] ] (12) [^with [.j.ppro [.J..0 [yp[y. 0[ppanewbaby]]]]]] The structural analysis of what in what with constructions in Felser & Britain's account makes use of Cinque's fine-grained system of functional heads (Cinque 1999), particularly the existence of Moodg^^^^^^^^ in an elaborated C system. Drawing comparisons with expressions like what a beautiful day, and how come he's always late, Felser & Britain (2007:122) analyse what as a factive operator which functions as specifier to a (null) evaluative head, an operator which is "neutral with respect to how exactly the proposition in its scope is evaluated". In sum the formal structure of what with patterns in this model is as (13) below; what with structures are therefore treated as Evaluative Phrases: what [g^^, 0 [(,p [ç. with [^.p...]]]]]

Theory and data in diachronic Construction Grammar 583

The neutrality of the operator with respect to the evaluation of the proposition in its scope is significant for an understanding of how the meaning of what with is interpreted by tbe bearer, and consequently, for bow tbe semantics and pragmatics of sucb patterns sbould be associated witb tbe formal structure above. In tbis minimalist account, tbe pragmatic meaning of particular instances of use (for example, wbetber a positive or a negative state cbaracterises tbe proposition in tbe matrix clause) is determined contextuaUy precisely because tbe abstract Evaluative bead is neutral in tbis regard. In otber words, tbe formal analysis provided by Felser & Britain (2007) gives a uniform account of tbe different complementation patterns of with and of tbe semantics of what, and leaves 'positive/negative' evaluation to tbe speaker/bearer's pragmatic knowledge. In tbe rest of this article, I explore some further issues regarding the nature of the synchronie variation, and the relationship between tbat syncbronic variation and tbe diacbronic evolution of what with patterns wbicb warrant investigation, and wbich suggest a constructional account may capture both the idiosyncracies and the generalizations which Felser & Britain accurately reflect in their paper. 2.2 A constructional analysis Tbe syncbronic analysis of what with constructions considers tbese patterns to be conventional symbolic units, a general deflnition of constructions (Langacker 1987, Croft 2005). Constructions are ordered in a taxonomic network; tbis network may be minimaUy described using tbe foUowing constructional levels, tbe macro-, meso-, and micro-construction, as weU as tbe notion of construct (Traugott 2007,2008a, b; see also Fried 2008). Tbe construct is an attested instance of language use; a set of constructs forms a type, tbe micro-construction; groups of similarly-bebaving micro-constructions form a meso-construction; and tbe macro-construction is the largest construction relevant to the process undergoing change. Elsewhere, constructions have been described in terms of dimensions of atomicity and scbematicity (Croft and Cruse 2004:255), fully atomic and substantive constructions being examples such as red 'red' (i.e. the form red arbitrarily linked to tbe meaning 'red'), traditionally associated with items in tbe lexicon, and fully complex and schematic constructions being examples such as tbe Double Object Construction (Goldberg 1995), traditionally associated witb syntax. Tbe what with construction sits somewbere between tbese two extremes: it is partiaUy specified by virtue of tbe initial pbonetic sequence [wüt wi9], but tbe remainder of the construction is schematic (taking a range of different complement types) and phonetically underspecified. It is a complex, not atomic, construction; crucially, it is a construction in which the range of potential complement-types has increased over time.

584 Graeme Trousdale

The particular argument proposed here is that the various what with constructions form a network of linguistic knowledge (Hudson 2007); specificaUy, that speakers abstract away from instances of use (constructs) to form constructions of varying degrees of generality. For instance, speakers abstract away from what with constructs such as (14a)-(14c), all taken from GOGA: (14) a.

Sometimes she wondered if Anastasia were gay, what with the piercing and diet. (2005 Jill Rosenberg, The Land of Sunshine and Flowers; COGA) b. Everyone was busy with the Rep and Dem candidates, what with the various scandals, the shooting and the attempted kidnapping. (2008 Wayne Wightman, A Foreign Country; COCA) c. Computers are discarded fairly easily these days, what with constant innovation and greater affordability. (2002 Jean Thilmany, Keeping in touch [Mechanical Engineering]; COCA)

to form a micro-construction, what with and a coordinated nominal complement. These formal properties of the pattern (that is, the specific phonetic opening sequence, the morphological form of the complements, and their syntactic behaviour with regard to co-ordination) are associated with particular meanings. With this micro-construction, the referents of the NPs denote the reasons associated with the state of affairs designated by the proposition in the matrix clause to which the what with construction is attached. This is an inherent part of the semantics of this micro-construction. However, as discussed above and elsewhere in the literature (for example Kortmann 1991, Felser & Britain 2007), the extent to which the speaker evaluates the matrix proposition neutrally, negatively or positively is pragmatically variable, and thus underspecified, and established only in a specific context of use. This pragmatic feature of the construction is complemented by various discourse features, namely the construction's typical association with informal speech or writing. The levels and properties of the micro-construction may be represented diagrammatically as in Figure 1. As we have seen, not all of the constructions involving what with have these properties; some do not have NP complements, some are absolute constructions. Figure 1. The what with NP and NP micro-construction Discourse Pragmatics Semantics Syntax Morphology Phonology

Typically informal Underspecified evaluation of reasons for state associated with proposition denoted by main clause predicate; typically negative evaluation Reason adjunct What with NP and NP wot wi6 ... and ...

Theory and data in diacbronic Construction Grammar 585

some involve the speaker's positive evaluation of the state of affairs denoted by the proposition in the main clause, and so on. Based on the existing studies of what with constructions discussed in this section, we can identify the following constructional types: a. those involving purely nominal complements (cf. (8a)); b. those where the non-finite verb has a different subject from the verb in the main clause (cf. (8b)); c. those where the non-finite eomplement has a subjeet eo-referential with that of the finite verb in the main elause (ef. (8e)). Eaeh of these represents the prototypieal meso-eonstruetion, a grouping together of various miero-eonstruetions. For instanee, within group (a) there are two types of miero-eonstruetion, one with eo-ordinated NP eomplements, and one with a single NP complement; each of these is a micro-construction. Furthermore, the nature of the constructional network allows intersection of the constructional types. This is iflustrated by example (15): (15)

The fort did look like a fair, what with all the tents and folks selling knives and blankets and stone jewelry. (COCA)

This construct is the product of the intersection of meso-eonstruetion types (a) and (b). As for the maero-eonstruetion, the relevant abstraetion here would be aeross the meso-eonstrustions (b) and (e), a maero-eonstruetion whieh eneompasses afl non-finite elausal eomplements of what with. And as for the diaehronie development, the hypothesis is that there wifl be a trajeetory from (a) to (e) above, sinee the development of eo-referential subjects shared between clauses is an indication of increased grammaticalization (see further Killie & Swan 2009 on the development of converb clauses involving -¿«g participles in the history of English). Evidenee of yet further grammatieal ehange would be a development of a new eonstruetional type (d), where the eomplement of what with is a finite elause. This would involve speakers abstraeting yet further: The maero-eonstruetion that is the generalisation aeross (b) and (e) would be insuffieient as the most general eonstruetion relevant to the ehange, beeause the most general eonstruetion would have to lieense both finite and non-finite elausal eomplements. I suggest, therefore, that what with eonstruetions are eonstruefions, in that they eonstitute eonventional symbolie units, formed as a result of abstraetions aeross usage events. They are infrequent, and involve a network of form-meaning pairings. This network is the produet of historieal evolution: The produetivity of the different synehronie patterns is a refleetion of aspeets of diaehronie change. The hypothesis regarding the historieal development is that the eomplexity oí what

586 Graeme Trousdale

with constructions observed by Eelser & Britain (2007) in tbeir synchronie corpus analysis is a product of grammatical constructionalization (see further Section 4.2 below). The micro-steps involved in change produce an array of related constructions at any synchronie moment. In order to test this hypothesis, the foUowing section provides a presentation and discussion ofthe bistorical data from corpora of Englisb.

3.

Data on the historical evolution ofthe what with construction

In tbis section, some data on tbe bistorical evolution of what with constructions is presented, and analysed from tbe perspective of grammatical constructionalization. To begin, however, I provide a brief summary of tbe arguments proposed by Eelser & Britain (2007) against treating what with as a grammaticalized prepositional complementizer (that is, where what with is treated as a constituent, a complex preposition), and restate the essence of their synchronie analysis ofthe what with construction described in detail in §2.2 above. Tbere are two principal reasons wby Felser & Britain do not treat what with as a grammaticalized prepositional complementizer. First, they adduce evidence from internet examples to support the claim that what with is not a constituent, because tbe sequence may be interrupted by adverbs like primarily, also, and even, as Ulustrated by (16): (16) As you can probably tell by now. Final Fantasy VIII is very different from its predecessors, what especially with the drastic innovations in its battle system. (= Eelser & Britain's example (37b)) Secondly, again using examples taken from tbe internet, tbey sbow that what in what with constructions can be 'expanded' to what all (which is itself a DP, McCloskey 2000), as in (17): (17) So ... what all with this desk job, it seems I have such various job tasks as data entry, switchboard operation, and signing someone else's name on letters. {= Felser & Britain's example (39a)) This is taken as evidence that what (all) with cannot be a phrasal head (tbat is, cannot be located in C). Recall from §2.1, above, tbat Eelser & Britain (2007) start from a position tbat {what) with is part of an absolute construction which takes a clausal complement. What look on the surface to be simple nominal complements are accounted for by virtue of a nuU clausal bead. A furtber historical development noted by Eelser & Britain (2007) is tbat tbe what with pattern with nominal complements (that is, where what with introduces a reason adjunct) is likely to have

Theory and data in diachronic Construction Grammar 587

preceded the use of what with as the first part of an absolute, given that the former were attested in the Middle English (ME) period (see §3.1 below), while absolute constructions were rare untU the Early Modern period (Denison 1998, Rissanen 1999, Rio Rey 2002). For this reason, I focus particularly on developments in the Late Modern English period. In the foUowing subsection, the development of the what with construction is explored in order to track the various micro-changes which have occurred: Section 3.1 is concerned with the very early history of the expression, and Section 3.2 uses a corpus of Late Modern English in order to track the change in more detaU in the recent history of English. Thereafter, an analysis of the development is provided, using the framework of grammatical constructionalization, focussing specificaUy on the kinds of expansion which typify the growth of grammatical constructions (Himmelmann 2004). 3.1 Up to Modern English Early evidence of the what with construction appears in the ME period. Example (18) below, from the late fourteenth century, Ulustrates the typical pattern associated with what with constructions noted in Section 2 above, namely that what with precedes coordinated elements (phrases or clauses), where what with is itself repeated. (18) So what with hepe and what with crok, so what with pruning hook and what with crook Thei make her maister ofie winne they make their master often win 'So by hook or by crook, they make it so their master ofien wins'. (c. 1393 Gower, Confessio Amantis, 5.2872) An even earlier example involves the use of Old English (OE) hwat 'what', with a different preposition, but nonetheless with a simUar meaning to the modern what with construction. In example (19) below, hwat 'what' cooccurs with/or (19) Allewebeoö in monifald wawe ine pisse all we be.p.iND in manifold woe in this wreche Hue, hwat for ure eldere werkes, wretched life, what for our elder.p work.p hwat for ure ajene gultes. what for our own.p guilt.p 'We are all in great unhappiness in this wretched life, what with the sins of our fathers, what with our own misdeeds'. (c. 1175 Lamb. Hom. 145)

588 Graeme Trousdale

Tbis use of hwat/what for persists into tbe ME period; further ME examples show yet more prepositions occurring with what {through in example (21)): (20)

What for calde & for holdyng in pe watir What for cold and for holding in the water I was nere-hand slayn. I was near-hand slay.p.PTCP 'What with the cold and being held in the water, I was very nearly killed'. {cl440 Alphabet of Tales, 13)

(21) For what thorugh werre and wikkede werkes For what through war and wicked work.p and wederes unreasonable... and weather.p unreasonable... 'For what with war, and wicked deeds, and unseasonable weather'. (1393 Langland, Piers Plowerman B-text, Passus XV.355) In the earlier history of Englisb, tben, tbere was a wide range of prepositions wbicb could occur foUowing hwat/what. The most frequent of tbese appears to be for, which is predominant in earlier texts {Oxford English Dictionary, s.v. what, II adv. or conj. 2. b) but otber examples include because of, between, by,from, in case (of), of, through, and of course with (see also Visser 1972:1158, 1271-1277 for more on tbe augmentation of absolute constructions in tbe bistory of Englisb). Tbe what P pattern could be repeated after tbe conjunction (as in (18) and (19)) or not (as in (20) or (21)). 3.2 Late Modern English 3.2.1 Method For the Late Modern period, a searcb of CLMETEV (De Smet 2005) was undertaken. Because tbe corpus is not parsed, it is not possible to searcb for any preposition foUowing what. In order to investigate tbe development of what with, a number of different searcbes were undertaken. Tbe first was a simple searcb for tbe string "wbat witb". Tbis returned a number of false positives of tbe type It's poisoned — I don't know what with [COCA]; tbese were discounted. Eacb instance of tbe what with construction was classified based on tbe following complement,' as ülustrated below: (22) NP complements and what with my diligent labour, and your poor mother's spinning, and your needle-work, I make no doubt we shall do better and better. {1740, Samuel Richardson, Pamela; CLMETEVl '")

Theory and data in diachronic Construction Grammar 589

Non-finite -/«^clauses without an overt subject Yes, I only got the order yesterday at noon; and there's three girls beside the mother; and what with trying on and matching the stuff (for there was not enough in the piece they chosefirst),I'm above a bit behindhand. (1848, Gaskell, Mary Barton; CLMETEV2) Non-finite -ing clauses with an overt subject and how, what with many being strangers coming by sea, and others being serfs fiedfrom home, they were a nameless, masterless sort, who knew not where to seek a parish priest. ( 1870, Yonge, Caged Lion; GLMETEV3) SmaU clauses a concealment so complete, what with the fane in front and the trees behind, that no unsuspicious passenger could possibly have detected him. (1834, Bulwer-Lytton, The Last Days of Pompeii; CLMETEV2) Mixed complements (for instance, a non-finite -ing clause with no overt subject coordinated with an NP) What with being tired waiting so long for convoy, and the knowledge that arriving before the other West Indiamen would be very advantageous, I made up my mind that, instead of beating up into the bay again, I would run for England without protection, trusting to the fast sailing of my vessel and the guns which I had on board. (1841, Marryat, Masterman Ready; GLMETEV2) The second search was for specific instances listed by Felser & Britain (2007) as instances of'interrupted' what with, that is, where an adverb intervenes. (RecaU that this was a feature of some of the internet data that Felser & Britain discovered). The specific strings that were searched for were what all with, what also with, what especially with, what primarily with, what now with, what just with, and what therefore with, that is, those strings which were attested in Felser & Britain (2007). The third search was for any prepositions which were attested in the OED as occurring foUowing whatlhwat in earlier stages of the language, such as those Ulustrated in examples (19)-(21) above. 3.2.2 Results Table 1 below gives the proportion and number of instances for each of the coordinated complement types which foUowed the sequence what with in GLMETEV, subdivided by a seventy-year period. Because the focus of the discussion concerns the types of complements which may be listed affer a single instance of what with, instances where what with is repeated, as in examples (23) and (24), are excluded. For similar reasons.

59O Graeme Trousdale

Table 1. What with eonstruetions in the CLMETEV corpus: eoordinate/Usted complements Period 1710-1780 1780-1850 1850-1920

NP 100% (20) 66.7% (12) 78.1% (25)

ing -subj 0 16.7% (3) 3.1% (1)

SC 0 5.6% (1) 0

ing +subj 0 0 6.2% (2)

en 0 0 0

Mixed 0 11.1% (2) 12.4% (4)

instanees with general extenders, sueh as and all in (25), are also exeluded from the frequeney eounts. (23) And what with the gipsy affair, and what with this, J could not think ofgoing down to dinner. (1740, Samuel Riehardson, Pamela; CLMETEVl) (24)

What with the arrestments on this night of the Twenty-ninth, what with those that have gone on more or less, day and night, ever since the Tenth, one may fancy what the Prisons now were. (1837, Thomas Carlyle, French Revolution; CLMETEV2)

(25) but I'm such a poor weak old body, and my head's so gone, and I'm so dazed like, what with Alice and all, that J think and think, and can do nought to help my child. ( 1848, Eflzabeth Gaskell, Mary Barton; CLMETEV2) For the seeond seareh, no instanees of any of the 'interrupted' sequenees, that is, what all with, what also with, what especially with, what primarily with, what now with, what just with and what therefore with, were found. For the third seareh, the other prepositions whieh oeeurred with what are between, by, and from, as illustrated in (26)-(28) below: (26) What between the frite and sickness, J thought I should have brought my heart up. (1771, Tobias Smoflett, Humphrey Clinker; CLMETEV 1 ) (27) for in a few minutes, what by her looks, and the case itself, Jfound myself as much embarrassed as it was possible the lady could be herself. (1768, Thomas Sterne, Sentimental Journey; CLMETEVl) (28) we might there procure Hogs, Fowls, and other refreshments. Articles that we have been very sparingly supply'd with at this last Jsland, as the Ship's Company (what from the Constant hard duty they have had at this place, and the two free use of Woman) were in a worse state of health than they were on our first arrival. ( 1773, James Cook, Journal; CLMETEV 1 ) Sueh forms appeared infrequently in all three subperiods of the eorpus.

Theory and data in diachronic Construction Grammar 591

3.3 Twentieth-century American English (COCA corpus) 3.3.1 Method A similar method was adopted for an investigation of contemporary American English, using the COCA corpus (Davies 2008-). Here, a random sample of 500 instances of what with patterns were subject to analysis. Tbe second and third searches (that is, for 'interrupted' what with, and for what foUowed by a preposition other than with such as what for and what between) were also carried out. 3.3.2 Results The results of the analysis of the 500 instances of what with are illustrated in Figures 2 and 3. Unlike the results for the Late Modern English corpus, where all the instances involved co-ordination of some kind (even if only with generalised extenders such as and all), the results from the COCA corpus have been subdivided into tbose involving co-ordinated (Figure 2) and non-coordinated (Figure 3) complements. Excluding a total of 23 examples oí what with one thing and another (n=17) and other idiosyncratic forms (n=6), only 52.4% of this sample of what with patterns appear in a hst/coordinated structure (n=251); furthermore, some of these co-ordinated complements contain grammaticalized extenders (Cheshire 2006) such as and all. 'Extender' forms {what with X and everything/and all/and the like) are by definition instances of coordinate structures. However, instances where only one explicit reason is given before tbe extender (for instance, what with the movie and all) outnumber instances wbere tbere is more tban one explicit reason listed (for instance, what with the explosion, car chase, Ukranians and all) by a ratio of 5:1. 'Verbless' variants (that is, those involving NP complements, or small clauses) are favoured in botb coordinate and noncoordinate structures. This is consistent with the analysis ofthe BNC dataprovided by Eelser & Britain (2007)." The second and third searches (i.e. for 'interrupted' what with, and for what followed by a preposition otber tban with sucb as what for and what between) returned no hits.

4

Grammatical constructionalization: A cognitive approach to language change

In this section, I summarise some of the changes that emerge from the data presented in the preceding section (§4.1 ). An account of those changes is given in §4.2 from the perspective of grammatical constructionalization.

592 Graeme Trousdale

4.1 Summary of the principal changes Two major types of change are evident in the development of what with constructions. One involves narrowing or reduction, the other broadening or expansion. The narrowing involves the set of prepositions which co-occur with what. In the earlier history of English, a more general what + P pattern was available to speakers; over time this has narrowed. Some of the patterns have simply faUen out of use (for example, what by and what from); others have developed a variant without what (for example, between his job and his family, it's no surprise they moved). This process suggests a greater entrenchment of a particular subvariant of the more general historical pattern what + P, where entrenchment is to be understood as the product of frequency of successful use (Langacker 1987:59) resulting

Distribution of complements in coordinate WWAs (%)

NP

Mixed

Extender ing + subj Ing-subj

SC

Figure 2. Percentage of complement types in coordinated what with constructions, COCA Distribution of compiements in non-coordinate WWAs (%) 100-1 9080 70 60 50403020100 NP

I

Ing -I- subj

SC

ing - subj

en

Other

Figure 3. Percentage of complement types in non-coordinated what with constructions, COCA

Theory and data in diachronic Construction Grammar 593

in a unit status — as what with is used witb greater and greater frequency in comparison to related patterns sucb as what between and what by as ways in English of introducing reason clauses, the status of what with as a grammatical unit increases. In traditional grammaticalization terms, this may be considered as a kind of univerbation, the obligatorification of a particular form selected from a set of related structures. Tbis aspect of tbe cbange fits weU with Haspelmath's definition of grammaticalization as "a diachronic change by wbich parts of a constructional schema come to have stronger internal dependencies" (Haspelmath 2004:26) — the bonds between what and otber prepositions in tbis construction are weakened, wbile tbe bond between what and with is strengtbened. Tbis narrowing is part of a more general cbange affecting tbe range of items wbicb could augment absolute constructions (Visser 1972:1158, 1271-7; Kortmann 1991:199). As Kortmann (1991:199) notes, augmentation of absolutes in contemporary Englisb is restricted to with and without, the use of what with and and being even more marginal. Tbis notion of grammaticalization — as obligatorification, and as reduction — is common in tbe traditional grammaticalization literature. But as Traugott (2010) observes, grammaticalization bas also been sbown to involve expansion and growtb, particvilarly witb reference to tbe development of discourse markers and clause connectives. We see tbis clearly in the development of the complements of what with in tbis construction. In tbe early history of the construction, the most frequent type of complement was a co-ordinated nominal (that is, what with NP and NP); over time, two constraints have been loosened — one constraint is tbat tbe form of tbe complements be NPs, tbe other is that tbe complements be coordinated. Tbe cbange may be represented scbematically as (29): (29)

what+with -1- NP + NP >what+with + XP {+ XP)

wbere XP is eitber a nominal or clausal complement (or, in tbe framework adopted by Felser & Britain (2007), wbere X stands for either D or T). Tbis diacbronic development of what with is consistent witb Himmelmann's expansion model (Himmelmann 2004) involving bost-class expansion, syntactic expansion, semantic-pragmatic expansion. Host-class expansion involves cbanges in collocation restrictions, exemplified bere by tbe sbift from nominal complements to clausal complements. Syntactic expansion often involves positional change, but here might be extended to include tbe possibility of a non-coordinated complement. Semantic-pragmatic extension may involve processes sucb as subjectification. In tbe case of what with, tbere appears to be an on-going change whereby the what with construction may attach to matrix clauses whose proposition does not denote a negative state. As noted by Kortmann (1991:202), noted above in Section 2, what with constructions do appear to be marked as subjective — tbe

594 Graeme Trousdale

negative (or other) evaluation of the implications of the proposition in the matrix clause are associated with the speaker's perspective or world view. The syntactic changes are summarised in Figure 4. 4.2 Grammatical constructionalization The process of grammatical constructionalization (Traugott 2008, Trousdale 2008, Traugott & Trousdale 2010) is concerned with the development of form-meaning pairings which develop a particular procedural, non-referential meaning, and which are characterised by changes associated with three particular parameters: generaUty, productivity and compositionality (see also Langacker 2005, and a related discussion in Bybee, Perkins & Pagliuca 1994). The foUowing changes to those parameters are: a.

an increase in semantic generality. As noted above, a semantic-pragmatic expansion is witnessed in the relaxation of the condition that the matrix proposition must denote a non-event/negative state (cf. Kortmann 1991). It might be observed that this is rather a weak expansion, when one considers, for instance, the extent to which semantic polysemies have developed. However, the greatest changes seem to have been in the realm of pragmatics. As Felser & Britain (2007) observe, even in their formal model of what with absolutes, the location of what as the head of an Evaluative Phrase underspecifies the positive or negative evaluation of the proposition. The diachronic evidence points to a change, whereby positive propositions associated with the matrix clause have increased in frequency. While semantic bleaching is often considered a hallmark of grammaticalization, there is little evidence of bleaching in the development of this construction. One possible reason for this is the fairly

Syntactic development I: narrowing ME hwat + P {for, ...]> EModE what + P {with,...} >PDE what with, between

> LModE what {with, between]

Syntactic development II: broadening To ModE LModE what Pj NP,8c what P. NP^ what with NP & NP {what P NP) (what with NP {8c TP}) {what PNP {8c what PTP}}

PDE what with NP 8c NP {what with NP 8c TP¡ {what with TP} {what with TP & TP}

Key: {} = alternatives; bold = most frequent variant in corpora used Figure 4. Syntactic developments associated with the what with construction

Tbeory and data in diachronic Construction Grammar 595

bleached origins of the various elements of the eonstruetion, that is, what and a preposition (see also Traugott 2008b, Patten 2010, this volume, on the grammatiealization of eonstruetions without lexieal souree); b. an inerease in syntaetie productivity. The host-class and syntactic expansions discussed above provide evidence for an increase in syntaetie produetivity. In the earliest CLMETEV period (1710-1780), only eoordinated NP eomplements were attested; by eontrast, in the COCA eorpus, we see evidence of noncoordinate structures, and an extended range of XP complements. Nonetheless, despite this expansion, the construction is stifl infrequent. Possible explanations for this infrequency must be tentative. As noted above, other grammarians who have written on the what with construction in English have comments on its infrequency and its soeiolinguistie patterns (that is, its tendeney to appear in informal diseourse), see for example Kortmann (1991); Quirk et al (1985); Huddleston & Puflum (2002). A possible reason for this infrequency may again be associated with entrenchment. While entrenchment links to the selection of what with over other eombinations oí what and a preposition, what with itself may be disfavoured over other very frequent (grammatiealized) heads of reason elauses whieh are faetive, like^¿ven or granted; e. a deerease in eompositionality. This deerease in eompositionality applies to the inereased idiomatieity that arises from the narrowing oí what H- P to what with. A eritical issue here is that ehanges do not simply affeet what + P + XP, but larger eonstruetional types. In grammatieal eonstruetionalization, different parts of a eonstruetion may ehange in different ways, with some parts showing the 'elassie' grammatiealization pattern of inereased dependeney and obligatorifieation (assoeiated with reduetion), and others showing how grammatiealization involves expansion and grov\^h. Felser & Britain (2007) argue that what with is not grammatiealized. Their argument is, first, that what with is not a eonstituent — adverbs may be inserted to give strings sueh as what also with, and seeond, that examples sueh as what all with suggest that what with eannot be loeated under a single (grammatiealized head). None of these forms were attested in the eorpus seareh; that said, the eonstruetion is fairly infrequent in the first plaee. But the data from the eorpus do provide evidenee that suggests that the what with eonstruetion is elearly taking part in the restrueturing of absolutes in English, developing stronger internal dependeneies, and licensing a wider range of complementation types. The oeeasional instanees where what and with do not form a unit show that there is variation, but does not provide evidence against the development of a process whereby a shift towards a more unit-like status is clearly in progress. There is clearly a gradual process of univerbation at work, even if the two items have not univerbated for every speaker.

596 Graeme Trousdale

Micro-Steps are an important part of establisbing bow grammatical constructionalization takes place. Eor instance, tbe minimalist position taken by Eelser & Britain (2007) suggests tbat a TP (clausal) complement sbould be considered basic, witb what appear to be DP complements being analysed as TPs witb nuU heads and pro in spec-TP position. However, as Eelser & Britain themselves note, the diacbronic evidence suggests tbat clausal complements of what with are newer, and nominal complements are the most frequently attested earlier forms. A constructional model, wbicb considers changes in sequences of form-meaning pairings, and focuses on contraction and expansion at different levels of the construction, provides an alternative analysis for the growth of the pattern, and the various grammatical changes involved. This approach, which concentrates on the "coevolution of meaning and form" (Bybee, Perkins & Pagliuca 1994:20) provides (a) a coherent account ofthe development ofthe construction and (b) a legitimate analysis of tbe syncbronic variation. Einally, these micro-steps in the development of the what with construction should, as noted in Section 2 above, be seen in tbe context of tbe development of free adjuncts and absolutes in tbe history of English. As Rio-Rey (2002) observes, corpus evidence in the Early Modern period shows the distinction between free adjuncts and absolutes to be less sbarp tban is tbe case in present-day Englisb. In tbe Early Modern period, free adjuncts witb non-coreferential subjects, and absolutes with overt subjects, while not common, were more widely attested. Over time, a functional split between tbe two constructional types has emerged, the development being "a gradual one towards tbe fulfilment of complementary tasks" (Kortmann 1991:103). The emergence of tbis more general pattern — the use of free adjuncts in cases of subject coreference, and tbe use of absolutes in cases of non-identity between tbe referents of subjects of tbe finite and non-finite verbs — suggests greater transparency and regularity in tbe form and function of the macro-constructions. At the end of Section 2 above, I bypotbesised that an increase in generality of the construction may involve yet further abstraction on tbe part of users of Englisb. Tbat hypothesis would be supported were tbere evidence of complements of what with that were finite clauses, rather than the non-finite clauses typicaUy associated with absolutes and free adjuncts. No such finite clauses appeared in the COCA corpus. However, there are some sporadic internet examples of what with preceding finite clauses, as in (30)-(32): (30) Paul Sereno knows this very well, what with he's been getting Todd Marshall to illustrate just about everything his team's beenfindingin Africa. [http://scienceblogs.com/tetrapodzoology/2008/02/day_6_silvisaurus.php; accessed 15th February 2011]

Theory and data in diachronic Construction Grammar 597

(31) My mom always mails a Christmas box to hubby and me, what with she's in the South and we're not. [http://saUanscorner.wordpress.com/2010/12/; accessed 15th February, 2011] (32) So I'm gonna quit and hand it in now. Leastways it don't matter, what with I'm outta here next week, but the hypothetical question is: if I was to turn this in for a grade, seeing as I spent two thirds of the time assigned, and carved up three quarters of the assigned surface, would that mean Iget a "B" on the project? Strictly hypothetically speakin', I mean. [http://wvkrw.wirelady.com/berrienwirecurranpage.html; accessed 15th February 2011] These could be early instances of further reanalyses of what with as a grammaticalized subordinator, capable of introducing both finite and non-finite clauses: This would constitute yet further constructional change. For speakers of English for whom constructs such as (30)-(32) above are grammatical, they are sanctioned by an even more abstract macro-construction, which allows complements of what with to be finite. Also at the end of Section 2, a further hypothesis was made regarding grammaticalization and the development of complement types of what with. Since non-finite clauses with controUed subjects (tbat is, where the covert subject of the non-finite verb is co-referential with the overt subject of the main clause verb) are said to be more integrated or bound to tbe main clause than is the case for non-finite clauses with overt, non-co-referential subjects, we might expect in cases of grammaticalization that the latter might develop after the former. The Late Modern English evidence does not support this hypothesis, since in Table 1 it is shown that there were three instances of constructions with co-referential subjects in the second period (1780-1850), and no instance of the other constructional type. However, the numbers involved are very smaU, and a study using a larger corpus of Late Modern English may reveal different patterns. The data from larger contemporary English corpora (that is, the BNC data used in Felser & Britain 2007, and the COCA corpus used in this article) show that what with constructions involving coreferential subjects are less frequent than those where each clause has a different subject. It seems to be the case that what with constructions involving nominal complements only are decreasing in frequency, with clausal complements on the increase over time.

5.

Conclusions

The research presented in this article contributes to ongoing investigations into the relationship between cognitive linguistics (especially construction grammars

598 Graeme Trousdale

of various kinds) and diachronic linguistics. Diachronic work is typicaUy 'data rich', and so it needs an appropriate theoretical formalism (of which Construction Grammar is just one kind). The focus in such accounts of language cbange is on micro-steps, piecemeal realignments in signs. This focus on micro-steps is an area of considerable convergence between generative and cognitive approaches to grammatical change (see further Roberts 2010). Given Felser & Britain's adoption (or adaptation) of Cinque's functional-bead projection, a minimalist analysis of the change (rather tban tbeir syncbronic analysis) is also likely to require attention to "structural micro-cbanges, witb gradualness emerging as tbe accumulation of sucb changes" (Roberts 2010:70). Formal syntacticians are increasingly engaging witb micro-variation (see for instance Henry 1995, Adger 2006, and Parrot 2007), and tbe relationsbip between I-, E- and S-languages (see furtber Adger and Trousdale 2007), allowing for convergence between variationist sociolinguistics and formal syntax; similar kinds of convergence between formal and cognitive approacbes to cbange would mean tbat "we may be getting sometbing rigbt" (Roberts 2010:70).

Corpus data CLMETEV = Corpus of Late Modern English Texts (Extended Version), see De Smet 2005 COCA = Corpus of Contemporary American English, see Davies 2008Oxford English Dictionary

Notes 1. I am grateful to Elizabeth Traugott, the editors of this special issue, and two anonymous reviewers for their comments on an earlier version of this article, and to Manuela Rocchi, whose research on the synchronie properties of what with prompted me to consider the historical development of the construction. All errors and shortcomings are my own. 2. The majority of the research presented by Eelser & Britain (2007) is primarily concerned with exploring the alleged constructional nature of the synchronie pattern discussed in this article, and not with its historical evolution. My purpose here is the converse: to focus on the historical evolution of the pattern, and to show how these are of relevance to the more general theoretical issues regarding the synchronie pattern raised by Eelser and Britain in their article. 3. I use the expression 'what with construction' to mean a conventional symbolic unit of English grammar, in the Construction Grammar sense of the term; I use the expression 'what with sequence' for non-Construction Grammar approaches to this aspect of English grammar. 4. Examples are taken from the Corpus of Contemporary American English (COCA; Davies 2008-) or the extended version of the Corpus of Late Modern English Texts extended version (CLMETEV; De Smet 2005) unless otherwise stated.

Theory and data in diachronic Construction Grammar 599

5. A further augmentation with and is possible (for instance, / was amazed that Moriarty was a ruthless criminal and him a professor as well!); Kortmann (1991:199) observes that in his corpus study, what with and and augmentation were both very rare, with about 1% (n=3) of all absolutes being augmented with these forms. 6. Examples such as (8c) suggests that there may be an ongoing change in the complementation patterns associated with what with, since Kortmann (1991:11) observes that free adjunct and absolute augmentors are in complementary distribution, as noted above. I return to this issue in Section 4, where the nature of the grammatical changes is explored in more detail. 7. Felser 8c Britain (2007), for purposes of their synchronie analysis, coUapse the distinction between free adjuncts (which are subjectless) and absolutes (which have an overt subject). 8. More recently, Goldberg (2006) has argued that a compositional form-meaning pairing may be considered a construction if it occurs sufficiently frequently. Tbe fact that what with patterns occur relatively infrequently in corpora would suggest that they are better analysed as idiosyncratic constructions, foUowing the earlier (Goldberg 1995) definition. 9. As Table 1 shows, no en clause complements were returned from this search. 10. GLMETEVl etc. refer to the subperiods of the corpus: CLMETEVl = 1710-1780, CLMETEV2 = 1780-1850 and CLMETEV3 = 1850-1920. 11. Felser 8cBritain (2007) split their BNC data into what with absolute [+predicate], and what with absolute [-predicate} and examine the distribution of complements witbin each group. They find that what with absolutes [+predicate} typically disfavour coordination, whue what with absolutes [-predicate} favour coordination.

References Adger, David. 2006. Combinatorial variability, journal of Linguistics 42. 503-530. Adger, David 8c Jennifer Smith. 2005. Variation and the minimalist program. In Leonie Cornips 8c Karen P. Corrigan (eds.). Syntax and variation: Reconciling the biological and the social (Current Issues in Linguistic Theory 265), 149-178. Amsterdam: John Benjamins. Adger, David 8c Graeme Trousdale. 2007. Variation in English syntax: Theoretical implications. English Language and Linguistics 11. 261-278. Bergs, Alexander 8c Gabriele Diewald (eds.). 2008. Constructions and language change (Topics in Linguistics. Studies and Monographs 194). Berlin8cNew York: Mouton De Gruyter. Brinton, Laurel 8c Elizabeth Closs Traugott. 2005. Lexicalization and language change (Research Surveys in Linguistics). Cambridge: Cambridge University Press. Bybee, Joan, Revere Perkins & William Pagliuca. 1994. The evolution ofgrammar: Tense, aspect, and modality in the languages of the world. Chicago: University of Chicago Press. Cheshire, Jenny. 2006. Discourse variation, grammaticaUsation and stuff like that. Journal of Sociolinguistics 11. 155-193. Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective (Oxford Studies in Comparative Syntax). Oxford: Oxford University Press.

6oo Graeme Trousdale

Croft, WUliam. 2005. Logical and typological arguments for radical construction grammar. In Jan-Ola Ostman & Mirjam Fried (eds.). Construction grammars: Cognitive grounding and theoretical extensions, 273-314. Amsterdam: John Benjamins. Croft, WiUiam & D. Alan Cruse. 2004. Cognitive linguistics. Cambridge: Cambridge University Press. Davies, Mark. 2008. The Corpus of Contemporary American English (COCA): 410+ miUion words, 1990-present. Available online at bttp://www.americancorpus.org De Smet, Hendrik. 2005. A corpus of late modern Englisb texts. ICAME Journal 29. 69-82. Denison, David. 1998. Syntax. In Suzanne Romaine (ed.). The Cambridge history of the English language volume IV, 1776-1976, 92-329. Cambridge: Cambridge University Press. Felser, Claudia & David Britain. 2007. Deconstructing what with absolutes. Essex Research Reports in Linguistics 53. 97-134. Fiscber, Olga. 2007. Morphosyntactic change: Functional andformal perspectives (Oxford Surveys in Syntax and Morphology). Oxford: Oxford University Press. Fried, Mirjam. 2008. Constructions and constructs: Mapping a sbift between predication and attribution. In Alexander Bergs & Gabriele Diewald (eds.). Constructions and language change (Topics in Linguistics. Studies and Monograpbs 194), 47-79. Berlin/New York: Mouton De Gruyter. Fried, Mirjam. 2009. Construction grammar as a tool for diacbronic analysis. Constructions and Frames 1. 261-290. Gelderen, EUy van. 2011. The linguistic cycle: Language change and the language faculty. Oxford: Oxford University Press. Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press. Goldberg, Adele E. 2006. Constructions at work: On the nature of generalization in language. Oxford: Oxford University Press. Haspelmatb, Martin. 2004. On directionality in language change witb particular reference to grammaticalization. In Olga Fiscber, Muriel Norde & Harry Peridon (eds.). Up and down the cline — the nature of grammaticalization (Typological Studies in Language 59), 17-44. Amsterdam: Jobn Benjamins. Henry, Alison. 1995. Belfast English and standard English: Dialect variation and parameter setting. Oxford/New York: Oxford University Press. Hilpert, Martin. 2008. Germanic future constructions: A usage-based approach to language change (Constructional Approaches to Language 7). Amsterdam: John Benjamins. Himmelmann, Nikolaus P. 2004. Lexicalization and grammaticization: Opposite or ortbogonal? In Walter Bisang, Nikolaus Himmelmann & Björn Wiemer (eds.). What makes grammaticalization?: A look from its fringes and its components (Trends in Linguistics. Studies and Monograpbs 158), 19-40. Berlin: Mouton de Gruyter. Huddleston, Rodney & Geoftrey K. Pullum. 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press. Hudson, Ricbard A. 2007. Language networks: The new word grammar. Oxford: Oxford University Press. Katz, Graham. 1993. The semantics of free adjuncts: Deriving tbe weak-strong distinction. In V. M. Lindbad & M. Gamon (eds.), MIT Working Papers in Linguistics 20. 129-143. KUIie, Kristin & Toril Swan. 2009. The grammaticaUzation and subjectification of adverbial -ing clauses (converb clauses) in English. English Language and Linguistics 13. 337-363.

Theory and data in diachronic Construction Grammar 601

Kortmann, Bernd. 1991. Free adjuncts and absolutes in English: Problems of control and interpretation. London: Routledge. Langacker, Ronald W. 1987. Foundations of cognitive grammar volume 1: Theoretical prerequisites. Stanford: Stanford University Press. Langacker, Ronald W. 2005. Construction grammars: Cognitive, radical and less so. In Francisco J. Ruiz de Mendoza & Sandra Peña Cervel (eds.). Cognitive linguistics: Internal dynamics and interdisciplinary interaction, 101-159. Berlin: Mouton de Gruyter. McCloskey, James. 2000. Quantifier ftoat and w/i-movement in an Irish English. Linguistic Inquiry n. 57-84. Parrott, Jeffrey K. 2007. Distributed morphological mechanisms ofLabovian variation in morphosyntax. Georgetown University dissertation. Patten, Amanda L. 2010. Cleft sentences, construction grammar and grammaticalization. University of Edinburgh dissertation. Pintzuk, Susan & Eric Haeberli. 2008. Structural variation in Old English root clauses. Language Variation and Change 20. 367-407. Ouirk, Randolph, Sidney Greenbaum, Geoft^rey Leech & Jan Svartvik. 1985. A comprehensive grammar ofthe English language. London: Longman. Rio-Rey, Carmen. 2002. Subject control and coreference in Early Modern English free adjuncts and absolutes. English Language and Linguistics 6. 309-323. Rissanen, Matti. 1999. Syntax. In Roger Lass (ed.). The Cambridge history ofthe English language, volume III, 1476-1776, 187-331. Cambridge: Cambridge University Press. Roberts, Ian G. 2010. Grammaticalization, the clausal hierarchy, and semantic bleaching. In Elizabeth Closs Traugott & Graeme Trousdale (eds.), Gradience, gradualness and grammaticalization (Typological Studies in Language 90), 45-73. Amsterdam: John Benjamins. Rupp, Laura. 2005. Constraints on nonstandard -s in expletive there sentences: A generativevariationist perspective. English Language and Linguistics 9. 225-288. Stump, Gregory. 1985. The semantic variability of absolute constructions. Dordrecht: Reidel. Traugott, Elizabeth Closs. 2007. The concepts of coercion and type-shifting from the perspective of grammaticalization. Cognitive Linguistics 18. 523-557. Traugott, Elizabeth Closs. 2008a. The grammaticalization of NP of NP constructions. In Alexander Bergs & Gabriele Diewald (eds.). Constructions and language change (Topics in Linguistics. Studies and Monographs 194), 23-45. Berlin/New York: Mouton De Gruyter. Traugott, Elizabeth Closs. 2008b. "All that he endeavoured to prove was...": On the emergence of grammatical constructions in dialogic contexts. In Robin Cooper & Ruth Kempson (eds.). Language in ftux: Dialogue coordination, language variation, change and evolution (Communication, Mind, and Language), 143-177. London: Kings College Publications. Traugott, Elizabeth Closs. 2010. Grammaticalization. In Silvia Luraghi & Vit Bubenik (eds.), A companion to historical linguistics. London: Continuum, 269-283. Traugott, Elizabeth Closs & Graeme Trousdale. 2010. Gradience, gradualness and grammaticalization: How do they intersect? In Elizabeth Closs Traugott & Graeme Trousdale (eds.), Gradience, gradualness and grammaticalization (Typological Studies in Language 90), 1944. Amsterdam: John Benjamins. Trousdale, Graeme. 2008. Grammaticalization, constructions and the grammaticalization of constructions. Paper presented at New Reflections on Grammaticalization 4, KU Leuven, 16-19 July 2008. Visser, E Th. 1972. An historical syntax ofthe English language: Syntactical units with one verb. Leiden: Brill.

6o2 Graeme Trousdale

Author's address Graeme Trousdale Linguistics and English Language School of Philosophy, Psychology and Language Sciences University of Edinburgh Dugald Stewart Building 3 Charles Street EdinburgbEH8 9AD, United Kingdom [email protected]

Copyright of Studies in Language is the property of John Benjamins Publishing Co. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.