Tracing Grammaticalisation in English Light Verbs A Corpus-based Study

This talk will examine the properties of LVs to reveal whether they are instances of ‘little v’. Tracing grammaticalisation in English light verbs: A corpus-based study 1. Introduction 1.1 The phenomenon a. b. c. (5) full verb > (vector verb) > auxiliary verb > clitic > affix (Hopper & Traugott 2003) (6) V > v > T > clitic > affix As an example, consider Roberts & Roussou’s (2003) account of the grammaticalisation of English will: This talk is concerned with English Light Verb Constructions (LVCs) of the type in (1a). (1) (ii) Roberts & Roussou (2003) and Roberts (2010) provide a Minimalist account of grammaticalisation. Under their analysis, the grammaticalisation cline in (5) involves ‘upward reanalysis’ as in (6). (7) She took a sip of champagne then put the glass down. (BNC, HA7 4115) take a sip (of …): Light Verb (LV) + a/an + zero-derived deverbal noun to sip a. [TP b. c. [TP (i) The focus in this talk is on the set of LVs given in (2a); other verbs may be added to this set, (2b) (from Huddleston & Pullum 2002: 296). (2) a. b. do, give, have, make, take pay (a visit), offer (an apology), raise (an objection), put (an end to), cast (a glance) These LVCs belong to a larger class of constructions often referred to as composite predicates (see e.g. Brinton & Akimoto 1999). LVs (and LVCs) raise some interesting synchronic and diachronic questions (see e.g. Butt 2003, 2010) and this talk addresses the following two questions concerning LVs: (i) (ii) What is the syntactic category of LVs? Do LVs grammaticalise? [T will ‘future’] [VP [V V]]] Structural change [TP V+T [VP tV TP]] > [TP T VP] (ii) Parametric change T*Move > T*Merge (iii) Cause Loss of infinitive marker (Roberts & Roussou 2003: 195) I II III IV ‘Merge’ (T) Semantic shift (desemanticisation) Morphosyntactic shift (decategorialisation) Morphophonological shift (cliticisation) Phonetic shift (erosion) The table in (9) gives the number of LVCs of the type in (1a) in Early Modern English and Late Modern English. (9) Corpus study of LVs featuring in LVCs of the type in (1a) in the Modern English period (1500-1900), when this type of LVC became more frequent (see Brinton 2008). The corpus study makes use of the syntactically annotated historical corpora in (3). The data searches were done using CorpusSearch 2 (Randall et al. 2005). (3) Corpus Penn-Helsinki Parsed Corpus of Early Modern English (Kroch et al. 2004) Penn Parsed Corpus of Modern British English (Kroch et al. 2010) Period 1500-1710 Subperiods 1500-1569 (E1) 1570-1639 (E2) 1640-1710 (E3) 1700-1769 1770-1839 1840-1914 1700-1914 Word count 183,397 (Penn 2) 222,396 (Penn 2) 180,924 (Penn 2) 298,764 368,804 281,327 EModE 1500-1569 0 4 5 4 0 Do Give Have Make Take (i) This study adopts Chomsky’s (1995 et seq.) Minimalist Program. In this approach, there is a straightforward categorial candidate for LVs, ‘little v’. [T auxiliary verb] [vP [v light verb?] [VP [V lexical verb]]]] 1 1570-1639 0 0 11 6 4 LModE 1700-1769 1 11 23 24 9 1640-1710 0 2 7 1 4 Total 1770-1839 0 10 28 13 20 1840-1914 2 11 19 10 6 3 38 93 58 43 An example of each LV is given in (10). (10) a. b. c. d. 1.4 Approach [TP ‘Move’ (V > T) 2. Data: the properties of Modern English light verbs 1.3 Method (4) [V will ‘wish, desire’]]] The examination of the properties of LVs will reveal whether they show signs of any of the grammaticalisation processes in (8). (8) 1.2 Research questions [T ] [VP e. Isa. I almost fancy this Mr. Welby has done a Work in a few Hours, … (DAVYS-1716, 34.363) And the people gave a shoute, sayinge: it is the voyce of a God and not of a man. (TYNDNEW-E1-P2, XII,20A.543) While Cuthred the West-Saxon had a fight with Ethelhun, … (MILTON-E3-P2, X,182.98) There happened to be some reflections upon the priests which the Whigs clapped extremely and the Tories made a faint hiss. (RYDER-1716, 181.579) The turnkey takes a fixed look at him, … (DICKENS-1837, 548.117) 2.1 Semantics The LVs being studied here belong to a group of ‘general action verbs’. As LVs, these verbs have a basic action/activity meaning ‘do, perform’, illustrated in (11) for the examples in (10). 2 (11) ‘do, perform’ (a) work a shout a fight a faint hiss a fixed look (=10a) (=10b) (=10c) (=10d) (=10e) There are some argument-structural differences between LVs and their full-verb counterparts, e.g. LV give can be 2place or three-place (full verb give is three-place), as illustrated in (18); and LVCs do not readily allow passivisation, as illustrated in (19), which gives the numbers for ModE, and by the examples in (20). (18) a. b. LVs are not empty of meaning, as illustrated by the PDE examples in (12) (from Quirk et al. 1985: 752). (12) a. b. c. She gave a shriek. [an involuntary shriek] She had a good shriek. [voluntary and for own enjoyment] She did a (good) shriek. [a performance before an audience] LVCs are often Accomplishments (compare the counterpart simplex verbs, which are typically Activities). (13) a. b. LV: Activity/Accomplishment ‘in X time’ (=10a) ‘has done a Work in a few Hours’ (=10d) made a faint hiss for/in ten seconds a’. b’. simplex V: Activity ‘for X time’ has worked for/*in a few hours hissed for/*in ten seconds (19) Passivisation Do Give Have Make Take (20) a. The zero-derived deverbal noun typically derives from a MANNER verb. b. (14) c. a. b. b’. (=10b) ‘gave a shute’ ‘to shout’ Event schema (Rappaport Hovav & Levin 1998: 109): manner → [x ACT] Lexical verbs tend to have more than one component of verb meaning (e.g. ACT and MANNER in the case of manner verbs), but LVs only have the component ACT. The event schema for the (intransitive) manner verb to shout in (14b’) represents Activity Aktionsart. The event schema for the LVC to give a shout represents Accomplishment Aktionsart and a possible schema is given in (15) (from Rappaport Hovav & Levin 1998: 104). (15) [[x ACT] CAUSE [BECOME [y ]]] The LVC nominal is an ‘effected object’, i.e. an object which comes into existence as a result of the action described by the verb. It typically measures out/delimits the event described by the verb. This state of affairs is in line with Heine’s (1993: 35) observation that the (transitive) Action schema “X does Y” often gives rise to “perfect or terminative aspect categories” and that “[p]rototypical verbs figuring in the Action Schema are action verbs such as “do”, “take,” etc.”. The thematic force is located in the LVC nominal rather than in the LV, as illustrated in (16). (16) a. b. He deliberately had a look. *He deliberately had a sneeze. LVs have a full inflectional paradigm; (17) presents some ModE examples. a. b. c. d. and left the prince to take a short and sad leave of his Imoinda; (BEHN-E3-P2, 171.19) to-inf Captain Davis could not give a favourable reply to Mr. Barker's question on the subject, (MONTEFIORE-1836, 162.447) inf We had a hearty laugh over it. (THRING-187X, 219.112) past thus, in reading the word, Composition, they would not read it at once, but bysyllables: thus, Com-po-si-ti-on; making a pause at every syllable. (LANCASTER-1806, 57.344) prog ptc 3 EModE 0 1 2 2 2 ModE 1 4 0 6 0 Total 1 5 2 8 2 but a still larger trade is done through the post, by means of exhibitions, and by advertising in the papers. (WEATHERS-1913,1,4.72) But an especial care must be taken, and a charge accordingly often given, that your Scholars do at no time play with any but their own Schoole-fellowes, (HOOLE-E3-P2,246.78) and that therein a care be had, that they ever touch both the lines of the ruling pen with the bodies of their letters. (BRINSLEY-E2-P2,39.208) In what shape they wou’d severally come out, tho’ a good Guess might be made, was not then demonstrable to the deepest Foresight; (CIBBER-1740,37.59) Passivisation is also restricted with Cognate Object Constructions (COCs, e.g. smile a sad smile). According to MacFarland (1995), the scarcity of COC passivisation is due to the non-referential status that the nominal is often associated with, which makes them non-suitable for being topics. This is true of LVC nominals too. The non-referentiality of LVC nominals is connected to their being effected objects: according to Hopper (1985), effected objects differ from affected objects in that effected objects are non-referential, which means that they cannot be described as Patients. There is some cross-linguistic evidence for the correlation between affectedness and accusativity: Hopper & Traugott (2003: 96-97) note that the West African language Gã has an accusative case marker k , which was originally a verb meaning ‘take’ (they refer to Lord 1993: 53-56) and which can only be used if the object is affected. This restriction is retained “from the historical antecedent of this grammatical morpheme in the lexical verb ‘to take’: only objects which can be taken are marked morphologically as accusatives” (Hopper & Traugott 2003: 97). LVs pattern like lexical verbs in the major syntactic developments of loss of finite verb movement and the rise of dosupport. They highly favour the VP-internal position, close to their nominal: there are no ModE examples showing verb-movement of the LV. Examples of do-support are presented in (21). (21) 2.2 Morphosyntax (17) d. and then, thrusting his head out of the window, gave a shrill whistle, … (DICKENS-1837, 561.432) when they see one, they give a smile which would bewitch a saint imagine the effect upon a sinner!, … (READE-1863, 223.554) a. b. c. so that if my Life did not then take a more laudable Turn, … (CIBBER-1740, 45.148) - damn me, if I $do $n't have a slap at him! (COLLIER-1835, 29.1079) for of thys last se~tence many me~ at the first hearing do make a doubt. (RECORD-E1-P2, 2,B4R.102) ModE LVCs are readily modifiable, as illustrated by the table in (22) and the examples in (23). 4 (22) EmodE + LModE Do Give Have Make Take (23) a. b. (28) Modified 2 (66.7%) 18 (47.4%) 50 (53.8%) 19 (32.8%) 23 (53.5%) Non-modified 1 (33.3%) 20 (52.6%) 43 (46.2%) 39 (67.2%) 20 (46.5%) Total 3 38 93 58 43 Mabel Chiltern gives a little nod. (WILDE-1895, 59.542) Who, albeit he made a discreete qualified awnswer, neuertheles was forthwith comitted to the tower. (ROPER-E1-P2, 74.64) The LVC nominal regularly takes a complement and appears to retain the argument structure of the lexical verb it derives from, as shown in (24) and illustrated in (25). (24) EmodE + LModE Do Give Have Make Take (25) a. b. Complement 0 21 (55.3%) 62 (66.7%) 19 (32.8%) 19 (44.2%) No complement 3 (100%) 17 (44.7%) 31 (33.3%) 39 (67.2%) 24 (55.8%) Total 3 38 93 58 43 And now I have a bite at another: (WALTON-E3-P2, 252.280) and when they had appointed where to meet the next time, they took a kind leave of one another, … (PENNY-E3-P2, 165.149) As (25b) shows, modifiers and complements can co-occur. 3. Analysis: the status of Modern English light verbs There is clear semantic and morphosyntactic evidence that LVs are grammaticalised lexical verbs, as indicated in (26) (cf. (8)). (26) I II Semantic shift Loss of lexical (referential) meaning, gain in aspectual meaning Morphosyntactic shift The verb’s complement is a nominalised verb and is a non-referential/non-Patient object (loss of prototypical transitive properties) LVs are in the initial stages of grammaticalisation (see also Brinton & Traugott 2005; Brinton 2008). While LVs are grammaticalised, they are different from Chomsky’s (1995) ‘little v’: LVs are not clearly associated with accusative Case and they are more than simply causal operators. I propose instead that LVs are instances of V, and that the properties of LVs can be understood in terms of ‘subsective gradience’ (Aarts 2007), defined in (27): (27) “Subsective gradience is found when X and Y are in a gradient relationship within the same form class. This is a question of prototype vs. marginal members of a category…” (Brinton & Traugott 2005: 150) Lexical verbs and LVs are in a gradient relationship within V. Lexical verbs are the prototype members of V, LVs are marginal members of V. Following Borer (1994), Slabakova (1997), Ritter & Rosen (2005), Travis (2010), among others, I propose the syntactic structure in (28). 5 vP ei v’ ei v AspP ei Asp’ ei Asp VP [+measure] ei V’ ei V DP lexical verbs light verbs The Aspect Phrase (AspP) directly contributes aspectual information to the clause: Asp stands for ‘event measure’, and SpecAspP licenses the measure interpretation associated with the object of accomplishment verbs (Borer 1994). Slabakova (1997) proposes that the BECOME operator in the LCS of accomplishment verbs (see e.g. (15)) projects an AspP in syntax. Ritter & Rosen (2005) propose that a feature [QUANT] is present on Asp when the event is quantized. I propose that the LV lacks an accusative Case feature (cf. the non-referential, effected object); the movement of the LV nominal to SpecAspP is motivated by the interpretive (aspectual) feature [+measure], not by the formal feature of Case (cf. Borer 1994). This means that the diachronic development of LVs involves the loss of the uninterpretable feature [+ACC] (accusative Case), but also the gain of an aspectual [+measure] feature. LVs move from V > Asp > v, as indicated in (29). This means that there has not (yet) been a reanalysis from V > Asp (Move to Asp > Merge in Asp) in the sense of Roberts & Roussou (2003) and Roberts (2010). (29) vP ei v’ ei v AspP ei DP Asp’ ei Asp VP [+measure]ei V’ ei V tDP Given Economy, which favours Merge over Move, reanalysis is expected to occur when (uninterpretable) features are lost. I assume that this reanalysis has not (yet) occurred because LVs still retain enough featural content of lexical verbs to be merged in V (i.e. to have a subsective relationship with lexical verbs). 4. Conclusions English LVs in LVCs with a indefinite zero-derived deverbal nominal are in the initial stages of grammaticalisation and appear to have been quite stable since Early Modern English. 6 Different LVs may be at different stages on the grammaticalisation cline. As an example of a LV that has grammaticalised further, consider the aspectual verb go (see also come) in (30): (30) a. b. Let’s go take a look at Green Meadow Farm.’ (BNC, CAM 380) [vP [v ] [ AspP [ Asp go] [VP [V take] [DP a look at …]]]] LVs of the type discussed in this talk are marginal members of the category V, which has lexical verbs as its prototypical members. This difference in membership status is reflected syntactically by a featural difference. We therefore have a case of ‘subsective gradience’ (Aarts 2007), which arose as a result of a grammaticalisation development from lexical verb (prototypical V) to LV (marginal V). The analysis proposed here combines two seemingly opposite views of gradience, one in which categories are considered gradient (‘subsective gradience’, Aarts 2007) and one in which categories are considered non-gradient (Roberts & Roussou 2003). However, as Roberts (2010: 52) points out, these two views can be combined by making use of the idea that categories are bundles of features: “subsective gradience involves a gradient relation between two feature bundles such that one is a subset of the other”. The question is why LVs have not (yet) grammaticalised further (e.g. from V > Asp > v in a Roberts & Roussou 2003 style approach). Brinton & Traugott (2005: 151) note that “the beginnings of structural auxiliarihood are identified when a verb has ceased to have prototypical morphology and distribution”. As we have seen, LVs in the type of LVC discussed here still have much of the morphosyntactic properties of lexical verbs. It is important to note that not all LVs necessarily travel the same distance along the grammaticalisation cline (see e.g. (30a)). LVs combine with a number of different complements, creating their own dynamics and giving rise to LVC templates allowing new additions to be made (cf. particle verbs). The focus of this talk has been on the development of LVs rather than on the development of LVCs. With regard to the latter, Brinton (2008) has argued that LVCs grammaticalise too. Evidence for this comes from e.g. their productivity and host-class expansion (i.e. the set of deverbal objects increases over time). 7