See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/259092209 Bilingual serial verb constructions: A comparative study of Ewe-English and Ewe-French codeswitching Article  in  Lingua · December 2013 DOI: 10.1016/j.lingua.2013.08.002 CITATIONS READS 3 39 1 author: Evershed Kwasi Amuzu University of Ghana 18 PUBLICATIONS   49 CITATIONS    SEE PROFILE Some of the authors of this publication are also working on these related projects: African Humanities Fellowship of the American Council of Learned Societies View project Self-sponsored project View project All content following this page was uploaded by Evershed Kwasi Amuzu on 05 December 2017. The user has requested enhancement of the downloaded file. Available online at www.sciencedirect.com ScienceDirectLingua 137 (2013) 19--37 www.elsevier.com/locate/lingua Bilingual serial verb constructions: A comparative study of Ewe-English and Ewe-French codeswitchingEvershed Kwasi Amuzu * Department of Linguistics, University of Ghana, Legon, Accra, GhanaReceived 10 May 2012; received in revised form 8 August 2013; accepted 8 August 2013 Available online 9 September 2013 Abstract The paper compares serial verb constructions (SVCs) in Ewe-English and Ewe-French codeswitching (CS) spoken in Ghana and Togo respectively. It argues that Ewe, the Matrix Language (ML) in both cases, sets the morphosyntactic frames of bilingual SVCs and, thus, determines their structural possibilities. It demonstrates this by looking at the various properties of SVCs in monolingual Ewe (including monoclausality and the expression of aspect and modality categories) and comparing them to the ones found in Ewe-English and Ewe-French CS structures. It also demonstrates this by looking at the expression of complex motion using Talmy’s (2000) typology. Although English and French belong to different types with English being satellite-framed and French being verb-framed, and, although neither language has SVCs, complex motion is expressed in CS with SVCs. The facts are accounted for by using Myers-Scotton’s (1993, 2002) Matrix Language Frame model. One major significance of the paper is that it is the first cross-linguistic study of bilingual SVCs. It predicts that for bilingual SVCs to be characteristic of CS, the ML has to have SVCs even if the other language, the embedded language, does not have them. © 2013 Elsevier B.V. All rights reserved. Keywords: Bilingual serial verb constructions; Complex motion; Codeswitching; Matrix language; Ewe-English; Ewe-French1. Introduction1 A good number of languages in West Africa, Southern Asia, Amazonia, Oceania, New Guinea, and the Caribbean are known (see the collection of papers in Aikhenvald, 2006) to express certain types of events by means of two or more verbs occurring in complex constructions called serial verb construction (SVC). The SVC may be described, following Aikhenvald (2006:1) and Ameka (2006:128--129), as having the following characteristics:i. it* 1 DE tive Pot ma 002 http is a monoclausal construction in which two or more verb phrases (including any complements and adjuncts) appear as a single predicate without any overt marker of coordination, subordination, or syntactic dependency of any other sort;ii. the verbs in sequence combine to describe what is conceptualized as a single event; iii. all the verbs in it share one syntactic subject that is expressed only once, with V1;Tel.: +233 243754406. E-mail addresses: a_evershed@yahoo.com, evershed2010@gmail.com. Abbreviations used in interlinear glosses are as follows: ADDR = Addressive; ALLAT = Allative; COMP = Complementizer; COP = Copula; F = Definiteness marker; FOC = Focus marker; HAB = Habitual aspect; INDEF = Indefiniteness marker; ITIVE = itive; IMPERF = Imperfec- ; LOG = Logophoric pronoun; MOD = Modal; NEG = Negative; NP = Noun Phrase; PL = Plural marker; POSS = Possessive linker; POT = ential; PROG = Progressive; PROSP = Prospective; PRES = Present tense Q = Question marker; REDU = Reduplication; SG = Singular rker; SVC = serial verb construction; TP = Topic marker; VENTIVE = Eventually; VP = verb phrase; 1, 2, 3 = 1st, 2nd, 3rd person. 4-3841/$ -- see front matter © 2013 Elsevier B.V. All rights reserved. ://dx.doi.org/10.1016/j.lingua.2013.08.002 20 E.K. Amuzu / Lingua 137 (2013) 19--37iv. w2 3 pro 4 lan Wo and Kaithin the SVC, the individual verbs may have the same or different transitivity values; v. where applicable when two (or more) verbs in sequence share an object argument, the object argument appears only once with the first of the verbs; vi. the verbs can be formally marked for the same or different aspect and modality categories. Most of these characteristics are observed in the following example from Ewe, a Gbe language of the Kwa family spoken in West Africa.(1) Amu The gres The guag lof (E Gu nuri Police-wo zu (2010) is comparative s. re are many es, notably En nglish in Gam inea), Kru (En (English in Nigme the publ analysis other cro glish an bia, Fre glish in L eria, Frlé1-na ished version o of bilingual S ss-border We d French: Akan nch in Senega iberia, French ench in Chad),wóf my 2 VCs in st Afric and Nz l), Yoru in Côte etc.tu2-na 005 Ph.D. Th Ewe-English an languages ema (English ba (English in d’Ivoire), Suso esis. vs. Aka that ar in Ghan Nigeria, u (Englita-e Police-PL NEG catch-HAB 3PL lock-HAB NEG head-FOC ‘it is because the police do not arrest and lock them up.’Occurring without any overt marker of coordination, the verbs lé ‘catch’ and tu ‘lock’, which are conceptualized together as a two-in-one event, share common subject (police-wo) and object (wó: 3PL) and are marked similarly for the habitual aspect. The example also illustrates another characteristic of Ewe SVCs, which is that negation is marked only once by the discontinuous negation morpheme me. . . o, which has scope over the two VPs. Detailed illustrations of all seven characteristics of the Ewe SVC appear in section 2. There is a large body of literature on SVCs, including much work on West African languages (e.g. Bamgbose, 1974; Osam, 1997; Collins, 1997; Dorvlo, 2008; Ameka, 2006) and on pidgins and creoles especially in the Caribbean that may have West African languages as substrates (e.g. Sebba, 1987; Mufwene, 1990; Veenstra, 1996). However, no major work has appeared in which SVCs involving codeswitching (CS) are the subject matter. The few that have touched on this subject matter include Forson (1979) and Quarcoo (2009), on Akan-English CS, and Amuzu (1998, 2005b[2010]2), on Ewe-English CS. Instructively, all of them have arrived at the same conclusion, that structural possibilities in bilingual SVCs closely relate to structural possibilities in SVCs in the West African languages involved; i.e. each of them have characterized the West African language involved as the language of grammatical structure (or matrix language/ML) of the bilingual SVCs. Essizewa (2007:233) provides an additional kind of insight about the use of bilingual SVCs in CS. In reporting that he did not find any in his Kabiye-Ewe CS data, he theorizes that their absence is due to the fact that Kabiye, the ML in Kabiye-Ewe CS, does not have SVCs (i.e. despite the fact that Ewe, the embedded language/EL, has them). Thus, it is evident from the literature that bilingual SVCs are possible in CS only when the ML has SVCs. No cross-linguistic study of bilingual SVCs has been carried out, i.e. to the best of my knowledge, although two kinds of such study are possible at least in West Africa. In one kind, bilingual SVCs involving a European language and different West African languages (as in Ewe-English CS vs. Akan-English CS) may be compared with the aim of determining whether and to what extent cross-linguistic similarities and/or differences are traceable to similarities and/or differences between the West African languages.3 In the other kind of study, bilingual SVCs involving a West African language and different European languages may be compared with a view to finding out how similarly or differently the West African language accommodates elements from the different European languages. The present study is of this kind. The paper discusses examples of SVCs that consist of elements from Ewe, and either English or French. That is, it is a qualitative comparative analysis of bilingual SVCs from Ewe-English CS, which as noted is spoken in Ghana, and Ewe-French CS, spoken in Togo. The principal research questions addressed are:a. How much of the grammatical similarities and/or differences between Ewe-English and Ewe-French SVCs may be traced to Ewe?b. How much of the similarities and/or differences may be traced instead to English and French (bearing in mind, however, that neither language has SVCs)? This is the first of this kind study in West Africa although Ewe is not the only cross-border West African language that is in a contact-situation with two colonial or non-African official languages.4 Previous studies that are cross-linguistic inn-English is the subject matter of another paper (by this author) in e also in separate codeswitching contact with at least two colonial a, French in Côte d’Ivoire), Hausa (English in Nigeria, French in Niger), French in Benin), Fula/Pulaar (English in Gambia, French in Senegal sh in Sierra Leone, French in Guinea, Portuguese in Guinea Bissau), E.K. Amuzu / Lingua 137 (2013) 19--37 21character, have instead compared bilingual constituents involving either English or French and different West African languages.5 Unless otherwise acknowledged, the examples of bilingual SVCs discussed in the paper come from recordings made between April 2009 and October 6 2011 by a team of researchers whom this writer led on a project called ‘‘Ewe Contact Research Project’’, which was sponsored by the Office of Research, Innovation and Development (ORID) of the University of Ghana. The Ewe-English CS data, totalling twenty six hours of conversation, were recorded in Accra, Ho, Keta, and Akatsi (all in Ghana) while the Ewe-French CS data, totalling twenty five hours of conversation, were recorded in Lome and Avepozo in Togo. The same methodology was employed in collecting both sets of data. Speakers (never more than six per recording session) were organized into a focus group to discuss various topics (including marriage ceremonies, marital life, domestic and religious issues, current affairs---especially political and economic---and the relevance of science and technology in human lives). Members of each group were native speakers of Ewe who are fluent in English (Ghana) or French (Togo). Each session was moderated by a research assistant who is competent in Ewe and English or in Ewe and French. Each session lasted at least one hour. The general format was for the moderator to introduce a topic and to ask a series of general questions that stimulate a debate. As the various speakers take turn to address the questions (an event that is often characterized by rebuttals, overlaps and reformulation of the questions by the speakers themselves), the moderator’s major task was to ensure that everyone gets an equal opportunity to speak. Speakers are not instructed to speak any particular variety of language (Ewe or English or French or Codeswitching) although the recordings show that most of the moderators used monolingual Ewe predominantly. The CS data that resulted can therefore be seen as recordings of spontaneous speeches by the speakers. All the conversations recorded have been transcribed and typed out, making it easy to pick out the examples of bilingual SVCs discussed in the paper. In section 2, structural possibilities in bilingual SVCs in Ewe-English CS and structural possibilities in bilingual SVCs in Ewe-French SVCs are described and compared, using monolingual Ewe SVCs as the points of reference. That section will therefore be organized in terms of the characteristics of SVCs listed above. By the end of the section, it is made clear that there are very few differences between the characteristics of bilingual SVCs across the data bases and that the differences closely mirror differences between dialects of Ewe spoken in Ghana (e.g. Anlo) and the dialect (i.e. Gengbe/ Mina)7 spoken by Ewe-French codeswitchers in Lome, Togo. Section 3 presents theoretical analysis of patterns of CS described in section 2. It is shown how the patterns support predictions made by Myers-Scotton (1993, 1995, 2002) in her Matrix Language Frame model, especially in the model’s ‘‘Morpheme Order Principle’’, ‘‘System Morpheme Principle’’, and Blocking Filter. In the last section, we turn our attention to the two research questions posed above. The significance of the paper is also outlined in that section. 2. Characteristics of the SVCs 2.1. The monoclausality of the SVCs Ewe SVCs are as noted elsewhere monoclausal, and this is a key feature that differentiates them from Ewe bi-/multi- clausal constructions such as coordinate clauses, and sentences showing subordination. An indicator of the monoclausality of an SVC is the fact that the sequence of VPs in it ‘‘act together as a single predicate, without any marker of coordination, subordination, or syntactic dependency of any kind’’ (Aikhenvald, 2006:1). For example, in (2a) the verbs fɔ ‘pick, gather’ and ma ‘distribute’ are not overtly linked by any conjunction and these verbs, which are transitive, share syntactic subject and object that are expressed with only fɔ, the V1. As soon as a conjunction is introduced to link the verbs, a biclausal construction results because the subject and object of V2 are now overtly expressed with the V2. Compare (2a) with (2b) which illustrates coordination and with (2c), an example of subordination:(2a) 5 Stu markin Englis widely 6 Se 7 Ge among[Ama dies of C g in Ewe h vs. Aka -cited co e ackno ngbe, al groupsfɔ1S invo -Englis n-Eng mparat wledge so know of noné-ƒe lving English in h, Akan-Englis lish. Haust (199 ive study of bil ment at the en n as Gen, or M -native speakeavɔ-wo clude Amuz h, Ga-Englis 5) also com ingual noun d of the pap ina, is a dia rs of Ewe.ma2] u (2005a) on h and Hausa pares Mandin phrases in F er. lect of Ewe (K non-ve -Engli ka-En ongbe ozelka(SVC) Ama pick 3sg-POSS cloth-PL distribute ‘Ama gathered all her cloths and distributed them.’rbal predication in Ewe-English vs. Akan-English; Amuzu (2009) on plural sh; and Amuzu (2005b [2010]) on the mixed copula constructions in Ewe- glish with Wolof- English in the Gambia. Poplack and Meecham (1995) is a -French and Wolof-French. , 1984; Essizewa, 2007) that evolved initially as a second language variety 22 E.K. Amuzu / Lingua 137 (2013) 19--37(2b) 8 é i 9 En[Ama s the var glish andfɔ iant of Frencé-ƒe the 3sg (i.e. in h verbs that ocavɔ-wo] stead of the cur in bilingEYE wo in 2 ual SVC[wo b) tha s arema t is used in s consistently wó] ubordin their in(Coordination) Ama pick 3sg-POSS cloth-PL and 3sg distribute 3PL ‘Ama gathered all her cloths and she distributed them.’(2c) [Ama fɔ é-ƒe avɔ-wo] EYATA [é- ma wó]8ate c finitive(Subordination) Ama pick 3sg-POSS cloth-PL therefore 3sg distribute 3PL ‘Ama gathered all her cloths so she distributed them.’Note that in (2a) Ama, the subject of both verbs, is expressed only with V1and that the shared object of the two transitive verbs (i.e. é-ƒe avɔ-wo ‘her clothes’) is also expressed only after V1. Note, however, that in (2b), where the coordinative conjunction eye ‘and’ is introduced to link the two VPs, the V2 carries its subject as well as its object, both of which are realized as pronouns that refer respectively to the subject and object of V1. The same pattern is observable in (2c) where the subordinative conjunction eyata ‘therefore’ is introduced to link the VPs. Similar distinctions may be made in Ewe-English and Ewe-French CS. For example, compare the Ewe-English SVC in (3a) with the biclausal versions of it 9 in (3b) and (3c) :SVC: (3a) Míe download1 journal article ma-wo katã xle2 nyitsɔlauses. forms.1PL that-PL all read a.day.removed ‘We downloaded all those journal articles and read [them] a few days ago.’COORDINATION: (3b) Míe download1 journal article ma-wo katã EYE míe xle2 wo Tnyitsɔhis assertion is clar1PL that-PL all and 1PL read 3PL a.day.removed ‘We downloaded all those journal articles and we read them a few days ago.’SUBORDINATION: (3c) Míe download1 journal article ma-wo kata ̃ EYATA míe xle2 wó nyitsɔ1PL that-PL all so 1PL read 3PL a.day.removed ‘We downloaded all those journal articles so we read them a few days ago.’Because of the presence of a conjunction in both (3b) and (3c), the V2 in each case is accompanied by its subject and object arguments, which are pronominal copies of the arguments of V1. This is also the difference between the Ewe- French SVC in (4a) on the one hand and the biclausal versions of it in (4b) and (4c):(4a) [Kofi réparer1 afokpa-a do2] (SVC) Kofi repair shoe-DEF wear ‘Kofi mended and wore the [pair of] shoes.’(Ewe-French)(4b) [Kofi réparer1 afokpa-a] EYE [wo do2 e] (Coordination) Kofi repair shoe-DEF and 3sg wear 3sg ‘Kofi mended the [pair of] shoes and them.’(4c) [Kofi réparer1 afokpa-a] EYATA [é do2 e] (Subordination) Kofi repair shoe-DEF therefore 3sg wear 3sg ‘Kofi mended the [pair of] shoes so he wore them.’2.2. The SVC as an expression of one event Whether an SVC is made up of just two verbs as in (5b) or of more than two verbs as in (5a), it expresses only one event.(5a) eyata [é-dzo1 he-va-dze2 anyigbá he-tsɔ3 atukpa i la fied in szi4ection 2ɖi].4.therefore 3sg-move.above.ground ITIVE-VENTIVE-contact ground ITIVE-take bottle DEF press down E.K. Amuzu / Lingua 137 (2013) 19--37 23(5b) eye [wo-wó1 he-kaka2]. And 3sg-explode ITIVE-scatter ‘Therefore he jumped onto the ground and smashed the bottle on the ground and it exploded and scattered [before he was relieved]’(Ameka, 2006:129)However, the event which an SVC expresses consists of subcomponents or subevents that are represented by the individual VPs or groups of VPs. As Ameka notes, in (5a) ‘‘the last two VPs. . . represent a subevent unlike the other two, which represent a subevent each’’ (p. 129). The event that an SVC describes may be viewed as taking place at one scene or at multiple scenes. Thus, a distinction is made between ‘single-scene’ and ‘multi-scene’ SVCs (Pawley and Lane, 1998). All the SVCs in (5a) and (5b) illustrate the single-scene type: the subevents are closely related in time and space and do complement each other as they describe a single event. With ‘multi-scene’ SVCs, a series of events describe what may be conceived as a macro (i.e. one complex) event or episode. The events in question may take place at different times and locations as long as they are construed as being segments of an episode. Example (6) contains a multi-scene SVC which has a series of eight verbs. The episode comprises six events, two of which are expressed in a combination of VPs (namely, V1 & V2, and V7 & V8). The remaining events are expressed in single separate VPs.(6) Yao dzo1 yi2 agble vâ-ho3 agbeli me4 ɖu5 he no6 tsi ɖe é-dzi mlɔ7 Yao leave go farm VENTIVE-uproot cassava roast eat ITIVE- drink water at 3sg-TOP lie anyi dɔ8 alɔ ɖe ati te ground enter slumber at tree under ‘Yao went to the farm and uprooted cassava, roasted and ate it and drank water on it and then lay down and slept under a tree.’Both the single-scene and the multi-scene SVCs are found in the Ewe-English and Ewe-French CS data. Example (7a) from Ewe-English CS and (7b) from Ewe-French CS are of the single-scene type:(7a) Ke ne é-nye be yé hã lɔ wo-a, [ke eya boŋ a-go1 ahead a-propose2 a- Then if 3sg-be COMP 3sg too love 2sg-TP then 3sg rather POT-go ahead POT-propose POT- dzi3 ahiã wò]. woo 2sg ‘Then if he also loves you, [then it is he who should go ahead and propose to you and woo you].’ (Ewe-English)(7b) Battérie be si couler1 ɖe mɔ-a me ɖu2 résistant-wo gble3 circuit --a kpata. Battery poss water flow to machine-DEF INSIDE eat resistor-PL spoil circuit-DEF all ‘Acid from the battery flowed into the machine and corroded the resistors thus spoiling the entire circuit.’ (Ewe-French)In (7a), three (phrasal) verbs---go ahead, propose and dzi ahiã ‘woo, court’---describe various aspects of one act, i.e. the act of courting or wooing a lover. In (7b), there are three subevents that can be construed as one integrated event: the discharge of battery acid into a machine, the corrosion of resistors in the machine by the acid, and the damaging of the entire circuit in the machine. The multi-scene SVC is illustrated from Ewe-English CS in (8a):(8a) Mie four-wheel drive-a refurbish1-ge a-dzra2. 1PL four wheel drive-DEF refurbish-PROSP POT-sell ‘We will refurbish the four wheel drive and sell it.’ (Ewe-English)This SVC illustrates the multi-scene type of SVC because the events expressed by V1 (refurbish) and V2 (dzra ‘sell’) are independent although logically connected. They may indeed be set apart from each other and seen as having taken place at different times and locations, as demonstrated in the (8b) version of the above SVC:(8b) Mie four-wheel drive-a refurbish1-ge le Kumasi Magazine a-dzra2 le Accra. 1PL four wheel drive-DEF refurbish-PROSP at K. POT-sell at Accra ‘We will refurbish the four wheel drive in Kumasi and sell it in Accra.’ 24 E.K. Amuzu / Lingua 137 (2013) 19--37Similarly, note that in (9), from Ewe-French CS, Korkor’s business activities that have culminated in her becoming a hotelier occurred at different times and locations:(9) Korkor ple1-na ha so nord va sa2-na na client-wo le Asiga ̃me Korkor buy-HAB pig from north VENTIVE-sell-HAB to client-PL at Asiga ̃me he sɔ3-na bénéfice-a construire4-na de Bè videvide kaka va ITIVE take-HAB benefit-DEF construct-HAB at Bè piece.meal until VENTIVE zu5 hotélière egbe. become hotelier today. ‘Korkor used to buy pigs from the north, sell them to traders at the Lome market and with the profit built one structure after another and have therefore become a hotelier today.’ (Ewe-French)Ewe SVCs may be classified into functional types depending on the sequence of the subevents and on the semantics of the verbs which express the subevents. Several SVCs simply emphasize the chronological ordering of subevents that make up an event. The governing principle is that the ordering is as natural or as logical as possible. For example, in (6), the three verbs, ho ‘uproot’, me ‘roast’, and ɖu ‘eat’, occur in what is a natural sequence leading to the eating of the cassava. Later in the same example, the subevents expressed---mlɔ (anyi) ‘to lie (down)’ and dɔ alɔ ‘enter sleep’---occur in a natural sequence. Also, as captured in (10: Ewe-English), it is natural for mangoes to drop to the ground before they scatter:(10) [Mango drop1 --na kaka2-na ɖe ati-a te], gake ame aɖeke me show -na interest Mango drop-HAB scatter-HAB at tree-DEF under but person none NEG show-HAB interest ɖe wo me o in 3PL inside NEG ‘[Mangoes drop and scatter under the tree] but no one shows interest in them’ (Ewe-English)Another common functional type of Ewe SVCs is what may be called ‘manipulative SVCs’ (Durie, 1997; Dorvlo, 2008). The V1 in this SVC type is typically a HANDLING verb (either tsɔ ‘take’ or kɔ ‘take, raise’) that assigns to its object the thematic role of either instrumental or theme. The V2may be one of a wide range of verbs, including a change of state verb (e.g. lã ‘cut’ in 11a), a positioning verb (e.g. ku ‘hang’ in 11b), a change-of-location verb (e.g. kɔ ‘pour’ in (11c), or a benefactive verb (e.g. na ‘give’ in 11d):(11a) é-tsɔ1 hε lã2 te-a-e 3sg-take knife cut yam-DEF-FOC ‘S/he used a knife to cut the yam’(11b) Kofi kɔ1 agblɔʋu kpekpe ma ku2 ɖe gli ŋu K. take drum heavy DEM hang at wall skin ‘Kofi lifted that heavy drum and hanged it on the wall.’(11c) Osɔfo-a tsɔ1 ami kɔkoe la kɔ2 ɖe atukpa me Priest-DEF take oil holy DEF pour at bottle INSIDE ‘The priest poured the holy oil into the bottle.’(11d) Me tsɔ1 nye agbe na2 Yesu 1sg take 1sg life give Jesus ‘I give my life to Christ.’Two bilingual manipulative SVCs appear in (12), i.e. (12a) from Ewe-English CS and (12b) from Ewe-French CS.(12a) Azumah use1 é-ƒe experience tsɔ knock2-é out. Azumah use 3sg-poss experience TAKE knock-3sg out ‘Azumah used his experience to knock him out.’(Ewe-English; Asilevi, 1990:34) (12b) Mi sɔ1-na mɔ ŋua repérer2-na problem ɖekpekpe le hu ŋu1PL TAKE-HAB machine that diagnose-HAB problem any at vehicle side ‘We use that machine to diagnose any problem with a vehicle.’ (Ewe-French)Also common is the functional type of SVCs that may be called ‘directional SVCs’. The V1 in this type encodes movement while V2 typically encodes direction, as in the following: E.K. Amuzu / Lingua 137 (2013) 19--37 25(13a) nyɔnu-a tsi1 srɔ-a yome yi2 kɔƒe-a me woman-DEF stay spouse-DEF behind go village-DEF INSIDE ‘the woman trailed the husband (and went) to the village.’(13b) avu-vi-wo gbɔ1 va2 aƒeme Dog-small-PL return come house ‘the puppies have returned home.’This sequence is illustrated in (13c: Ewe-English), where walk is followed by yi ‘go’.(13c) Éya ŋutɔ-e walk1-na yi2-na va fetch3-na vi-a-wó gbesiagbe 3sg self-FOC walk-HAB go-HAB VENTIVE fetch-HAB child-DEF-PL everyday ‘It is she herself who walks to fetch her children every day.’(Ewe-English, Amuzu, 1998:55)We shall return to this type of bilingual SVC in section 3 as we consider a critical issue, namely the inability of some English and French verbs of direction to occur in it. In ‘resultative SVCs’, the action expressed in V1 leads to the situation expressed in V2. In (2b), the explosion (expressed in V1, wo ‘explode’) of the subject entity is what has caused this entity to scatter, as expressed in V2 kaka ‘scatter’. In (7b: Ewe-French), we are told that the battery acid that flowed into the machine led to the corrosion of some resistors in the machine. Lastly, some SVCs are used to express comparison. V1 expresses the quality of the subject entity while the entity to which the subject is compared surfaces as the object argument of V2, which may be the comparative verb wú ‘exceed, surpass’ as in (14a) or the motion verb gbɔ ‘come.back’ in (14b):(14a) Kodzo lolo1 wu2 fo-a K. fat exceed older.brother-DEF ‘Kodzo is fatter than his older brother.’(14b) é-tsi1 gbɔ2 nɔvia ŋu 3sg-grow come.back sibling skin ‘he has grown more than his brother/sister.’(Ameka, 2006:136)An instance of this SVC type is the Ewe-French CS example below:(14c) amedzro-wo lε développer2 wu2 mi sasasa Visitor-PL PROG develop exceed 1PL by.far ‘‘immigrants are developing at a much faster pace than us.’’ (Ewe-French)2.3. Verb transitivity and the expression of objects in SVCs Verbs in Ewe SVCs may have the same or different transitivity values. For example, in (1a) the verbs have the same transitivity value (both lé ‘catch’ and tu ‘lock’ are transitive) while in (5b) the verbs, wo ‘explode’ and kaka ‘scatter’, are intransitive. On the other hand, in (5a) transitive and intransitive verbs are mixed: the intransitive verb dzo ‘jump’ participates in SVC with three transitive verbs, dze ‘contact’, tsɔ ‘take’ and zi ‘press’. We find all these verb combinations in Ewe-English and Ewe-French CS too. In (10: Ewe-English) and in (15: Ewe-French), the verbs in sequence are intransitive; in (12a) and (12b) they are transitive; and in (16) they are a mixture of transitive and intransitive verbs:(15) Wó refuser1 dzo2 enumake kple dziku. 3PL refuse leave immediately with anger ‘They refused and left immediately with anger.’ (Ewe-French)(16) Guy-ma a-teŋu a-dance1 a-entertain2 ame-wo well well well. Guy-that POT-can POT-dance POT-entertain folk-PL ‘That guy can dance and entertain people very well.’ (Ewe-English) 26 E.K. Amuzu / Lingua 137 (2013) 19--37When transitive verbs in sequence do not share an object, each verb is followed by its own object; we saw this in the monolingual Ewe SVC in (6) where V6, V7, and V8 are followed by their respective objects. However, when transitive verbs in sequence do share object, it is expressed only once with the first verb as is the case with agbeli ‘cassava’ after V3 in (5) although agbeli is also the object of V4 and V5. As elsewhere, what obtains in monolingual SVCs applies also in bilingual SVCs. In (3a) (Ewe-English), article mawo katã ‘all those articles’, the shared object of V1download and V2 xle ‘read’, occurs only after V1. However, in (12a), because the two verbs, tsɔ ‘take’ and knock out, do not share object, each of them is followed by its own object. In Ewe-French CS too, as we see in (17a), a shared object occurs once, i.e. in-between its two transitive verbs. On the other hand, note in (17b) that sɔ ‘take’ and repérer ‘find, locate’ are followed by their respective object because they do not share object:(17a) 10 Ther two asp 11 Verb consiste indicate English’Kofi e is an ectual s that nt with s that t ’ (p. 13réparer1 exception th categories, t are unmarked what Ameka he change of 8).afokpa ough. The he progres are said (2006:13 state it [thdo2re is an a sive -m to be in 8) says i e verb] euxiliar , and the aor s its fu ncodeKofi repair shoe wear ‘Kofi mended the [pair of] shoes and wore them.’ (Ewe-French)y pre the pr ist. T nction s has(17b) Mí sɔ1-na mɔ ŋua repérer2-na -verb le ‘b ospective he interpr in Ewe. occurredproblem e.at’ that ind ge . Le has etation of ao Thus, if an E before now ɖekpekpe icates tense (o two forms: le rist in relation t nglish/French and the post sle r dir (pr o En verb tate hu ection) on b esent tense glish and F in the aori is current; ŋu 1PL TAKE-HAB machine that find-HAB problem any at vehicle side ‘We use that machine to diagnose any problem with a vehicle.’ (Ewe-French)2.4. Expressing aspect and modality categories in the SVCs It is important to note right away that the category of tense---which in languages like English and French locates the time of an event expressed by a verb with reference to the moment of speaking or writing---is NOT expressed with Ewe verbs (Ameka, 2006; Ameka and Kropp Dakubu, 2008; Aboh and Essegbey, 2010)10 and accordingly not with English and French verbs in bilingual SVCs. It is aspect and modality categories that are expressed. Aspect ‘‘views the event referred to by the verb with respect to inception, completion, duration, etc., and modality. . . states the event’s relationship to truth, possibility, necessity, etc’’ (Johansson and Lysvåg, 1987:116). Modality categories are expressed via pre-verb markers in Ewe. They include the potential mood marker (l)a, the subjunctive (n)a, the auxiliary te ŋu ‘can’ among others. Aspectual categories include the repetitive ga, the aorist,11 and the habitual --(n)a. Two other aspectual categories, the progressive and the prospective, are expressed differently in varieties of Ewe that are in contact with English (e.g. Anlo) and Gengbe or Mina, the variety that is used in the Ewe-French CS recordings in Lome, Togo. The progressive is expressed as the post- verb --m in Ewe-English CS but as the pre-verb lε in Ewe-French CS. The prospective is expressed as ge, a post-verb, in Ewe-English CS but as la, a pre-verb, in Ewe-French CS. Verbs in Ewe SVCs can be formally marked as similar or different depending on the aspect and the modality categories. The verbs in many of the monolingual Ewe SVCs discussed above are illustrative of SVCs in which verbs are formally marked similarly for aspect or for modality. For example, in (2a), (5a and 5b), and (6), which are monolingual Ewe examples, all the verbs are in the aorist, a perfective aspect (cf. Aboh and Essegbey, 2010:42). The verbs in (1) are also similarly marked for habitual aspect. However, the verbs in (18) are marked differently for aspect and/or modality. In (18a), V1 is in the aorist while V2 carries the progressive aspect marker --m. In (18b), V1 carries a-, ‘potential mood marker’ (Essegbey, 1999, 2008) while V2 is in the aorist. In (18c), V1 carries the prospective aspect marker ge, while V2 carries the potential a-.(18a) Kofi ƒle2 ʋu yeye aɖe le kuku2-m fifia Name buy vehicle new INDEF be.at:PRES REDU-drive-PROG now ‘Kofi bought a new car, which he now drives’(18b) Ama â-ɖa1 nu ɖu2 Name POT-cook thing eat ‘Ama will cook and eat.’ (Ameka, 2006:138)(18c) Ama le nú ɖa1 -ge a-ɖu2 Name be.at:PRES thing cook PROSP POT-eat ‘Ama will cook and eat.’ (Ameka, 2006:138)ehalf of verbs which carry either of ) and nɔ (non-present). rench verbs in bilingual clauses is st is an inchoative verb, ‘‘the aorist hence, it is translated as present in E.K. Amuzu / Lingua 137 (2013) 19--37 27Similar patterns are reflected in the bilingual data. For example, in (10 and 13a, from Ewe-English CS), both verbs carry the habitual just as the verbs in (9 and 12b, from Ewe-French CS) do. However, in the Ewe-English CS examples in (8a) and (19), verbs are marked differently---the V1 carries the prospective aspect marker while the V2 carries the potential mood marker:(19) Dr. Annan be [ye meeting yɔ1-ge a-discuss2 nya ma-wo kata ̃ kple mi] Dr. A. say LOGO meeting call-PROSP POT-discuss matter that-PL all with 1PL ‘Dr. Annan said [he will call a meeting to discuss all those issues with us].’ (Ewe-English)Similarly, in the Ewe-French CS example below, while the V1 takes the progressive lε the V2 takes the potential a-:(20) Ettiene lε préparer1 géographie a-fiɔ2 ebe etudient-wo Lundi Ettiene PROG prepare geography POT-teach his student-PL Monday ‘Ettiene is preparing geography to teach his students on Monday.’ (Ewe-French)As was noted above, the forms used to express the progressive and the prospective in dialects of Ewe (e.g. Anlo) used in Ewe-English CS are different from the forms used to express those categories in the Gengbe dialect used in Ewe- French CS. While the post-verbs --m and -ge are used to express the progressive and the prospective respectively in Anlo, the pre-verb lε and la are used in Gengbe to express the respective categories. This morphological distinction is the basis for a major syntactic distinction (i.e. with respect to SVCs) between dialects like Anlo on the one hand and the Gengbe dialect on the other. In Gengbe, when two transitive verbs in SVC share object, the object appears only with the first verb and the result is an S-V-O-V clause structure. This structure is used irrespective of the modality or aspectual categories the transitive verbs carry. It is thus this structure we find in the examples in (21): in (21a) the verbs take the habitual -na, in (21b) they take prospective la, and in (21c) they take progressive lε:(21a) Ama ɖa1-na nu ɖu2-na (S-V-O-V) Name cook-HAB thing eat-HAB ‘Ama cooks and eats.’(21b) Ama la ɖa1 nu a-ɖu2 (S-V-O-V) Name PROSP cook thing POT-eat ‘Ama will cook and eat.’(21c) Ama lε ɖa1 nu a-ɖu2 (S-V-O-V) Name PROG cook thing POT-eat ‘Ama is cooking to eat.’In dialects like Anlo, however, the S-V-O-V structure is used only when the first of the object-sharing verbs takes an aspectual form other than the progressive m or the prospective ge. Thus, (21a) may as well come from Anlo. As soon as the first of the object-sharing verbs carries either m or ge, the shared object has to precede (rather than follow) the first verb, resulting as it were in an S-O-V-V clause structure as we find in the following versions of (21b) and (21c):(22a) Ama nu ɖa1 -ge a-ɖu2 (S-O-V-V) Name thing cook-PROSP POT-eat ‘Ama will cook to eat.’ (compare with 21b)(22b) Ama nu ɖa1 -m a-ɖu2 (S-O-V-V) Name thing cook-PROG POT-eat ‘Ama is cooking to eat.’ (compare with 21c)Ewe-English CS and Ewe-French CS have ‘inherited’ these dialectal differences: while SVCs in Ewe-English CS mirror the patterns found in dialects like Anlo, the SVCs in Ewe-French CS mirror the patterns found in Gengbe. In the Ewe- English CS data the S-V-O-V vs. S-O-V-V distinction is kept: compare (23a) in which V1 carries the potential a with (23b) in which V1 carries the progressive m(23a) Wo dzi be yewo a- manufacture1 chocolate a-export2 (S-V-O-V) 3PL want that LOGO POT-manufacture chocolate POT-export ‘They want to manufacture chocolate and export it.’ 28 E.K. Amuzu / Lingua 137 (2013) 19--37(23b) 12 Actu V1. Its s 13 Caro held in OAccountant-a ally, it is an S-V-O-V-O econd object is ebe e l Myers-Scotton discu slo. The article versinɔ stru tudie ssed on ofga-a cture that is used nt-wo ‘his stude this issue gener that presentatiocash1--m in (24b). Fiɔ, the nts’. ally in a paper sh n will soon appenɔ V2, i e pre ar.ɖu-ɖu2-m. . . s a double object verb th sented in June 2011 at (S-O-V-V) Accountant-DEF be money-DEF cash-PROG be REDU-eat-PROGat shares its first object, géographie, with the the International Symposium on Bilingualism‘The accountant was cashing and spending the money. . .’But in Ewe-French CS, only the S-V-O-V structure is used following the situation in monolingual Gengbe: compare (24a) in which V1 is in the aorist with (24b) in which V1 takes the progressive 12 lε.(24a) Kofi réparer1 afokpa do2 (S-V-O-V) Kofi repair shoe wear ‘Kofi mended the [pair of] shoes and wore them.’(24b) Ettiene lε préparer1 géographie a-fiɔ2 ebe etudient-wo Lundi (S-V-O-V) Ettiene PROG prepare geography POT-teach his student-PL Monday ‘Ettiene is studying geography to teach his students on Monday.’It is important to draw attention to the fact that in the bilingual clauses Ewe modality and aspectual forms occur with English and French infinitive verbs and that this fact supports Myers-Scotton’s recent stipulations that in intrasentential CS it is EL infinitive verbs (not their finite counterparts) which take ML finite inflections.13 Strong evidence for this assertion in regard to Ewe-English CS are examples in which English infinitive verbs whose finite forms are irregular verbs, as with go in (25a) where it takes the potential; neither of the finite forms of go (went, gone) can replace it in (25a):(25a) Ke ne é-nye be yé hã lɔ ̃ wo-a, [ke eya boŋ a-go1 ahead a-propose2 a- Then if 3sg-be COMP 3sg too love 2sg-TP then 3sg rather POT-go ahead POT-propose POT- dzi3 ahia ̃ wò]. woo 2sg ‘Then if he also loves you, [then it is he who should go ahead and propose to you and woo you].’Another kind of overt evidence is English verbs in bilingual SVCs which have past tense reading. In (12a), for example, both use and knock occur without any overt marking (= are in the aorist per Ewe morphology) and are interpreted as encoding past actions; the past tense forms of these verbs are unacceptable as shown in the version of (12a) even when we imagine they are in the aorist:(25b) Azumah *used1 é-ƒe experience tsɔ *knocked2-é out. Azumah used 3sg-poss experience TAKE knocked-3sg out ‘Azumah used his experience to knock him out.’ (Ewe-English; compare with 12a)Similarly, in Ewe-French CS it is infinitive French verbs that take the Ewe inflections. However, because the pronunciation of French infinitive verbs that end with --er could be the same as for a number of French finite verb forms and because several French verbs showing Ewe morphology that are discussed in this paper are from the -er category, it is regularly not obvious that they are infinitive forms. For example, the infinitive verb réparer ‘to repair, to mend’ in (24a), which is in the aorist, has the same pronunciation as its past tense form, réparé ‘repaired, mended’. Much clearer examples of our claim are verbs with ending other than -er (e.g. -re, -ir, -oir) since they are pronounced differently from their finite counterparts. The --re verb construire ‘to construct’ (27b) is one such verb: its past tense form, construit, is unacceptable in the following version of that example (construire would instead have occurred in the aorist in this example):(25c) Wo mu utiliser1 fer sɔ2 *construit3 gli o] 3PL NEG utiliser iron_rod take construct wall NEG ‘They didn’t use iron rods to construct (the) walls.’, E.K. Amuzu / Lingua 137 (2013) 19--37 292.5. Expressing negation in SVCs Negation in the Ewe SVC is marked once by the bipartite morpheme me. . . o. Me comes before V1 and o comes at sentence final position, as in (26):(26) 14 Calli 2005a,b constitu is a ‘comɖevi-a ng Ewe a one , 2009, 2010, ents, it does so posite ML’.me-ta1-language ‘ML’ in press), it will in conjunction yi2in Ew be de with inxɔ-a e-English CS monstrated th puts from Engme is not entirely accurate at although Ewe plays lish/French. In other woo Child-DEF NEG-crawl go room-DEF containing.region NEG. ‘the child didn’t crawl into the room.’ . As wa a predo rds, it w(Ameka, 2006:138)As elsewhere in the grammar of bilingual SVCs, this morphosyntactic characteristic of the Ewe SVC finds expression in bilingual SVCs too. In examples in (27a: Ewe-English) and (27b: Ewe-French), one finds that mu, the Gengbe variant, duly occurs before V1 while o occurs at the sentence final position:(27a) Nye me venture1 do2 coat le heat ya me kpɔ o. 1sg NEG venture wear coat at heat this inside see NEG ‘I have never ventured to wear a coat in this heat.’(27b) [Wo mu utiliser1-na fer sɔ2 construire3-na gli o], s do mina ill beta ne in nt ro comewo previo le in s clear be us stud etting th that whsamu-wo ies by this author (cf. A e grammatical frame of at obtains in the bilingualmu 3PL NEG utiliser-HAB iron_rod take construct-HAB wall NEG; so 3PL poss storey_building-PL NEG do-na nuse o pile-HAB strength NEG ‘[They don’t use iron rods to construct (the) walls] so their storey buildings are not strong.’3. Principles which guide the combination of Ewe and English/French morphemes in the bilingual SVCs Evidence discussed in the preceding section points to a firm conclusion that structural possibilities in bilingual SVCs in both Ewe-English and Ewe-French CS relate directly to structural possibilities in monolingual Ewe SVCs. This section aims at explaining why there is this kind of relationship, and it pursues this aim using Carol Myers-Scotton’s (e.g. Myers- Scotton, 1993, 2002) Matrix Language Frame/MLF model. It is claimed that the relationship stems from the application of the central principles of the MLF model---namely the Morpheme Order Principle and the System Morpheme Principle---by both groups of bilinguals. A third principle, the Blocking Filter (Myers-Scotton, 1995:247) is also explored in the explanation. We begin with the Morpheme Order Principle (MOP). It says:In ML + EL constituents consisting of singly-occurring EL lexemes and any number of ML morphemes, surface morpheme order (reflecting surface relations) will be that of the ML. (Myers-Scotton, 1993:82)The default ML in Ewe-English and Ewe-French CS is Ewe,14 so the MOP anticipates that verbs participating in bilingual SVCs will share one subject in accordance with the system in monolingual Ewe SVCs. The MOP also anticipates the pattern of distribution of object arguments in bilingual SVCs. It anticipates that similarities and differences among dialects of Ewe with respect to the distribution of objects in SVCs will show up in bilingual SVCs involving the dialects. We encountered support for this prediction in bilingual SVCs in Ewe(=Anlo)-English CS and their versions in Ewe(=Gengbe)- French CS in which two transitive verbs share one object. In Ewe(=Anlo)-English CS, following the patterns in Anlo, either the basic S-V-O-V structure or the re-structured S-O-V-V structure is used depending on whether the first of transitive verbs takes the progressive or the prospective (see the discussion of (23)). In contrast, in Ewe(=Gengbe)-French CS, following the patterns in Gengbe, only the S-V-O-V structure is used irrespective of whether the first of the transitive verbs takes the progressive or the prospective (see the discussion of (24)). The System Morpheme Principle (SMP) also anticipates that verbal morphology in the bilingual SVCs will be consistent with patterns found in the monolingual Ewe dialects involved. The principle says:In ML + EL constituents, all system morphemes which have grammatical relations external to their head constituent (i.e., which participate in the sentence’s thematic role grid [namely outsider system morphemes]) will come from the ML. (Myers-Scotton, 1993:82)muzu, mixed SVCs 30 E.K. Amuzu / Lingua 137 (2013) 19--37‘Outsider system morphemes’ include such grammatically-active morphemes as aspect and modality markers, negation markers, and case-marking prepositions; and the principle anticipates that in bilingual SVCs all such morphemes will come from the Ewe dialect involved. The principle thus explains why English/French verbs routinely occur with Ewe aspect, modality and negation markers and not with English/French-origin aspect, modality and negation markers (see sections 2.4 and 2.5). It is instructive to note that a distinction is made between outsider system morphemes and ‘early system morphemes’ (Myers-Scotton, 2002). Early system morphemes include verb satellites, e.g. out in knock out (12a, reproduced below as 28), and forms like the reflexive pronoun se in French pronominal verbs (e.g. se in se soumettre ‘to submit oneself to, to subject oneself to’ in example 29). Such morphemes contribute to the lexical meaning of the verbs they accompany and they often get lexicalized 15 with their verbs. The absence of such forms normally alters the meaning of the verb; e.g. while knock out---in boxing---means ‘to defeat by flooring an opponent with punches’, knock simply means ‘to punch, bang, tap’. Similarly, without se, soumettre’s meaning changes to ‘to subdue, to subjugate (someone)’. The important point here is that the SMP does not block early system morphemes from occurring with their head verbs in bilingual SVCs. This is why out occurs with V2 knock in the bilingual SVC in (28) and se occurs in (29) with V2 soumettre:‘(28) 15 In oth 16 Six ou also unce prepositio become cAzumah er words, su t of ten respo rtain that the n in CS is do lear during tuse1ch form ndents y can p ne thro he discué-ƒe s do not conv categorically roduce such ugh similar co ssion of motexperience ey grammatical ruled out the gra a bilingual cons nsultations. Tha ion events belowtsɔ relations mmatical truction. t forms lik . knock2-é information ab ity of (32b) whi The determina e into and throut. Azumah use 3sg-poss experience TAKE knock-3sg out ‘Azumah used his experience to knock him out.’ out th le the r tion o ough (Asilevi, 1990:34)(29) mí gba lε dzi1 la se soumettre2 na mía be e ve ema f the as uparents-wo rbs in ways outs ining four who ar acceptability or sed here are cono 1PL REP PROG want PROSP submit to 1PL poss parents-PL NEG ‘we no longer want to submit to our parents.’Note however that because aspect markers and case-marking prepositions are examples of outsider system morphemes, they are expressed using Ewe morphemes: in (29) la (the prospective) occurs with se soumettre while the dative preposition na introduces the Recipient object argument mía be parents-wo ‘our parents’. The dative na would have been unacceptable in (29) if soumettre (without se) were the verb because in that case it would assign to the object the thematic role of Patient rather than Recipient:(30) mí gba lε dzi1 la soumettre2 (*na) mía be parents-wo o 1PL REP PROG want PROSP subjugate to 1PL poss parents-PL NEG ‘we no longer want to subjugate our parents.’Two related points need to be stressed with respect to the occurrence of the dative case-marker na in (29). One is that it is present there because of the abstract lexical structure (subcategorization features) of se soumettre. The second is that it (rather than its French counterpart) has to occur as the dative marker because, as noted, the SMP empowers only Ewe to contribute such outsider system morphemes in bilingual VPs. Thus, na in (29) may not be replaced by its French counterpart à:(31) mí gba lε dzi1 la se soumettre2 *à mía be parents-wo o 1PL REP PROG want PROSP submit to 1PL poss parents-PL NEG ‘we no longer want to submit to our parents.’Although the SMP is what blocks English and French case-marking prepositions from occurring in bilingual constituents and allows their Ewe case-marking preposition counterparts to occur, it is not what blocks another category of English and French prepositions, which are content morphemes. Examples of content morpheme prepositions in English are into, 16 which is blocked in (32b), and through, which is blocked in (33b).(32a) Fiafi-a dive1 ge2 dze3 tɔ-a-me. . . Thief-DEF dive drop contact river-DEF-inside ‘The thief dived into the river. . .’ (Amuzu, 1998:55)ider system morphemes do. e below twenty years of age, are otherwise of an English/French tent-morpheme prepositions will E.K. Amuzu / Lingua 137 (2013) 19--37 31(32b) 17 The 18 Actu (i.e. othe example ‘‘The bo to casesFiafi-a concept of fixi ally, Figural en r-directed mo s in Fig. 1. In t y threw the bal of spontaneodive ty is no tity (in m tion). In he caus l over th us moti*into t explore otion e spontan ed motio e wall’’ w on.tɔ-a-me. . .d in this paper becau vents) is conceptualiz eous motion, the figu n events, however, so here ball, the object NThief-DEF dive into river-DEF-inside(33a) é leak1-ge a-to2 do ma me se it is n ed in two ral entity me agen P, is thelo! 3sg-leak-PROSP POT-pass hole that inside EXCL ‘(mind you) it will leak through that hole.’(33b) é leak-ge *through do ma me lo! 3sg-leak-PROSP hole that inside EXCLExamples (32a) and (33a) are bilingual SVCs that express motion events. In (32a), three verbs together express a motion event: dive encodes a subject entity’s motion as well as the manner of its motion, ge ‘drop, fall’ encodes the change-of- location of the subject entity, and dze ‘contact’ encodes the end-state of the motion. Similarly, multiple verbs express a motion event in (33a): leak expresses motion and manner of motion while the Ewe verb to ‘pass by, pass through’- expresses path of motion (including boundary crossing). In order to understand why the English motion verbs have to be used as V1 with V2 Ewe path/direction verbs, and in order to understand why those English verbs may not occur with the English prepositions into and through, we need to refer to work done by Leonard Talmy (e.g. Talmy, 2000) on cross-linguistic differences among languages with regard to how they express motion events. Armed with insight about motion event expression, we will also be in position to appreciate the fact that the blocking of such English prepositions (and of their French counterparts in Ewe-French CS) is due not to the SMP but to another principle of the MLF model, which Myers-Scotton (1995:247) has called the ‘‘Blocking Filter’’. According to Talmy (2000:221), the conceptualization of a motion event may be analysed as integrating four distinct conceptual components, namely ‘‘a figural entity, an activating process, an association function, and a ground entity’’. The conceptualization is 17 captured in Fig. 1, which is an adaptation from Talmy (2000:221). Figural entity is generally the component on which attention or concern is currently centred. It is typically expressed in the subject NP of a motion expression, e.g. ‘the thief ’ in example (a) and ‘it’ in (b).18 The Activating process, which is typically expressed in the verb group (e.g. dived and will leak), is the conceptual part of a motion event by which the figural entity makes transition with respect to the Ground entity, which is the spatial environment in respect to which the figural entity is moving. The Association function relates the figural entity to the ground entity, as into and through respectively do in the examples in Fig. 1. In other words, the functions which into and through perform in these examples are conceptual in nature, and it is why from the standpoint of the MLF model they are content morphemes (and not late system-morpheme prepositions which merely indicate grammatical case). Talmy (2000) notes that although in expressing motion events verbs typically encode the concept of motion itself, it is often the case that the verbs additionally encode another motion related concept, e.g. the manner of motion or the path/ direction/boundary crossing/end-state of motion. One group of verbs (e.g. dive and leak) conflate motion and manner, leaving a content-morpheme preposition (satellite in Talmy’s terminology) to express the association function; they are henceforth called motion+manner verbs. For example, in (a) in Fig. 1 into expresses the direction and end-state of the motion expressed by dive and in (b) through expresses the path as well as the boundary crossing of the figural entity in the motion encoded in leak. The other group of verbs conflate motion and an association function (i.e. the path/direction/ boundary crossing/end-state of the motion). One such verb is enter (=go/come into, as used in Fig. 2) which conflates motion and boundary crossing. Unlike enter, which does not require any preposition to introduce or case-mark the ground NP, some verbs in this category do require such a preposition. An example is deviate (=move away), which conflates motion and direction. It requires the preposition from to introduce and thus case-mark the ground NP as Source entity, as shown in (b). With these verbs (henceforth motion+path verbs), the manner of motion is optionally expressed through adverbials. For instance, gingerly expresses manner of motion in ‘‘He entered the room gingerly’ just as deliberately does in ‘‘He deliberately deviated from the theory’’.ot relevant to the phenomenon of SVC. ways: entity in spontaneous (or self-directed) motion and entity in caused initiates its own movement resulting in a change of location, as with the tive entity causes the figural entity to move from one place to another, as in figural entity. Because of space limitations we shall restrict our illustrations 32 E.K. Amuzu / Lingua 137 (2013) 19--37 [Figu ral entity Activat ing proce ss Associatio n fu nction Gro und en tity] Transition fixity17 (a) The t hief div ed i nto the river. (b) It will leak through that hole. Fig. 1. Conceptual structure of a motion event. [Figural entity Activating proce ss Associati on function Ground e nt ity] Transition (a) He ent ered [go/c ome into] the room. (b) He deviated [move away] from the theory. Fig. 2. Conceptual structure of a motion event in which the verb conflates motion and path/direction/direction/boundary crossing/end-state of motion. [Figural entity Activating process Associati on function Ground entity] Transition sodza-a-wo z va k e-a me soldier-DEF-PL walk c ome village-DEF in ‘the soldie rs walked to the village.’ Fig. 3. Conceptual structure of a motion event expressed in an Ewe SVC.Talmy then observes that although languages have both motion+manner and motion+path verbs (e.g. English has both dive and enter), they differ with regard to which category they have predominantly. Some languages have more motion+manner verbs while others have more motion+path verbs. Languages which typically conflate motion and manner in verbs are, in his terminology, satellite-framed (or S-framed) languages; they are so-called to draw attention to the fact that with such verbs it is satellites which express path (i.e. perform the association function). He places English in this group. Languages which typically conflate motion and path are, in his terminology, verb-framed (or V-framed) languages, and attention is thereby drawn to the fact the motion verbs take care of the association function on their own. French is a member of this other group. Incidentally, Ewe belongs to neither group (Ameka and Essegbey, 2007; Adjei, forthcoming), a point which has direct bearing on our understanding of the nature of the bilingual directional SVCs in (32a) and (33a). Ameka and Essegbey (2007) find that Ewe belongs to a third group of languages, equipollently-framed languages,19 which typically express motion events through SVCs. In the SVCs, an initial verb conflates the concepts of motion and manner while a following verb conflates another concept of motion and, this time, path/direction, etc. The verbs would however be of equal syntactic status. This is demonstrated in Fig. 3 in which zɔ ‘walk’ conflates motion and manner while va ‘come’ conflates motion and direction. However, a motion+path verb may occur alone (i.e. not in an SVC with an initial motion+manner verb). In this distribution, there is no difference between the Ewe motion+path verb and the English/French counterpart. As in English (see Fig. 2), there are two sub-types of Ewe motion+path verbs. Members of one type require the ground NP to be case- marked by a preposition. Ge ‘drop, fall’, exemplified in Fig. 4, is a member of this sub-type. In the example, the preposition ɖe ‘into’ case-marks (as Goal) the ground NP kɔƒe-a me. Members of the second sub-type of Ewe motion+path verbs do not require the ground NP to be case-marked. An example is ɖo ‘reach, arrive at’, illustrated in Fig. 5; note that no preposition case-marks the ground NP kɔƒe-a me.19 It was Slobin (2004:249) who first pointed to this third group of languages. Equipollently-framed languages are those in which Manner and Path are expressed by forms that are accorded equal morphosyntactic status. E.K. Amuzu / Lingua 137 (2013) 19--37 33 [Figural entity Activating proce ss Associati on function Ground e nt ity] Transition sodza -a- wo ge ɖe k e-a me s oldier-DEF-PL dro p ALLAT villa ge-DEF i n ‘the s oldiers entered the villag e.’ Fig. 4. A motion+path/direction verb that requires a case-marking preposition to introduce the ground NP. [Figural entity Activating proce ss Associati on function Ground e nt ity] Transition Sod za-a-w o ɖó k e-a me child -DEF-PL reach/arri ve a t villa ge-DEF in ‘the soldiers reac hed the villag e.’ Fig. 5. A motion+path/direction verb that does not require a case-marking preposition to introduce the ground NP.Concerning the expression of motion events using CS, it is to be expected in line with predictions of the MLF model that Ewe morphosyntactic procedures would prevail, not English or French ones. First, let us return to the examples in (32) and (33) in which motion events are expressed. If English morphosyntactic procedures were to prevail, then it should be possible for the preposition into to express the association function of path which the motion+manner verb dive does not entail; it should also be possible for through to express the association function that leak does not entail. But as we find in (32b) and (33b), these prepositions are unacceptable as CS forms. We find instead that it is the Ewe morphosyntactic procedures which prevail: the motion+manner verbs occur as V1 in SVC with Ewe motion+path verbs which thus perform association function (see 32a and 33a respectively). The unacceptability of the prepositions into and through (which as pointed out are content morphemes because they encode association function, a core conceptual element in motion events) is anticipated in what Myers-Scotton has termed the ‘‘Blocking Filter’’. The filter20 Tha. . .blocks an EL content morpheme not congruent with its corresponding ML morpheme. (Myers-Scotton, 1995:247)Congruency is crucial in the MLF model and is defined as a match between the EL content morpheme and the ML content morpheme in two ways. First, if the ML counterpart of an EL content morpheme is a system morpheme, there is no match at all and a switch to the EL content morpheme in a mixed constituent is blocked. Second (and more relevant here), singly- occurring content morphemes from the EL can be inserted into constituents framed by the ML only if they show sufficient congruence with their ML content morpheme counterparts at the three levels of abstract lexical structure: lexical- conceptual structure, predicate-argument structure, and morphological realization patterns. The preposition into and through are only partially congruent with their Ewe content morpheme counterparts, which happen to be motion+path verbs. The prepositions and the motion+path verbs share lexical-conceptual structure (they all encode the core concept of association function); they also share predicate-argument structure (they all assign thematic role to the NP which encodes the Ground entity in motion event expressions); but they mismatch in morphological realization patterns (they are prepositions vs. verbs). What this really means is that Ewe does not have a syntactic 20 model that may be used to accommodate the English prepositions. The Blocking Filter thus ensures that they do not occur as CS forms. Instead, the ML calls for Ewe motion+path verbs to express the association function which the blocked prepositions express. A similar explanation goes for CS patterns involving French motion+manner verbs in Ewe-French CS. Let us consider (34a) where marcher ‘walk’ occurs with the Ewe deictic verb yi ‘go’. As expected, example (34b) is considered by respondents to be unacceptable, and the reason is that à ‘to’, the French prepositional equivalent of the verb yi, cannot replace yi.(34a) t wo is, a sla yntactic struté cture dmarcher1evoted to enca-yi2oding motiofi n eventsŋu in wha 2sg PROSP can walk POT-go place that Q ‘can you walk to that place?’ich a path-encoding preposition may perform an association function. 34 E.K. Amuzu / Lingua 137 (2013) 19--37(34b) wo la té marcher *à fi ŋu a 2sg PROSP can walk to place that QWe now turn to the distribution of motion+path verbs in predications that do not involve motion+manner verbs. We have noted with Figs. 4 and 5 that in this case Ewe motion+path verbs occur in the V-framed construction type and that there are two subtypes of this construction. There is one in which the motion+path verb requires a preposition to case- mark the ground NP and there is the other in which the verb does not make any such request. These two construction types mirror those in which English and French motion+path verbs occur. This means therefore that Ewe has the right syntactic models that may be used to accommodate English and French motion+path verbs. English motion+path verbs which do not require case-marking prepositions to introduce the ground NP behave like ɖo ‘reach’ in Fig. 5. This is the case with approach in (35: Ewe-English):(35) vi-nye-wo me te-a-ŋu approach-na wó fofo kple ga nya o child-1sg-PL NEG can-HAB approach-HAB 3PL father with money word NEG ‘my children cannot approach their father with issues related to money.’But English motion+path verbs which require a preposition to case-mark their ground NP occur in Ewe-based V-framed constructions in which Ewe case-marking prepositions introduce the ground NP, as we find in the occurrence of sink and ɖe ‘ALLATIVE’ in (36).(36) me dzi be message ya na-sink ɖe mia ƒe dzi-wo me 1sg want COMP message this POT-sink ALLAT 2PL poss heart-PL inside ‘I want this message to sink into your hearts.’Sink assigns case (=Goal) to the ground NP and requires a preposition to do the case-marking. Instructively, while it is the Ewe allative ɖe which actually expresses the required case, its English counterpart into cannot replace it, as we see in (37):(37) me dzi be message ya na-sink *into mia ƒe dzi-wo me 1sg want COMP message this POT-sink 2PL poss heart-PL insideBecause into is meant to function here as a case-marking preposition (=late outsider outsider system morpheme), it is blocked by the System Morpheme Principle/SMP, which prohibits the occurrence of such EL morphemes. As should be clear by now, the English case-marking prepositions that the SMP blocks in connection with the distribution of English motion+path verbs are replaced by equivalent Ewe case-marking prepositions while in contrast the English path-encoding content-morpheme prepositions that the Blocking Filter blocks in connection with English motion +manner verbs are replaced by Ewe motion+path verbs, which are content morphemes too. The distribution of French motion+path verbs in bilingual V-framed constructions is not different. The verb inonder ‘to flood’ (38) occurs without any overt case-marking of the ground NP because this is consistent with its distribution in French.(38) tsi inonder aƒe-wo le tokɔ-a me water flood house-PL at suburb-DEF inside ‘water flooded all the houses in the suburb.’In contrast, with disperser ‘to scatter’ (39a), the Ewe allative ɖe is used to case-mark (as Goal) the ground NP because this verb requires the presence of such a case-marker. The SMP duly blocks the French equivalent preposition, dans ‘on, within’, from doing what the allative ɖe does in (39a); see (39b):(39a) wo da tu-a kpo-a mi-a mi disperser ɖe agble-a dzi 3PL throw gun-DEF only-TOP 1PL-TOP 1PL scatter ALLAT farm-DEF top ‘The moment they fired the guns, we (in contrast to other people) scattered on the farm.’(39b) wo da tu-a kpo-a mi-a mi disperser *dans agble-a dzi 3PL throw gun-DEF only-TOP 1PL-TOP 1PL scatter on farm-DEF top E.K. Amuzu / Lingua 137 (2013) 19--37 35Four quick points remain to be made in connection with CS motion verbs. One is that English/French motion+manner verbs do not have to occur in SVCs if the conceptual components of association function and ground are assumed in the discourse context. This is seen in the occurrence of the motion+manner verb dive in (40):(40) le South Africa kple Uruguay-wo ƒe match-a me a, [Suarez dive hafi goalkeeper at South Africa and Uruguay-PL poss match-DEF inside TOP, S. dive before goalkeeper lé aƒɔ nε] gake referee-a ku penalty. catch foot DAT_3sg but referee-DEF blow penalty ‘In the match between South Africa and Uruguay, [Suarez dived before the goalkeeper caught his leg] but the referee whistled for penalty.’The second point is that although motion+path verbs entail the association function, they do not need to occur with a ground NP if the ground entity is assumed. This is illustrated in (41):(41) ne é- ɖi ko, é-ya ŋutɔ drop-ge. if 3sg ripe ADDR 3sg-FOC self drop-PROSP ‘When it ripens, it will drop by itself.’ (Amuzu, 1998:59)The third point is that an English/French motion+path verb may also occur as V1 in at least two types of SVC: where the V2 is another motion+path verb: in (42a), spread occurs in a directional SVC with ge ‘drop’. The implication of (42a) is that the spread of the cancer originated from elsewhere and entered the stomach. If spread were used alone (as in cancer-a spread ɖe dɔme nε), the implication would simply have been that the cancer spread inside the stomach. where the V2 is a non-motion verb: in (42b) collapse occurs in a resultative SVC with wu ‘kill’, which is a change-of-state verb.(42a) Julie kpe fu: cancer-a spread1 ge2 ɖe dɔme nε J. suffer: cancer-DEF spread drop ALLAT stomach DAT.3sg ‘‘Julie suffered: the cancer spread and entered her stomach.’’(42b) wó be storey building aɖe collapse1 he wu2 tens of people. 3PL COMP INDEF ITIVE kill ‘it is said that a storey building collapsed, killing tens of people.’Finally, a non-motion Ewe/English/French verb may join verbs in a bilingual directional SVC. For instance, one may expand the SVC in (32a) with the following two non-motion verbs: no ‘drink’ and ku ‘die’:(43) Fiafi-a dive1 ge2 dze3 tɔ-a-me [he no4 tsi ku5]. Thief-DEF dive drop contact river-DEF-inside ITIVE drink water die ‘The thief dived into the river [and drunk water and died].’4. Conclusions This paper began with two sets of research questions, one about the role that Ewe plays in bilingual SVCs and the other about the role that English/French also plays in the bilingual SVCs. Concerning the first question, the paper establishes that Ewe sets the morphosyntactic frames (i.e. acts as the default ML) of bilingual SVCs in both Ewe-English and Ewe-French CS. The paper demonstrated that Ewe performs this ML role in near- identical ways cross-linguistically. Through the MLF model of Myers-Scotton (1993, 1995, 2002), it was shown that Ewe performs the ML role principally via the operations of the System Morpheme Principle (SMP) and the Morpheme Order Principle (MOP). The SMP ensures that Ewe provides all outsider system morphemes (e.g. aspect and modality categories, case-marking prepositions, and the negation marker) in bilingual SVCs and that the MOP ensures that morpheme/constituent order in the bilingual SVCs is consistent with what obtains in analogous Ewe SVCs. A third principle was also shown to be operational in favour of Ewe morphosyntactic procedures. That is the Blocking Filter. It is what ensures that when an English/French motion+manner verb is used in a CS context, a required association function is not expressed via an English/French path-encoding preposition (as would be the case were English the ML) but via an Ewe motion+path verb. 36 E.K. Amuzu / Lingua 137 (2013) 19--37As noted, the second question concerns whether morphosyntactic differences between English and French reflect in the distribution of morphemes in bilingual SVCs. The default answer to this question is no, and the reason is that the SMP and the MOP consistently steer English and French verbs into the slots in Ewe-based SVCs where Ewe verbs which share their subcategorization features occur. For example, we saw in section 2.3 that English and French verbs with a given transitivity value are distributed in the same way that Ewe verbs with that transitivity value are distributed. In section 2.4 we saw that the only differences in verb distribution in bilingual SVCs are traceable to differences in the varieties of Ewe (e.g. Anlo vs. Gengbe) that are used in Ewe-English CS vs. Ewe-French CS.21 However, the point has also been made that English and French play a limited set of roles in defining the details of the Ewe-based bilingual SVCs. It is these languages which provide details of their verbs’ subcategorization information (e.g. transitivity values and semantic features) that are utilized in the projection of verb slots for them in Ewe-based constructions. For example, we saw that whether an English/ French motion verb will or will not occur in a bilingual SVC that expresses directional motion depends on its semantic features, i.e. on whether it is a motion+manner verb or a motion+path verb (see section 3). It emerged that English case- marking prepositions that the SMP blocks in connection with the distribution of English motion+path verbs are replaced by equivalent Ewe case-marking prepositions while in contrast English path-encoding content-morpheme prepositions that the Blocking Filter blocks in connection with English motion+manner verbs are replaced by Ewe motion+path verbs, which are content morphemes too. The same dichotomy obtains in Ewe-French CS. On the basis of the data analysed, it may be pointed out that the principal reason why bilingual SVCs are produced in Ewe-English and Ewe-French CS is that Ewe (=ML) has SVCs (although English and French do not have them). The importance of this fact is clearer when we consider Essizewa’s (in press) finding from Kabiye-French CS. He finds that bilingual SVCs are not used in Kabiye-Ewe CS ‘‘because Kabiye [the ML] does not have SVCs (i.e. despite the fact that Ewe has them)’’. What these opposite scenarios point to is the prediction that for bilingual SVCs to be characteristic of CS, the ML has to have SVCs even if the EL does not have them. Acknowledgements I am grateful to the Office of Research, Innovation and Development (ORID) of the University of Ghana for giving me a grant with which the Ewe-English and Ewe-French CS data were collected. I am also grateful to the American Council of Learned Societies (ACLS) for offering me a one-year Post-Doctoral fellowship through their African Humanities Program in the 2011--2012 academic year during which the draft of this paper was done. References Aboh, E.O., Essegbey, J., 2010. General properties of the clause. In: Aboh, E.O., Essegbey, J. (Eds.), Topics in Kwa Syntax. Studies of Natural Language and Linguistic Theory, vol. 78. pp. 39--64. Adjei, F.A., forthcoming. Learning how to express motion events in Ewe. Linguistics Department, University of Ghana, Ph.D. Dissertation. Aikhenvald, A.Y., 2006. Serial verb constructions in typological perspective. In: Aikhenvald, A.Y., Dixon, R.M.W. (Eds.), Serial Verb Constructions: A Cross-Linguistic Typology. Oxford University Press, Oxford, pp. 1--68. Ameka, F.K., 2006. Ewe serial verb constructions in their grammatical context. In: Aikhenvald, A.Y., Dixon, R.M.W. (Eds.), Serial Verb Constructions: A Cross-Linguistic Typology. Oxford University Press, Oxford, pp. 124--143. Ameka, F.K., Essegbey, J., 2007. Serializing languages: verb-framed, satellite-framed or neither? In: Maddeson, I., Hyman, L. (Eds.), Linguistic Comparison and Reconstruction in Africa. Africa World Press, Trenton. Ameka, F.K., Kropp Dakubu, M.E., 2008. Imperfective constructions: progressive and prospective in Ewe and Dangme. In: Ameka, F.K., Kropp Dakubu, M.E. (Eds.), Aspect and Modality in Kwa. John Benjamins Publishing Company, Amsterdam/Philadelphia, pp. 215--290. Amuzu, E.K., 1998. Aspects of grammatical structure in Ewe-English codeswitching. University of Oslo, (M.Phil. Thesis). Amuzu, E.K., 2005a. Revisiting the classic codeswitching -- composite codeswitching distinction: a case study of nonverbal predication in Ewe- English codeswitching. Australian Journal of Linguistics 25 (1), 127--151. Amuzu, E.K., 2005b. Ewe-English codeswitching: a composite rather than classic codeswitching. Australian National University, (Ph.D. Dissertation). Amuzu, E.K., 2009. Double plurality in codeswitching. Legon Journal of the Humanities 20, 151--180. Amuzu, E.K., 2010. Composite Codeswitching in West Africa: The Case of Ewe-English Codeswitching. Lambert Academic Publishing, Saarbrucken. Amuzu, E.K., Producing composite codeswitching: the role of the modularity of language production. International Journal of Bilingualism, in press. Available at: http://ijb.sagepub.com/content/early/2013/04/18/1367006913481139. Asilevi, F., 1990. English-Ewe code-mixing in conversational discourse: a case of English as a second language in Ghana. Linguistics Department, University of Ghana, (M.A. Thesis). Bamgbose, A., 1974. On serial verb constructions and verbal status. Journal of West African Languages 9, 17--48.21 We found in section 2.4 that cross-data differences in the expression of aspect closely mirror differences in how aspect is expressed in e.g. Anlo, which is used in Ewe-English, and Gengbe, which is used in Ewe-French. E.K. Amuzu / Lingua 137 (2013) 19--37 37Collins, C., 1997. Argument sharing in serial verb constructions. Linguistic Inquiry 28 (3), 461--497. Dorvlo, K., 2008. A Grammar of Logba (Ikpana). Landelijke Ondeerzoekschool Taalwetenschap, Utrecht, The Netherlands. Durie, M., 1997. Grammatical structures in verb serialization. In: Alsina, A., Bresnan, J., Sells, P. (Eds.), Complex Predicates. CSL1, Stanford, pp. 289--354. Essegbey, J., 1999. Inherent Complement Verbs Revisited: Towards an Understanding of Argument Structure in Ewe. MPI Series. Ponsen & Looijen bv, Wageningen. Essegbey, J., 2008. The potential morpheme in Ewe. In: Ameka, F., Kropp Dakubu, M.E. (Eds.), Aspect and Modality in Kwa Languages. John Benjamins, Amsterdam, pp. p.195--p.214. Essizewa, K., 2007. Sociolinguistic aspects of Kabiye-Ewe bilingualism in Togo. New York University, (Ph.D. Dissertation). Essizewa, K., in press. Lexical insertions in Kabiye-Ewe codeswitching in Lome. International Journal of Bilingualism Forson, B., 1979. Code-switching in Akan-English bilingualism. University of California, (Ph.D. Thesis). Haust, D., 1995. Codeswitching in Gambia. Köppe, Cologne. Johansson, S., Lysvåg, P., 1987. Understanding English Grammar, Part II: A Closer View. Universitetsforlaget, Oslo. Kozelka, P., 1984. The development of national language planning: a case study of language planning in Togo. Stanford University, (Ph.D. Dissertation). Mufwene, S., 1990. Serialization and subordination in Gullah: toward a definition of serialization. In: Joseph, B.D., Zwicky, A.M. (Eds.), When Verbs Collide: Papers from the 1990 Ohio State Mini-Conference on Serial Verbs. Working Papers in Linguistics 39. Ohio State University, Columbus, OH. Myers-Scotton, C., 1993. Duelling Languages: Grammatical Structure in Codeswitching. Clarendon Press, Oxford. Myers-Scotton, C., 1995. A lexically-based model of codeswitching. In: Milroy, L., Muysken, P. (Eds.), One Speaker, Two Languages: Cross- linguistic Perspectives on Code-switching. Cambridge University Press, Cambridge, pp. 233--256. Myers-Scotton, C., 2002. Contact Linguistics: Bilingual Encounters and Grammatical Outcomes. Oxford University Press. Osam, E.K., 1997. Serial verb constructions and grammatical relations in Akan. In: Givon, T. (Ed.), Grammatical Relations: A Functionalist Perspective. John Benjamins, Amsterdam, pp. 253--279. Pawley, A.P., Lane, J., 1998. From event sequence to grammar: serial verb constructions in Kalam. In: Siewierska, A., Song, J.J. (Eds.), Case, Typology and Grammar. John Benjamins, Amsterdam, pp. 201--228. Poplack, S., Meecham, M., 1995. Patterns of language mixture: nominal structure in Wolof-French and Fongbe-French bilingual discourse. In: Milroy, L., Muysken, P. (Eds.), One Speaker, Two Languages: Cross-Disciplinary Perspectives on Code-Switching. Cambridge University Press, Cambridge, pp. 199--232. Quarcoo, M., 2009. Grammatical constraints on noun and verb phrases in Twi-English codeswitching. Department of English, University of Ghana, Legon, (M.Phil. Thesis). Sebba, M., 1987. The Syntax of Serial Verbs: An Investigation into Serialization in Sranan and Other Languages. John Benjamins, Amsterdam. Slobin, D.I., 2004. The many ways to search for a frog: linguistic typology and the expression of motion events. In: Stromqvist, S., Verhoeven, L. (Eds.), Reading Events in Narratives: Typological and Contextual Perspectives. Lawrence Erlbaum Associates, Mahwah, NJ, pp. 219--257. Talmy, L., 2000. Towards a Cognitive Semantics. MIT Press, Cambridge, MA. Veenstra, T., 1996. Serial Verb Constructions in Saramaccan: Predication and Creole Genesis. Holland Institute of Generative Linguistics, Dordrecht.View publication stats