West Coast Conference on Formal Linguistics (WCCFL) (38th : 2020)

Verb doubling and Cyclic Linearization Lee, Tommy Tsz-Ming 2020-03-06

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata


73804-Lee_Tommy_Doubling_cyclic_WCCFL38_2020.pdf [ 90.92kB ]
JSON: 73804-1.0389862.json
JSON-LD: 73804-1.0389862-ld.json
RDF/XML (Pretty): 73804-1.0389862-rdf.xml
RDF/JSON: 73804-1.0389862-rdf.json
Turtle: 73804-1.0389862-turtle.txt
N-Triples: 73804-1.0389862-rdf-ntriples.txt
Original Record: 73804-1.0389862-source.json
Full Text

Full Text

Verb doubling and Cyclic Linearization WCCFL-38Verb doubling and Cyclic Linearization*Tommy Tsz-Ming LeeUniversity of Southern Californiaat West Coast Conference in Formal Linguistics 2020 (WCCFL-38)University of British Columbia | Mar 6-8, 2020AbstractThis talk proposes that verb doubling is a consequence of the interaction between Cyclic Lineariza-tion (Fox and Pesetsky 2005, where syntactic structure is linearized cyclically) and Chain Reduction (CR,Nunes 1995, 2004). Substantially, I propose CR is constrained by Linearization Preservation such that CRis suspended as a last resort if it violates Linearization Preservation. The proposal explains the asymmetriesbetween verbs and objects with regards to doubling possibilities in Cantonese.1 Doubling possibilities in CantoneseMovement generally leaves behind a gap, but there are cases where a copy is employed (i.e. doubling of themoved element). This talk tries to model when a copy is prohibited, required, or allowed.Asymmetry 1 While both verbs and objects can be topicalized, verbsmust be doubled, objectsmust not.(1) a. 想 (呢),阿明係*(想)食魚嘅Verb topicalization (Cheng and Vicente 2013)soengwant(ne),topAamingAaminghaifoc*(soeng)wantsikeatjyufishge2sfp‘As for (whether he) wants, Aaming wants to eat fish (but...)’b. 呢條魚(呢),阿明想食(*呢條魚)Object topicalizationni-tiuthis-cljyufish(ne),top,AamingAamingsoengwantsikeat(*ni-tiuthis-cljyu)fish‘This fish, Aaming wants to eat.’Asymmetry 2 While both verbs and objects can be right dislocated, verbs are optionally doubled, objectsmust not be doubled.(2) a. 阿明(食)呢啲野呀食RD of verbs (Lee 2017)AamingAaming(sik)eatni-dithis-cljethingaa4Qsik?eat‘ Aaming eats this thing?’b. 阿明食 (*呢啲野) 呀呢啲野RD of objectsAamingAamingsikeat(*ni-dithis-clje)thingaa4Qni-dithis-clje?thing‘Aaming EATS this thing?’*For comments and discussions, I thank Audrey Li, David Pesetsky, Hajime Hoji, Deniz Rudin, Andrew Simpson, Ka Fai Yip andthe audience in LSA Linguistics Institute 2019. Special thanks go to Stefan Keine for both detailed comments and encouragement. Allremaining errors are mine.1Verb doubling and Cyclic Linearization WCCFL-38Asymmetry 3 Topicalized subjectsmust not be doubled, whereas right-dislocated subject are optionally doubled.(3) a. 阿明(呢),(*阿明)想食呢種魚Subject topicalizationAamingAaming(ne),top(*Aaming)Aamingsoengwantsikeatni-tiuthis-cljyufish‘As for Aaming, (he) wants to eat this fish.’b. (阿明)想食呢種魚呀阿明RD of subjects (Cheung 2009, 2015)(Aaming)Aamingsoengwantsikeatni-tiuthis-cljyufishaa3 Aaming‘Aaming wants to eat this fish.’Summary (i) Object doubling is generally banned.(ii) Verb doubling is obligatory in topic constructions, but optional in RD.(ii) Subject doubling is banned in topic constructions, but optional in RD.Subject Verb ObjectTopic construction S, (*S-)V-O V, S-*(V-)O O, S-V-(*O)Right dislocation (S-)V-O, S S-(V-)O, V S-V-(*O),OTable 1: Doubling possibilities (Keys: dark gray - banned; light gray - optional; white - obligatory)Goal The talks accounts for these asymmetries based on a more elaborated version of Cyclic Linearization (CL,Fox and Pesetsky 2005) and Chain Reduction (CR; Nunes 1995, 2004). In particular, I propose that doubling isa consequenceof suspensionofChainReduction,which is constrainedbyLinearizationPreservation.2 A primer on Cyclic Linearization(4) Cyclic Linearization (Fox and Pesetsky 2005)a. Syntactic structure is linearized cyclically, by establishing Ordering Statements (OS) at each domain.b. Movement across domains is allowed, as long as it obeys Linearization Preservation.c. Linearization Preservation (a “filter” on word order):OS must be obeyed by overt elements in the final output.d. OS are cumulative, and once established, cannot be over-written.Scenarios: licit and illicit movements under CL(5) A copy-theoretic implementation of CL (Chomsky 1995; Nunes 1995, 2004)At each Spell-Out, two independent operations apply one after the other:(i) Chain Reduction (CR, standardly targeting the low copies), followed by(ii) Linearization (LIN, establishing OS).Under CL, ‘edge’ movements (i.e. X-movement), but not ‘non-edge’ ones (Y-movement), are allowed.2Verb doubling and Cyclic Linearization WCCFL-38(6) Scenario 1 (LIND !Moveedge ! CR! LIND’)OSD’: X <  < D(X < Y < Z)[D’ ... X  [D <X> Y Z ]](7) Scenario 2 (LIND !Movenon-edge ! CR! LIND’)*OSD’: Y <  < D(X < Y < Z)* [D’ ... Y  [D X <Y> Z ]]Y-movement is allowed if Y moves to the edge of D before it moves out (i.e. successive cyclic movement).(8) Scenario 3 (Move within D! CR! LIND !Moveedge ! CR! LIND’)OSD’: Y <  < D(Y < X < Z)[D’ ... Y  [D <Y> X <Y> Z ]]Alternatively, a non-edge movement from within D is licit if followed by some ‘compensating movement’.(9) Scenario 4 (LIND !Moveedge +Movenon-edge ! CR! LIND’)OSD’: X < Y <  < D(X < Y < Z)[D’ ... X ... Y  [D <X> <Y> Z ]]Ellipsis ( ̸= CR) also rescues Scenario 2 by not pronouncing some elements, q.v. ‘Salvation by Deletion’.(10) Scenario 5 (LIND !Movenon-edge ! Ellipsis! LIND’)OSD’: Y <  < D(X < Y < Z)[D’ ... Y  [D X Y Z ] ]3 A constraint on Chain Reduction3.1 Proposal and assumptionsProposal First, I propose that Chain Reduction is constrained by Linearization Preservation.(11) Chain Reduction suspensionChain Reduction on a copy is suspended as a last resort if it violates Linearization Preservation.In effect, it opens up a new way for non-edge movement: Multiple pronunciation (i.e. doubling) of Y(12) Scenario 6 (LIND !Movenon-edge ! CR suspension! LIN)OSD’: Y <  < D(X <Y <Z)[D’ ... Y  [D X Y Z ]]Second, I specify how OS is obeyed. 11. (13) follows the spirit of Principle of Minimal Compliance, with can be regarded as its linearization counterpart.(i) Principle of Minimal Compliance (Richards 1998, p.601)For any dependency D that obeys constraint C, any elements that are relevant for determining whether D obeys C can be ignoredfor the rest of the derivation for purposes of determining whether any other dependency D′ obeys C.(ii) Principle of Minimal Compliance, linearization versionFor any two elements  and  that obey an Ordering Statement (OS), any other (identical) copy that is relevant for determiningwhether it obeys that OS can be ignored for the rest of the linearization for purposes of determining whether it obeys that OS.3Verb doubling and Cyclic Linearization WCCFL-38(13) Minimal Compliance to Ordering StatementsFor successful linearization, each OS only needs to be satisfied once.Substantially, (13) suggests that if any one copy in a chain f, g satisfied the established OS (that involves), the other one copy is set free from that OS. In (12), the higher copy Y is free from the OSX < Y.Assumptions I also make the following assumptions:(14) Assumptionsa. vP and CP are Spell-Out domains.b. There is V-vmovement in Cantonese.c. v cannot move to Spec vP.(15) Illicit v-movementvPv vPSubject v’v VP3.2 Doubling effects in verb topicalizationDerivation Verb doubling in topic constructions is derived via (17), abstracting away from the topic marker ne andsentence-final particles ge. I assume that they are both vP-external.(16) Obligatory verb doubling想 (呢),阿明係*(想) 食魚嘅Verb topicalization, =(1a)soengwant(ne),topAamingAaminghaifoc*(soeng)wantsikeatjyufishge2sfp‘As for (whether he) wants, Aaming wants to eat fish (, but...)’(17) Derivation of (16)a. Building of vP (headed by soeng ‘want’)[vP Aaming hai soeng sik jyu ] (14c) bans verb movement to Spec vPb. Spell-Out of vP(CR)! LINvP; OSvP: Aaming < hai < soeng < sik < jyuc. (Non-edge) verb movement for topicalization2[TopicP soeng ... [vP Aaming hai soeng sik jyu ]]d. Spell-Out of TopicPCR suspension! LINTopP; OSTopP: soeng < Aaming < hai < soeng < sik < jyuLow copy : CR is suspended and the lower copy of soeng is not deleted. TheOS hai < soeng in (17b) is obeyed.High copy : by virtue of Minimal Compliance to OS in (13), it is free from the OS hai < soeng because it isalready satisfied once.Predictions What allows the low copy of the verb to escape from CR is the fact that it is preceded by some element inthe vP. We predict that the presence of Aaming or hai is crucial to doubling.2. I abstract over the standard subject movement for its irrelevance.4Verb doubling and Cyclic Linearization WCCFL-38(18) a. 想 (呢) ,係想食魚嘅soengwant(ne),tophaifocsoengwantsikeatjyufishge2sfp‘As for (whether I) want, (I) want to eat fish.’b. 想 (呢),阿明想嘅soengwant(ne),topAamingAamingsoengwantge2sfp‘As for (whether he) wants, Aaming wants.’(19) *想 (呢),想(食魚)嘅Absence of both the subject and hai*soengwant(ne),top,soengwant(sikeatjyu)fishge2sfp‘As for (whether I) want, (I) want (to eat fish).’The structure allowing for doubling is schematically represented as follows:(20) Verb topicalizationCR suspension! LINTopP; OSTopP: V < S/hai < V < XP[TopP V [ ... [vP *(S/hai) V XP ]]]3.3 Subject and object topicalizationObject Verbs are different from objects in terms of movement possibility to Spec vP. In cases where the objectmoves to Spec vP, CR applies as usual (not suspended), resulting in absence of doubling.3(21) 呢條魚(呢),阿明想食(*呢條魚)Object topicalization, =(1b)ni-tiuthis-cljyufish(ne),top,AamingAamingsoengwantsikeat(*ni-tiuthis-cljyu)fish‘This fish, Aaming wants to eat.’(22) Object topicalizationObject movements! CR! LINTopP; OSTopP: O < S < V[TopP O ... [vP <O> S V <O> ]]Upshot Asymmetry 1 observed in (1) is derivable from the structural position (i.e. the launching site) of the verbs andobjects.Verbs are “special” not because they are heads, but because they cannot stop at Spec vP. Objects can stop atSpec vP, hence the absence of doubling (note that CR suspension is the last resort).The head-phrase distinction bears a limited role in accounting for doubling possibility.3. The same line of reasoning applies to subjects in topic constructions, with no suspension of CR.(i) Subject topicalizationSubject movement! CR! LINTopP; OSTopP: S < V < O[TopP S ... [vP <S> V O ]]5Verb doubling and Cyclic Linearization WCCFL-384 Right dislocation4.1 Licit doubling in RDVerbs If doubling of a leftward-moving verb is triggered by preceding vP-internal elements, we expect to see thatdoubling of a rightward-moving verb will be triggered by vP-internal elements that follow it. This is borne out:(23) a. 佢食呢啲野呀食keoihesikeatni-dithis-cljethingaa4Qsik?eat‘ He EATS this thing?’b. *佢食呀食*keoihesikeataa4Qsik?eatIntended: ‘He EATS?’Similarly, an embedded verb enables doubling of an embedding verb (but not vice versa):(24) a. 佢想去架想keoihesoengwantheoigogaa3sfpsoengwant‘He WANTS to go.’b. *佢想去架去*keoihesoengwantheoigogaa3sfpheoigo‘He wants to GO.’Since the established OS in vP dictates that V must precede O/the embedded V in the final output, CR issuspended, or it would violate Linearization Preservation, resulting in verb doubling.Assuming a rightward movement analysis of RD,4 these cases are schematically represented below:5(25) a. The simplified structure of (23)OSvP: S <V < (O)[vP S V *(O) ]] aa4 Vb. The simplified structure of (24)[vP S V1 [TP ... V2 ... ]] gaa3 fV1/*V2gOSvP: S <V1 <V24.2 No doubling in RDVerbs Why is doubling for verbs (and subjects) in RD optional?(26) a. 佢 _ 呢啲野呀食keoiheti ni-dithis-cljethingaa4Qsiki?eat‘He eats THIS THING?’b. The simplified structure of (26a)[vP S O <V> <O> ] aa4 VObject movement! CR! LINvP;OSvP: S < O < VI suggest that verbmovement is preceded by objectmovement before the Spell-Out of vP. After the objectmovement, the verb is on the right edge of vP, subsequent rightward movement does not suspend CR, resultingin no doubling.64. Rightward movement is compatible with Cyclic Linearization, which only imposes restrictions on the final word order.5. The same applies to subjects in RD, as a subject is followed by V and O.This is the case for (3b), with the subject doubled.6. I assume the objectmovement is achieved by some ‘tucking-in’ operation, landing on a position below the subject (Richards 2001).6Verb doubling and Cyclic Linearization WCCFL-38Theobjectmovement is independentlymotivated. First, an object preposing rule is necessary to derive SOVword order in Chinese, which renders the object a contrastive focus (Ernst and Wang 1995, i.a.).(27) 佢呢啲野食 _ 㗎keoihe[ni-dithis-clje]ithingsikeatti gaa4?Q‘He eats THIS THING?’Objects that cannot under object preposing (e.g. bare noun indefinites) cannot undergo RD either.(28) a. *阿明野食 _ 呀*AamingAamingjeithingsikeatti aa4QInt.: ‘Aaming eats?’b. *阿明食 _ 呀野*AamingAamingsikeatti aa4Qjei?thingInt.: ‘Aaming eats?’Second, the movement in (26a) has a similar effect of object focus. In the absence of such movement, as in(23a), the verb receives focus interpretation. In contrast, in (26a), the object is focused.74.3 Illicit doubling in RDObjects The illicit doubling specific to objects can be attributed to the fact that objects are at the right edge of vP.8Rightward movement after the Spell-Out of vP would be subject to CR.(29) a. 阿明食 (*呢啲野) 呀呢啲野=(2b)AamingAamingsikeat(*ni-dithis-clje)thingaa4Qni-dithis-cljething‘Aaming EATS this thing?’b. The simplified structure of (29a)LINvP; OSvP: S < V < O[vP S V <O> ] sfp O7. As for RD of subjects without doubling, I suggest that the VP is fronted to the edge of vP such that the subject is on the rightedge of the vP.The VP in (3b), without subject doubling, receives focus interpretation, a reading that is extensively discussed in Cheung(2009) and earns it the name of Dislocation Focus Construction. If the subject is doubled (i.e. no VP fronting for focus), then the subjectreceives focus interpretation.(i) The simplified structure of RD of subjectsVP movement! CR! LINvP; OSvP: VP < S[vP VP <S> <VP> ] sfp S8. Note that Lai (2019) points out that object doubling cannot be ruled out by avoidance of phonological identity .(i) 佢中意佢呀佢keoiishezungjilikekeoijheraa3sfpkeoiiher‘She likes her.’ p.246, with adaptations7Verb doubling and Cyclic Linearization WCCFL-385 Predictions on object topicalizationFirst, if the object does move (e.g. for topic), doubling is required for RD.This is because the movement withinvP establishes the OSvP: O < S < V. (30a) involves both object topicalization and object RD.(30) a. 呢啲野阿明食呀呢啲野ni-dithis-cljethingAamingAamingsikeataa4Qni-dithis-cljething‘Aaming eats THIS THING?’b. The simplified structure of (30a)[TopP O ... [vP <O> S V <O> ] sfp OObject movement! CR!LINvP;OSvP: O < S < VSecond, we predict that a base generated topic cannot be doubled: as it does not originate within the vP, itis not linearized relative to elements within vP. When it is right-dislocated, CR applies.(31) (*水果)阿明中意梨呀水果(*seoigwo)fruitAamingAamingzungjilikeleipearaa3sfpseoigwofruit‘As for fruits, Aaming likes pears.’Third, object topicalization would consequently block verb doubling in RD, since, in case of object move-ment, the verb is on the right edge of vP upon Spell-Out. Doubling is disallowed in a way similar to a non-toplicalized object.(32) a. 阿明食呢啲野呀食=(23)AamingAamingsikeatni-dithis-cljethingaa4Qsikeat‘ Aaming EATS this thing?’b. *呢啲野阿明食呀食*ni-dithis-cljethingAamingAamingsikeataa4Qsikeat‘ Aaming EATS this thing?’6 Extension: when is verb movement allowed?The current proposal is too strong in predicting that verb movement across the subject must be doubled.(33) Verb movement without doublinga. Swedishhittadefoundhanhefaktistactually(*hittade)foundpengarnamoney.theunderundersängen?bed.the‘Did he actually find the money under the bed?’ (Takita 2010, p.40, with adaptations)b. BulgarianrazkazvalatoldbešewasčestooftenMarijaMaria(*razkazvala)toldtazithisistorijastory‘Maria had often told this story.’ (Harizanov 2016, with adaptations)Parameter I suggest, following Takita (2010), that these languages have a different linerization domain from Cantonese.For these langauges, upon Spell-Out, only the complement of v but not the whole vP is linearized. Accordingly,the order between S and V is not fixed upon Spell-Out of vP.8Verb doubling and Cyclic Linearization WCCFL-38(34) Spell-Out Domain Parameter for vP (Takita 2010)When Spell-Out applies to vP,a. Linearize the whole vP, including the elements on its edge, orb. Linearize the complement of v.If Swedish and Bulgarian take the value of (34b), when Spell-Out applies to the vP, only the VP is linearization(as opposed to vP in Cantonese), illustrated with the Swedish data:(35) The structure of (33a)LINVP; OSVP: V < O < PP[CP hittade [TP han faktist [vP <han> [VP <hittade> pengarna under sängen? ]]]Predictions Interestingly, the parameterwas originally proposed to explain illicit cases of remnantmovements in Japaneseand licit ones in English and German. Takita (2010) suggests (36), whereas the current proposal suggests (37).(36) Remnant movement possibilitya. e.g. JapaneseLanguages that disallow remnant movement must take the value of (34a);b. e.g. German, EnglishLanguages that allow remnant movement must take the value of (34b).(37) Verb doubling possibilitya. e.g. CantoneseLanguages that allow verb doubling must take the value of (34a);b. e.g. Swedish, BulgarianLanguages that disallow verb doubling must take the value of (34b).(36) and (37) combine to predict (38):(38) Predicted complementary distribution of verb doubling and remnant movementa. Languages that allows verb doubling will disallow for remnant movement.b. Languages that allows remnant movement will disallow for verb doubling.(38a) is borne out by the unavailability of remnant vP movement in Cantonese, as in (39a). Note that vP-fronting is allowed if the vP does not contain a trace, as in (39b).(39) a. Raising*變黑,個天開始喇*[vP ti binbecomehakdark,]j , go-tinicl-skyhoicibegintj laa3sfpIntended: ‘To become dark, the sky begins.’b. Control跑⻑跑,佢開始咗喇[vP PRO pauruncoengpaulong.run,]i, keoikeoihoici-zobegin-perfti laa3sfp‘To run long distance, he began.’(38b) is borne out in Swedish: remnant VP topicalization is possible in (40) (Fox and Pesetsky 2005).(40) Remnant movement in Swedish, from Fox and Pesetsky (2005, p.25)? [Gettgivenhenneherti] harhavejagIdeniitintenot...‘I have not given it to her.’9Verb doubling and Cyclic Linearization WCCFL-38Something similar is observed in English, where remnant movement is possible and verb doubling is disal-lowed:(41) Remnant movement in English[Criticized ti by his boss]j , Johni has never been tj.(Takita 2010)(42) Verb doubling in English*Criticize(d), John criticized his boss.7 Take-homemessagesThis talk proposed that verb doubling is a consequence of interaction between Cyclic Linearization and ChainReduction. I showed that the proposal (repeated below) explained doubling possibilities for S, V and O in bothtopic constructions and right dislocations in Cantonese.(43) Chain Reduction suspensionChain Reduction on a copy is suspended as a last resort if it violates Linearization Preservation.(44) Minimal Compliance to Ordering StatementsFor successful linearization, each OS only needs to be satisfied once.ReferencesCheng, Lisa Lai-Shen, and Luis Vicente. 2013. “Verbdoubling in Mandarin Chinese.” Journal of EastAsian Linguistics 22 (1): 1–37.Cheung, Lawrence Yam-Leung. 2009. “Dislocation fo-cus construction inChinese.” Journal of East AsianLinguistics 18 (3): 197–232.. 2015. “Bi-clausal Sluicing Approach to Dislo-cation Copying in Cantonese.” International Jour-nal of Chinese Linguistics 2 (2): 227–272.Chomsky, Noam. 1995.The minimalist program. Cam-bridge, Massachusetts: MIT Press.Ernst, Thomas, and Chengchi Wang. 1995. “ObjectPreposing in Mandarin Chinese.” Journal of EastAsian Linguistics 4 (3): 235–260.Fox, Danny, and David Pesetsky. 2005. “Cyclic Lin-earization of Syntactic Structure.” TheoreticalLinguistics 31 (1-2): 1–46.Harizanov, Boris. 2016. Head movement to specifier po-sitions in Bulgarian participle fronting. Paper pre-sented at the 90thAnnualMeeting of the Linguis-tic Society of America, Jan 7-10,Washington, DC.Lai, Jackie Yan-ki. 2019. “Parallel copying in dislo-cation copying: evidence from Cantonese.” Jour-nal of East Asian Linguistics 3:243–277. doi:10.1007/s10831-019-09195-3.Lee, Tommy Tsz-Ming. 2017. “Defocalization in Can-tonese right dislocation.” Gengo Kenkyu 152:59–87. doi:10.11435/gengo.152.0_59.Nunes, Jairo. 1995. “The Copy Theory of Movementand Linearization of Chains in the MinimalistProgram.” PhD diss., University of Maryland.. 2004. Linearization of Chains and Side-ward Movement. Linguistic Inquiry Monographs.Cambridge, Massachusetts: MIT Press.Richards, Norvin. 1998. “The principle of minimalcompliance.” Linguistic Inquiry 29 (4): 599–629.. 2001. Movement in language: interactions andarchitectures.Oxford: Oxford University Press.Takita, Kensuke. 2010. “Cyclic Linearization and Con-straints on Movement and Ellipsis.” PhD diss.,Nanzan University.10


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items