A COMP-less Approach to Hungarian Complement Clauses

A COMP-less Approach to Hungarian Complement Clauses Péter Szűcs University of Debrecen Proceedings of the LFG’18 Conference University of Vienna Miriam Butt, Tracy Holloway King (Editors) 2018 CSLI Publications pages 325–342 http://csli-publications.stanford.edu/LFG/2018 Keywords: complement clauses, LMT, grammatical functions Szűcs, Péter. 2018. A COMP-less Approach to Hungarian Complement Clauses. In Butt, Miriam, & King, Tracy Holloway (Eds.), Proceedings of the LFG’18 Conference, University of Vienna, 325–342. Stanford, CA: CSLI Publications. Abstract This paper engages in the (X)COMP debate in LFG. It argues that the view from Hungarian supports a “reductionist” position, as Hungarian complement clauses are easily amenable to an analysis with non-COMP functions. I also remark on the wider picture and side with those who would like to maintain a parsimonious inventory of grammatical functions in LFG, in conjunction with a reworked theory of functional and anaphoric control. 1. Introduction: the (X)COMP debate1 There has been a debate in LFG about the grammatical function(s) (GFs) that complement clauses may have. The necessity of the COMP function has been in the heart of the debate. In their seminal paper, Dalrymple & Lødrup (2000) argue that finite complement clauses in English may have either OBJ or COMP grammatical function, depending on the lexical properties of the given predicates. They motivate this bifurcation with differing grammaticality patterns with regards to alternation with NP/DP objects, passivization and coordination, among others. Thus for them, the data in (1)-(6) justifies an analysis where the complement clause of believe is an OBJ, while that of hope is a COMP. Similar views are expressed in Lødrup (2012) and Belyaev (2017). (1) a. I believe that Kate is the winner. b. I believe the story/ it. (2) I believe the story and that it means a lot to you. (3) That Kate won was believed by no one. (4) a. I hope that Kate is the winner. b. *I hope the story/it. (5) *That Kate would win was hoped by no one. There is a more “reductionist” alternative proposal, put forward by Alsina et al. (2005), according to which the COMP function should be dropped from the inventory of grammatical functions in LFG and every finite complement clause should receive some other GF. This position is supported by Forst (2006) and Patejuk & Przepiórkowski (2014, 2016). In this view, the lack of direct object nominals for hope means that the complement function is actually an OBLθ. From this perspective it is not surprising that we find PP-alternatives to it. 1 The Project no. 111918 (New approaches in the description of the grammar of Hungarian pronominals) has been implemented with the support provided from the National Research, Development and Innovation Fund of Hungary, financed under the K funding scheme. 326 (6) Kate hopes for a better result next time. Less attention has been paid to nonfinite complements. Alsina et al. (2005: 41) mentions that as “XCOMP may be considered a special case of COMP, XCOMP should probably go the same way as COMP”, but no detailed investigation is carried out. Patejuk & Przepiórkowski (2014, 2016) in their argumentation for a radical reduction of GFs in LFG explicitly push for the elimination of (X)COMP. They cite examples like (7) to argue that OBJ can also be controlled. They also show that there is a way to implement such an analysis in XLE. (7) I just want friends and to be happy. From an entirely different (“expansionist”) perspective, Falk (2005) proposes that the inventory of grammatical functions in LFG should be enriched, to properly model the cross-linguistic category-function correlations outlined in (8). (Note that Patejuk & Przepiórkowski 2016 explicitly argue against the existence/significance of such correlations.) Falk (2005) puts forward an expanded version of Lexical Mapping Theory (see Table 1), in which he posits two new open functions, XOBLθ and XOBJθ for the complements like the ones in (9) and (10), respectively. (8) a. NP, DP – OBJ b. PP – OBL c. S, IP, CP – COMP d. InfP2 – XCOMP (9) The transformationalist strikes me as crazy. (10) The transformationalist stayed crazy. -r +r +s -o SUBJ OBLθ XOBLθ +o OBJ OBJθ XOBJθ COMP XCOMP -c +c -s +/-o Table 1. Grammatical functions in Falk (2005). (r: restricted, o: objective, c: complement function, s: saturated). The goal of this paper is to add Hungarian to the set of languages that are considered from these perspectives. Overall, I align myself more with the 2 The syntactic category of infinitival clauses may be IP, CP or VP, depending on the particular analysis. I remain neutral on this issue, so the abbreviation “InfP” is used as a shorthand throughout the paper. 327 “reductionist” camp in that I argue that in Hungarian the (X)COMP function is not justified: any possible occurrences may be reduced to non-COMP functions. Whether Falk’s (2005) new functions are justified depends on one’s take on the nature of functional control and the already mentioned categoryfunction correlations. Theoretical and cross-linguistic considerations weigh rather against than for Falk’s (2005) “expansionist” view. 2. Closed complement functions in Hungarian I this section a discuss those cases of Hungarian whereby the complement is functionally complete, so no GF is predicated from the outside. This happens because the subject/object/oblique argument is a simple nominal or a clause with its own subject. The basic pattern is this: Hungarian complement clauses may function either as SUBJ, OBJ or OBLθ arguments of their respective predicates. The primary evidence for this is that there is a systematic alternation in Hungarian whereby the respective grammatical function is realized as a) a lexical noun; b) a pronoun; c) a that-clause; d) an infinitival clause. Let us take a look at the case of SUBJ first. The pattern described above is illustrated in (11) below. That is, derogál (‘feels derogatory’) has the subcategorization frame outlined in (12). Parallel examples could be construed with kellemetlen (‘unpleasant’), sikerül (‘successfully works out’), bejön (‘be appealing’), etc. (11) a. A vereség derogál Katinak. the defeat feels.derogatory Kate.DAT ‘The defeat feels derogatory to Kate.’ b. Az derogál Katinak, hogy vereséget szenvedett. that feels.derogatory Kate.DAT that(c)3 defeat.ACC suffered. ‘It feels derogatory to Kate that she was defeated.’ c. Derogál Katinak, hogy vereséget szenvedett. feels.derogatory Kate.DAT that(c) defeat.ACC suffered. ‘That she was defeated feels derogatory to Kate.’ d. Derogál Katinak vereséget szenvedni. feels.derogatory Kate.dat defeat.ACC suffer.INF ‘To be defeated feels derogatory to Kate.’ (12) derogál <(SUBJ)(OBL)> 3 The “c” stands for “complementizer”. This is to avoid any confusion with the demonstrative in such sentences. If not indicated otherwise, nominative case and present tense assumed in the glosses. 328 As pointed out by Rákosi & Laczkó (2005), this pattern makes straightforward sense if we assume that the underlined constituents uniformly function as the SUBJ of derogál (‘feels derogatory to’), regardless of their categorial status. This is quite straightforward in the case of (11a). In (11b) the subject is the nominative pronoun, and the that-clause is in an appositive relation to this, functioning as an ADJUNCT. 4 If there is no pronoun, just a that-clause, as in (11c), the clause itself is the SUBJ argument. In (11d) the infinitival clause itself functions as the subject of the main predicate. The understood subject of the infinitival is obligatorily controlled by the second, dative argument of derogál (‘feels derogatory to’). This is a major difference compared to the English translation, where the infinitival subject has arbitrary reference. The contrast may be seen from the fact that an explicit subject may be added in English in the form of a for-phrase, but not in Hungarian (Rákosi 2006: 212). (13) For Peter to be defeated feels derogatory to Kate. (14) *Derogált Katinak Péternek felt.derogatory. 3SG Kate.DAT Peter.DAT vereséget szenvedni. defeat.ACC suffer.INF One might suggest that (14) is ungrammatical because there is simply no structural place in the infinitival clause for the overt subject Peter. However, as known since Szabolcsi (2009), Hungarian infinitival clauses do provide a slot for overt subjects, as long as they are pronominal in form and co-referent with the controller. That is, the ‘pro’ subject of the infinitival may be overt as long as it conforms to the normal requirements of the obligatory anaphoric control relations. Szabolcsi (2009) discusses regular, nonsubject clauses, but the argument smoothly carries over to subject infinitivals (the subject of these bears dative case). The additional requirement is that the overt pronominal has to be under the scope of some discourse or quantificational operator. This is just the standard requirement for overt pronominals in such positions in a prodrop language like Hungarian. (15) Derogált Katinaki csak nekii/*j vereséget felt.derogatory.3SG Kate.DAT only her.DAT defeat.ACC szenvedni. suffer.INF ‘Only for her to be defeated felt derogatory to Kate.’ Interestingly, the extraposition-version (as in the translation of (11b)) does not work in Hungarian for the infinitival. Rákosi & Laczkó (2005) explains this by 4 This claim is related to the debate in Hungarian linguistics about the status of such pronouns. Here they are treated as contentful demonstratives and not expletives (contra Kenesei’s (1994) more or less standard account). For a detailed argumentation for this position, see Szűcs (2015) and references therein. 329 stating a requirement that the clause functioning as the adjunct for the demonstrative cannot be headed an infinitival. Hence the contrast in (16) vs. (11b). In sum, the various structures in (11) are realizations of the basic schema shown in (12), the underlined parts of (11) being the SUBJ argument of derogál (‘feel derogatory to’). (16) *Az derogált Katinak vereséget szenvedni. that felt.derogatory.3SG Kate.DAT defeat.ACC suffer.INF Moving on to object clauses, a parallel pattern emerges. The object argument of a verb like akar (‘want’) may be realized as an NP/DP (pronoun, lexical noun), a finite clause or an infinitival. The straightforward approach here is also to posit a single lexical entry. Similar examples could be construed with a próbál (‘try’), utál (‘hate’), szeret (‘like’), etc. (17) a. Kati ételt akar. Kate food.ACC wants. ‘Kate wants food.’ b. Kati azt akarja, hogy együnk. Kate that.ACC wants that(c) eat.3PL.SBJV ‘Kate wants (it) that we eat.’ c. Kati akarja, hogy együnk. Kate wants that(c) eat.3SG.SBJV ‘Kate wants that we eat.’ d. Kati enni akar. Kate eat.INF wants ‘Kate wants to eat.’ (18) akar <(SUBJ)(OBJ)> Finally, the same pattern emerges with OBLθ complements: fél (‘fear’) may occur with a lexical noun or pronoun marked with ablative case, a finite or a non-finite complement clause. Other example verbs are készül (‘prepare’), törekszik (‘strive’) or vonakodik (‘be reluctant’). (19) a. Kati fél a kutyáktól. Kate fears the dogs.from ‘Kate fears dogs.’ b. Kati attól fél, hogy a kutya megharapja. Kate that.from fears that(c) the dog bites.DEF ‘Kate fears that the dog may bite her.’ c. Kati fél, hogy a kutya megharapja. Kate fears that(c) the dog bites.DEF ‘Kate fears that the dog may bite her 330 d. Kati fél kutyát tartani. Kate fears dog.ACC keep.INF ‘Kate fears keeping a dog.’ (20) fél <(SUBJ)(OBLθ)> An interesting contrast between OBJ and OBLθ infinitives may be observed in the so-called “long-distance object definiteness agreement”-phenomenon in Hungarian (first described by É. Kiss 1989 and Kálmán et al. 1989). Szécsényi & Szécsényi (2017) observes that a finite verb may agree in definiteness with the object of its infinitival clause, but this only happens if the main verb is what Szécsényi & Szécsényi (2017) calls an “agreeing verb”.5 The distance between the agreement trigger (the embedded object) and the agreement target (the finite matrix verb) may be arbitrarily long as long as the path only contains “agreeing verbs”. This distinction finds a natural home in an LFG setting as the “agreeing” category shows a near perfect correlation with OBJ-taking verbs while the “non-agreeing” category may be equated with OBLθ-verbs.6 The phenomenon is illustrated in (21)-(22) below. (21) a. Kati akar olvasni egy könyvet. Kate wants.INDEF read.INF a book.ACC ‘Kate wants to read a book. b. Kati akar-ja olvasni a könyvet. Kate wants-DEF read.INF the book.ACC ‘Kate wants to read the book.’ (22) a. Kati fél olvasni egy könyvet. Kate fears read.INF a book.ACC ‘Kate fears reading a book. b. Kati fél(*-i) olvasni a könyvet. Kate fears(-DEF) read.INF the book.ACC. ‘Kate fears reading the book.’ In (21a) the object of the infinitive (a book) is indefinite and the matrix verb is in the default indefinite conjugation. In contrast, the definite object in (21b) (the book) triggers definite conjugation on akar (‘want’). No such variation may be observed with fél (‘fears’): regardless of the definiteness of the embedded object, it is in the default indefinite paradigm. This may be modelled with the following lexical entry on definiteness suffixes in Hungarian. (23) ensures that the agreement path may traverse through OBJ functions, but an See Bárány (2015) for a detailed investigation about definiteness-agreement in Hungarian. 6 The picture is slightly blurred by the fact that some auxiliaries also participate in a long-distance agreement process (e.g. fog (‘will’), talál (‘happen’), etc., see also example (40)). I leave this complication to further research. 5 331 intervening OBL will block it. If (23) is not satisfied, the default indefinite paradigm appears (as elsewhere case). (23) (OBJ+ DEF)=c + The above solution has the drawback of introducing non-locality into agreement, which is theoretically dispreferred. An alternative would be to posit a feature-sharing agreement mechanism, as Haug & Nikitina (2016) suggests for Latin dominant participles. This essentially means that the definiteness feature from the most embedded object “percolates” up to the infinitival itself if it bears the OBJ function and the main verb agrees in definiteness with the infinitival. This may happen in an arbitrary number of steps and locality is ensured. Under this approach the following equation would be available on every infinitival verbal lexical item as an option.7 See Figure 1, for (21). (24) (DEF) = (OBJ DEF) PRED akar ‘want’ <(SUBJ)(OBJ)> SUBJ PRED Kati ‘Kate’ DEF OBJ PRED olvas ‘read’ <(SUBJ)(OBJ)> SUBJ PRED pro DEF OBJ PRED a/egy könyvet ‘the/a book’ DEF +/- Figure 1. F-structure for (21), with feature-sharing. It is to be noted that in all the scenarios above, the different manifestations of the respective grammatical functions may be coordinated, which provides evidence for the uniform functional analysis. Some examples demonstrating this are shown below: in (25) an infinitival is coordinated with a pronoun, in 7 As this is not the main concern of this paper, the ramifications for the overall system of Hungarian agreement are left for further research. For example, finite clauses are not “transparent”, they always count as definite, regardless of their object: (i) Ígére-m/*-k promise-DEF/*INDEF.1SG ‘I promise I read a book.’ elolvasok egy könyvet. read.1SG.INDEF a book.ACC 332 (26) a lexical noun is coordinated with an infinitival, in (27) a lexical noun is coordinated with a finite that-clause. Other combinations are also possible. (25) Derogál Katinak vereséget szenvedni és az, feels.derogatory Kate.DAT defeat.ACC suffer.INF and that hogy ez ilyen gyakran megtörténik. that(c) this so often happens ‘To be defeated and that it happens so often feels derogatory to Kate.’ (26) Kati ételt és azzal jóllakni Kate food.ACC and that.with satisfied.become.INF ‘Kate wants food and to be satisfied with it.’ (27) Kati fél a kutyáktól és Kate fears the dogs.from and megharapják. bite.3PL ‘Kate fears dogs and that they might bite her.’ akar. wants hogy that(c) azok those The general conclusion to be drawn from this section is that the COMP function need not be invoked in the analysis of Hungarian that-clauses. In every case, they are straightforwardly amenable to an analysis in terms of SUBJ, OBJ or OBLθ. The systematic alternation and the coordination possibilities make the alternative, COMP-based alternative unlikely. 3. Open complement functions in Hungarian An open argument function contains a grammatical function (usually the SUBJ) which is the target of a functional control equation, i.e. it is predicated from outside. The stock example for this is the raising construction, where the non-thematic matrix subject is functionally identified with the subject of the infinitival clause.8 In standard LFG, the infinitival bears the XCOMP grammatical function. (28) Kate seems to be happy. As often noted in the literature, Hungarian seems to make a restricted use of InfP in such raising structures. The literal equivalent of (28) is ungrammatical and the state of being happy is expressed as a case-marked adjective.9 (A finite clause along the lines of It seems that Kate is happy is also an option (see 36c below), but that is irrelevant at this point.) 8 Though it is less recognized, such raising structures do occur with finite clauses as well. For an overview, see Ademola-Adeoye (2010). 9 See also Laczkó (2012: 50) for similar points about raising in Hungarian. 333 (29) a. *Kati boldog lenni tűnik. Kate happy be.INF seems. b. Kati boldog-nak tűnik. Kate happy-DAT seems. ‘Kate seems happy.’ It must be noted that some examples of the pattern verb.INF+tűnik may be found in the Hungarian National Corpus. However, this is quite limited: the Hungarian National Corpus returns 41 hits (on closer investigation, even some of these are irrelevant examples). In comparison, the adj.DAT+tűnik pattern returns 4210 sentences. A the closely related látszik (approx. ‘appears’) shows a much more balanced distribution (ca. 3000 hits with both patterns). Nominals are also acceptable if they are predicative and not referential, as shown in (30). (31) is an example with a transitive main verb (“raising to object”). (30) Kati (*az) okos lány-nak tűnik. Kate the smart girl-DAT seems ‘Kate seems a smart girl.’ (31) Kati-t boldog-nak/ zseni-nek tartom. Kati-ACC happy-DAT genius-DAT consider.1SG ‘I consider Kate happy/ a genius.’ Falk (2005: 139) notes that in English, prepositional phrases with an adjectival meaning may be complements of seem. This seems to be barred in Hungarian (magán kívül van ‘to be outside of oneself’ is a fixed expression in Hungarian, meaning ‘to be mad/dazzled’). (32) ?Kate seems out of his mind. (33) *Kati magán kívül Kate herself outside látszik / tűnik. seems appears Furthermore, English seems to allow non-adjectival PPs as in (34), but Hungarian lacks this option as well. (34) I want you out of the room. (35) *Ki/ Kint akarlak téged a szobából. out outside want.1SG you.ACC the room.from What can we distill from this distribution? My position is that the XCOMP function may be dispensed with, regardless of our decision of “reductionist” (Alsina et al. 2005, Patejuk & Przepiórkowski 2014, 2016) or the “expansionist” path (Falk 2005). In Falk’s (2005) approach, the grammatical function of raising infinitivals would be XOBJθ, as their most natural realization is AP and NP. I suggest that 334 even the InfPs may be analyzed as this GF. This should not be a controversial idea since the strict correlation of InfPs with a grammatical function has already been broken in the previous section (there they are SUBJ, OBJ or OBLθ). The reductionist take is that any of the standard grammatical functions may be functionally controlled. In other words, there is “XSUBJ”, “XOBJ”, “XOBLθ”, and “XOBJθ”, but instead of supplying the “X” label, we just need an appropriate theory of functional control. I will briefly look into these matters in the next section. At any rate, the (X)OBJθ seems to be an appropriate function for raising in Hungarian and XCOMP is not needed. Now we have eliminated both “comp” functions from the inventory of the GFs in Hungarian. This again could make sense from both the “reductionist” and the “expansionist” perspective. This is trivial for the “reductionist” camp, but Falk (2005) also mentions that the presence of the +/-c feature could be a matter of cross-linguistic variation, suggesting that Hebrew is a language without +c functions and according to Falk 2005 (referring to Dalrymple and Lødup 2000) Norwegian also makes a very restricted use of COMP and XCOMP. 4. Argument-structure Now that I have outlined my position on the general situation in Hungarian, now it is possible to elaborate on some details of the emerging general picture. In particular, I comment on how the lexical entries may be handled in terms of argument structure. Two issues arise: uniformity of the lexical entries required for the various realizations of the GFs, and the perspective of Lexical Mapping Theory (LMT). 4.1 Lexical uniformity Under the conclusions reached in sections 2-3, a problem with the functional subcategorization of the lexical entries emerges: how to attribute the same lexical entry to the controlled (the infinitival) and the uncontrolled (CP, DP, NP) manifestations of the respective predicates? In standard LFG, f-structural identities are encoded by annotations of identity for raising (36 (=29b)) or coreference for equi (37(=17d)). But then, such annotations are clearly inoperative in (36c) and (37c (=17a)) and would result in invalid f-structures for these sentences. (Note the parallel in the English translations.) (36) a. Kati boldognak tűnik. Kate happy.DAT seems ‘Kate seems happy.’ 335 b. tűnik <(XOBJθ)>SUBJ (SUBJ)= (XOBJθ SUBJ) c. Úgy tűnik, hogy Kati boldog. so seems that(c) Kate happy ‘It seems that Kate is happy.’ (37) a. Kati enni akar. Kate eat.INF wants ‘Kate wants to eat.’ b. want <(SUBJ)(OBJ)> (SUBJ INDEX) = (OBJ SUBJ INDEX)10 c. Kati ételt akar. Kate food.ACC wants ‘Kate wants food.’ To maintain a uniform analysis, a modification is needed in how to establish the control relationship. I find the ideas expressed in Alsina (2008) attractive in this matter.11 He argues that LFG should abandon the lexically encoded annotations of the kind expressed in (36b) and (37b) and the identity-relations should be the results of general constraints like the ones in (38). in addition to the wellestablished Completeness and Coherence conditions of LFG. 10 As want is a control verb, there is a referential identity between the main clause subject and the implicit (“PRO”) subject of the embedded predicate. Thus, for want, I subscribe to anaphoric control. For an illuminating discussion on functional and anaphoric control in equi-type constructions, see Falk (2001: 136-139). 11 While I largely agree with the spirit of Alsina’s (2008) account, certain aspects of it seem too restrictive, e.g. forbidding structure sharing into a finite clause (his SUBJ Binding Condition). As already mentioned (footnote 8), finite raising does exist. Also, finite equi-like structures also seem to be possible, see e.g. Ince (2006) on Turkish. A possible way to reconcile these with Alsina (2008) is to rely on constraint-ranking, whereby certain constraints allowing finite control outrank the SUBJ Binding Condition. Additionally, Alsina’s (2008) account makes a strict correlation between raising and functional control on the one hand, and equi and anaphoric control on the other (as only non-thematic arguments may be structure-shared). However, it is likely that certain equi-verbs establish functional control (see Falk 2001: 136-139). The resolution of these issues is a task for the future. 336 (38) a. Subject Condition12 Every verbal f-structure must have a SUBJ and no f-structure may have more than one SUBJ. b. Nonthematic Condition on GF Identification Structure-sharing of GF s is well-formed only if, in the minimal fstructure containing two structure-shared GFs, one of them: a) is nonthematic and b) is more f-prominent than any GF identified with it. The nonfinite clauses at hand are predicative f-structures, so they must have a SUBJ. This SUBJ must be provided with a PRED-value, otherwise the fstructure becomes semantically incomplete.13 Hence, structure-sharing (functional identification) is mandated. This is what happens with raising sentences like (36a). However, the same process would violate the nonthematic condition in equi-structures like (37a), since both subjects (the matrix and the infinitival) are thematic arguments. In Alsina’s (2008) view, this triggers the appearance of the dummy “pro” PRED value for the embedded subject, bypassing direct structure-sharing in favor of anaphorically binding this ‘pro’. While some aspects of the theory will have to be modified to capture the full range of the data (see footnote 10), the main point is that there is a possibility in the LFG framework to posit uniform lexical entries, by getting rid of the equality-annotations in (36)-(37). 4.2 Lexical Mapping Theory Standard LMT is trivially incompatible with the standard inventory of grammatical functions, as it only provides four options (with r and o specifications), leaving COMP and XCOMP out of the picture. So either the inventory has to be reduced or LMT has to be augmented. In this section I briefly examine these two options. Note that the “verbal” part in (38a) may well be too narrow, given the existence of nonverbal raising structures, see e.g. (32) and (34) above. Also, a reviewer raised the issue of possibly subjectless verbal clauses in German and Polish. This could mean that the Subject Condition is a matter of parametric variation. Alsina (2008, footnote 7) suggests that this may be modelled with an Optimality Theory-based approach to constraint satisfaction. Alternatively, as the reviewer noted, it may well be that the Subject Condition is superfluous, given that the Coherence and Completeness conditions are satisfied. 13 Note that this does not mean that every predicator must have a thematic subject. For instance, raising verbs subcategorize only for a propositional argument, which may be realized as a finite IP/CP. In this case, Completeness is satisfied and an expletive is only inserted because of the Subject Condition (e.g. (36c), it seems that Kate is happy). The nonfinite clauses in (36a, 37a) do not contain a raising predicate, so this is not an option for them. 12 337 In the “reductionist” system, nothing special needs to be added. All the GFs are standardly available: SUBJ, OBJ, OBLθ, OBJθ. The “open” versions of these are handled by the system outlined in the previous section. With a wider array of GFs, changes obviously have to be made. As already shown in Table 1, Falk (2005) adds the features c (complement function) and s (saturated) to make room for the extra functions and sets up a fairly complex mapping system to accommodate the various subcategorizations. A more mainstream conception of LMT is Kibort’s (2007) system, which works with a fixed valency template and a single mapping principle: map the argument to the least marked available grammatical function, markedness defined as having + specifications in the feature-space. As noted, this is entirely compatible with the “reductionist” approach. It is not at all straightforward how Falk’s c and s would fit into this system. Crucially, the main problem is that as long as there is a distinction between open and closed functions, no matter how one places the features into Kibort’s (2007) valency frame, the controlled and the noncontrolled lexical entries ((36a) and (37a) vs. (36c) and (37c)) will always represent two separate lexical entries at the functional level. This is a clear disadvantage compared to the “reductionist” position. Another problematic aspect of Falk’s (2005) expanded LMT is its asymmetry in two respects. For instance, c is neutral with respect to o. Falk justifies this by pointing out that COMP alternates with OBJ, OBJ θ and OBLθ. However, even though COMP also alternates with XCOMP ((36a), (36c)) the very same argument is not used by Falk (2005) to argue that c is neutral with respect to s as well. Thus the argument from alternation is only selectively employed, as an artificial barrier from having to postulate further grammatical functions in the +c realm. The empirical side also seems to militate against the exclusion of open SUBJ and OBJ. Arka & Simpson (1998) analyze certain subject clauses in Balinese as functionally controlled.14 Furthermore, Patejuk & Przepiórkowski (2014) argue that Polish contains functionally controlled OBJ clauses, as in (39), where the controlled infinitive is coordinated with a direct object.15 (39) Chcę pić i papierosa. want.1SG drink.INF and cigarette.ACC ‘I want to drink and (I want) a cigarette.’ There is also some data in Hungarian which point in the direction of functionally controlled OBJ clauses. Based on the long-distance definiteness 14 See Falk (2006) for a differing view. Based on case transmission facts, Patejuk & Przepiórkowski (2014) argue that control in Polish is functional. 15 338 agreement facts discussed in section 3, one may reach the conclusion that the complement of kezd (‘begin’) is an OBJ. The key fact is that kezd (‘begin’) has a nonthematic subject in this example, as evident from the English translations.16 Thus, the identification is functional, yielding a raising structure. (40) a. János kezd szeretni egy könyvet. John begins.INDEF like.INF a book.ACC ‘John is beginning to like a book.’ b. János kezd-i szeretni a könyvet. John begins-DEF like.INF the book.ACC ‘John is beginning to like the book.’ Overall, it seems to me that the “reductionist” approach is theoretically more elegant and is also better equipped to handle cross-linguistic data. 5. Conclusion and future perspectives In this paper I examined the landscape of complement clauses, from the perspective of Hungarian. I argued that complement clauses in Hungarian do not necessitate the (X)COMP function. Finite and non-finite complement clauses are analyzable in terms of SUBJ, OBJ, OBLθ and (X)OBJθ. In my investigation, I surveyed recent trends in LFG’s approach toward the possible grammatical functions and while I cannot say that the debate is settled, the overall picture seems to favor the “reductionist” approach. As one of my reviewers notes, a potential avenue for future research is the extension of the discussion to the analysis of copular clauses. It is important to recognize that copular sentences are not a unitary phenomenon, but several subtypes are to be distinguished, possibly with different versions of the copula (see e.g. Laczkó (2012) and references therein). Some instances lend themselves for a straightforward analysis in terms of OBJ. According to Falk (2005), (41a) is to be analyzed as including a COMP, but given the NP/DP alternative, OBJ is an equally likely option. (41) a. b. The problem is that the hamster will eat the cat. The problem is the cat. 16 The following alternative, with an expletive subject, makes the non-thematic nature of begin’s subject in (40) explicit: ‘it is beginning to be the case that John likes a/the book’. This is equivalent in meaning to the sentences in (40). Note that begin also has a use with a thematic subject, as in (i). Crucially, here the embedded predicate is agentive. (For a scope-based argument on this issue, see Szabolcsi (2009: 254-255)). (i) John began to run. (≠It began to be the case that John ran.) 339 Following Laczkó (2012), existential and locational sentences like (42) might include an OBL (this might be implicit in the case of existentials). (42) a. There are witches (on Earth). b. The cat is in the room. In attributive sentences, the copula might be a pure formative, without subcategorized grammatical functions. (43) The cat is hungry. However, Dalrymple et al. (2004: 193) contends that the PREDLINK function is better suited for sentences like (43) in English. Laczkó (2012) also argues for a PREDLINK-analysis of certain copular constructions in Hungarian. Both Falk (2005) and the “reductionist” approach are reluctant to recognize this GF as a distinct entity, as neither one can naturally accommodate it. Only a careful consideration of the cross-linguistic data and the theoretical consequences can settle this issue. References Ademola-Adeoye, Feyisayo Fehintola. 2010. A Cross-linguistic Analysis of Finite Raising Constructions. Doctoral dissertation, University of KwaZulu-Natal. Alsina, Alex. 2008. In Uyechi, Linda, Lian-Hee Wee (eds.), Reality Exploration and Discovery: Pattern Interaction in Language and Life. Stanford: CSLI Publications. Alsina, Alex, KP Mohanan, Tara Mohanan. 2005. How to get rid of the COMP. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG05 Conference. Stanford: CSLI Publications. 21-41. Arka, I Wayan, Jane Simpson.1998. Control and complex arguments in Balinese. In Butt, Miriam, Tracy H. King (eds.), The proceedings of the LFG’98 conference, University of Queensland, Brisbane: CSLI Publications. Bárány András. 2015. Differential object marking in Hungarian and the morphosyntax of case and agreement. Doctoral dissertation, University of Cambridge. Belyaev, Oleg. 2017. In Defense of COMP Complementation in Moksha Mordvin. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG17 Conference. Stanford: CSLI Publications. 84-103. Dalrymple, Mary & Helge Lødrup. 2000. The grammatical functions of complement clauses. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG00 Conference. Stanford: CSLI Publications. 104-121. 340 Dalrymple, Mary, Helge Dyvik & Tracy H. King. 2004. Copular complements: Closed or open? In Butt, Miriam & Tracy H. King (eds.) Proceedings of the LFG04 Conference. Stanford: CSLI Publications. 188-198. É. Kiss Katalin. 1989. Egy főnévi igeneves szerkezetről. In Kiefer Ferenc (ed.), Általános Nyelvészeti Tanulmányok XVII. Budapest: Akadémiai Kiadó. 153–69. Falk, Yehuda N. 2001. Lexical-Functional Grammar. Stanford: CSLI Publications. Falk, Yehuda N. 2005. Open argument functions. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG05 Conference. Stanford: CSLI Publications. 136-153. Falk, Yehuda N. 2006. Subjects and Universal Grammar: An explanatory theory. Cambridge: Cambridge University Press. Haug, Dag, Tatiana Nikitina. 2016. Feature-sharing and agreement. Natural Language & Linguistic Theory 34(3): 865-910. Ince, Atakan. 2006. Direct Complement Clauses as Object Control Structures in Turkish. In Bainbridge, Erin, Brian Agbayani (eds.), Proceedings of the thirty-fourth Western Conference On Linguistics. Department of Linguistics, California State University, Fresno. 208-221. Kálmán C. György, Kálmán László, Nádasdy Ádám,, Prószéky Gábor. 1989. A magyar segédigék rendszere. In Kiefer Ferenc (ed.), Általános Nyelvészeti Tanulmányok XVII. Budapest: Akadémiai Kiadó. 49–103. Kenesei István. 1994. Subordinate clauses. In Kiefer Ferenc, É. Kiss Katalin (eds.), The syntactic structure of Hungarian. San Diego: Academic Press. 141-165. Kibort, Anna. 2007. Extending the applicability of Lexical Mapping Theory. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG08 Conference. Stanford: CSLI Publications. 250–270. Laczkó Tibor. 2012. On the (un)bearable lightness of being an LFG style copula in Hungarian. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG12 Conference. Stanford: CSLI Publications. 341-361. Lødrup, Helge. 2012. In search of a nominal COMP. In Butt, Miriam, Tracy H. King (eds.), Proceedings of the LFG12 Conference Stanford: CSLI Publications. 383-404. Rákosi György, Laczkó Tibor. 2005. Verbal category and nominal function – Evidence from Hungarian Subject Clauses. In Butt, Miriam and Tracy H. King (eds.), Proceedings of the LFG05 Conference. Stanford: CSLI Publications. 353-370. Patejuk, Agnieszka, Adam Przepiórkowski. 2014. Control into selected conjuncts. In Butt, Miriam and Tracy H. King (eds.), Proceedings of the LFG14 Conference. Stanford: CSLI Publications. 448-460. Patejuk & Przepiórkowski. 2016. Reducing grammatical functions in LFG. In Arnold, Doug, Miriam Butt, Berthold Crysmann, Tracy H. King, Stefan 341 Müller (eds.), Proceedings of the Joint 2016 Conference on Head-driven Phrase Structure Grammar and Lexical Functional Grammar. Stanford: CSLI Publications. 541–559. Rákosi György. 2006. Dative experiencer predicates in Hungarian. Doctoral dissertation, University of Utrecht. Szabolcsi Anna. 2009. Overt nominative subjects in infinitival complements in Hungarian. In den Dikken, Marcel, Robert M. Vago (eds.), Approaches to Hungarian 11. John Benjamins. 251-276. Szécsényi Kriszta, Szécsényi Tibor. 2017. Definiteness agreement in Hungarian multiple infinitival constructions. In Emonds, Joseph, Markéta Janebová (eds.), Language Use and Linguistic Structure: Proceedings of the Olomouc Linguistics Colloquium. Olomouc: Palacký University. 75-89. Szűcs Péter. 2015. On pronouns in Hungarian complex sentences. Argumentum 11: 292-313. 342