The syntax of Hungarian -vA adverbial participles: A single affix with variable merge-in locations1 Huba Bartos 1. Introduction The Hungarian language has for long had two related constructions, labeled as ‘adverbial participle’ in traditional grammatical terms: the -vA and the -vÁn participles.2 It seems that while they may originally have emerged as dialectal variants, by the early modern times their functional distribution became complementary: the -vÁn participle was rather consistently used as an adverbial of time or reason, while the -vA participle usually assumed the role of manner or state adverbial. Later on, however, this division of labor was gradually lost, and by now, in standard colloquial Hungarian, (i) the -vÁn type has almost entirely disappeared: for many speakers it is clearly archaic, and for the rest its use is very limited (stylistically marked) and rare, its earlier functions basically taken over by the -vA type; (ii) for those whose dialect has retained the -vÁn forms, the functions of the two type heavily overlap, with the exception that the -vÁn participle cannot have a state adverbial reading – the other relevant adverbial functions (time, rea- son, manner, purpose) can in principle be assumed by both of them; (iii) when both are available, -vÁn has a strong tendency for a temporally anterior reading (with respect to the time of the matrix), while -vA is more neutral in this respect.3 This chapter will be exclusively devoted to the discussion of the -vA par- ticiples, for the following reasons: (i) the -vÁn participles play a very limited role (or, for some speakers, no role at all) in present-day Hungarian; (ii) their use is more restricted, with few complications; (iii) arguably, they always project a full-fledged participial clause, whose internals never involve any voice alternation (unlike the -vA participles, where this is a key problem; see below); and (iv) they have received an essentially satisfactory account al- ready, with -vÁn as an inflectional affix, licensing its own (potentially overt) subject, in Sárik (1998). Just for the sake of illustration, (1) gives two exam- ples of the -vÁn participle, by which I part with them for now and turn my attention to the -vA participles alone. 2 Huba Bartos (1) a. Beesteledvén(,) hazaindultunk. in-dusk-VÁN home-started-1pl ‘Night having fallen, we left for home.’ b. A földre feküdvén álomba merült. the ground-onto lie-VÁN sleep-into sank(3sg) ‘Having lain on the ground, he fell asleep.’ Let us now take a look at a few introductory examples of -vA participle, with traditional function identification given for each occurrence: ‘comple- ment’: (2a), ‘predicate’: (2b), simultaneous state adverbial: (2c), anterior state adverbial: (2d), manner adverbial: (2e), and purpose adverbial: (2f). (2) a. Bezárva találtuk az ajtót. in-lock-vA found-1pl the door-ACC ‘We found the door locked.’ b. Zárva volt az ajtó. lock-vA was(3sg) the door ‘The door was locked.’ c. Laci a karosszékben ülve várta a vendégeket. Laci the armchair-in sit-vA waited-3sg the guests-ACC ‘Laci waited for the guests sitting in the armchair.’ d. Laci teljesen felöltözve várta a vendégeket. Laci completely up-dress-vA waited(3sg) the guests-ACC ‘Laci waited for the guests completely dressed.’ e. Futva igyekeztünk haza. run-vA hurried-1pl home ‘We hurried home running.’ f. Kinyitotta az ablakot, utat engedve a füstnek. out-opened(3sg) the window-ACC way-ACC allow-vA the smoke-DAT ‘He opened the window, giving way to the smoke.’ One remark is immediately in order for the function labels: the adverbial in (2a) is easily analysable as a small clause predicate, rather than a comple- ment of the matrix verb talál ‘find’; in fact, all instances of this type can be The Syntax of Hungarian -vA Adverbial Participles 3 analysed this way, conflating the types of (2a) and (2b), simplifying the pic- ture somewhat: adverbial participial phrases/clauses can thus be either predi- cative complements (small clauses), or adverbial modifiers (as in the rest of the examples), such that the participle itself is a predicate within its phrase/clause domain — as will be argued later in detail. Given this surface variability of functions and loci in the syntactic struc- ture, as well as the different sizes of the participial, ranging from a single word (2a, b) to a whole clause (2c,d), it is tempting to make use of several different variants of the participial affix in an analysis, but the more chal- lenging (hence probably more interesting) option is to try to stick to the ‘sin- gle -vA’ hypothesis, seeking a more unified account og the full set of con- structions. In this vein, I will propose to treat the syntax of the -vA partici- ples in an antilexicalist framework, which will be an exercise in pushing the idea of having a single affixal lexical item with variable locus of merging it into the syntactic projection of the host category, rather than positing several homonymic lexical items (as a more traditional and/or lexicalist analysis would have it). The conceptual advantage of this single-item approach is obvious (no need to multiply lexical entries, with different alleged selec- tional properties and semantic effects), but it is a viable alternative only so far as no significant price is paid somewhere else — which I hope to be able to show here not to be the case. 2 Previous treatments of the -vA participles in the generative tradition There have been several attempts to analyse the (morpho)syntax and seman- tics of the -vA participles in the literature, and the discussion has revolved around two main issues: (i) how far do these participles project: are they word-level, phrase-level, or clause-level entities; (ii) is there a single -vA affix, or do we need (at least) two distinct such affixes in the lexicon to ac- count for various subtypes of the -vA participle constructions — the crucial point here is an active ~ passive alternation, in an oft-alleged correlation between the manner ~ state distinction in construal: (3a) = unergative base V, ‘active’; (3b) = unaccusative base V, ‘active’; (3c) = transitive base V, ‘ac- tive’; (3d) = transitive base V, ‘passive’.4 (3) a. Laci mosolyogva válaszolt. Laci smile-vA answered(3sg) ‘Laci answered smiling.’ 4 Huba Bartos b. Teljesen megszáradva(,) a festék védi a fát. completely dry-vA the paint protect(3sg) the wood-ACC ‘Completely dried, the paint protects the wood.’ c. Laci az újságot olvasva jött be a szobába. Laci the newspaper-ACC read-vA came(3sg) in the room-into ‘Laci came into the room reading the paper.’ d. (Az ellenségtől) bekerítve a csapat megadta magát. the enemy-from in-surround-vA the team surrender itself-ACC ‘Surrounded (by the enemy), the team surrendered.’ The manner ~ state split apparently obtains between unergatives and active transitives (manner) vs. unaccusatives and passive transitives (state) – a seemingly clear distinction, easily captured by referring to argument struc- ture (presence vs. absence of an external argument). As will be shown later, in section 4., the semantic distinctions (partly concerning argument struc- tural properties) are much more complicated and fine-grained than this, which suggests (to me) that the relevant semantic variables are interpretive, rather than generative, in the case of adverbial participles: they apply (non- distinct) syntactic structures at the level of interpretations, and cannot rea- sonably be assumed to drive the syntactic derivations. Most of the earlier accounts have been built up on the manner~state dual- ity, without paying attention to the facts that (i) not all instances of the -vA participle fall into either of these two readings (see, e.g., (2f)), and that (ii) the two readings cannot always be neatly told apart, especially in the ‘trans- parent adverbial’ (cf. Geuder 2004, and section 4.1. below) cases: (4) a. [Kezében puskát tartva] bukkant fel a vadász hand-3sg-in gun-ACC hold-vA emerged(3sg) up the hunter a bozótból. the bush-from ‘The hunter emerged from the bush holding a gun in his hand.’ b. Az ingujját sem tűrve fel(,) a kondérban turkált. the sleeve-3sg-ACC nor roll-vA up the cauldron-in poked(3sg) ‘Not even rolling up his sleeves, he was poking around in the caul- dron.’ The Syntax of Hungarian -vA Adverbial Participles 5 In this section, I will sketch the most important accounts, to set up the scene for my own analysis. 2.1. Lexical solutions 2.1.1. Komlósy (1992) – a single -vA suffix Komlósy’s lexical solution uses a single suffix -vA, which can attach either to an active stem, or a lexically passivized one, where passivization involves a null-affix (or conversion), to account for the paradigm in (3): (5) a. LEXICAL PASSIVIZATION 1. subject → oblique (demotion) 2. object → subject (promotion) 3. AGENT is existentially bound (optional; if it applies, it yields an agentless passive) b. ADV-FORMATION -vÁn: [V ___ ] stem/non-passive Adv -vA: [V ___] stem Adv As is clear from these rules, for Komlósy, participle forming is a lexical derivational operation, yielding an adverb. For unergatives, unaccusatives, and non-passivized transitives, the rule is a simple case of V → Adv deriva- tion, while for cases like (3d), the passivization defined in (5a) must precede the Adv-formation. Two points of criticism apply to this account: − It needs a stipulation to the effect that the null-passives formed by (5a) cannot surface without undergoing some further operation, because we do not find null-passivized finite verb forms in Hungarian. The stipula- tive constraint he offers is given in (6). − It does not tell us how/why the Adv’s formed by (5b) project their own modificational domain, which can reach full clausal status (see the ex- amples in (2) and (3)), and how certain argument roles of the base V are identified with arguments in the matrix domain. The reason for the lack of information on these points resides in the sketchy nature of the analy- sis, since Komlósy’s focus was on adjectival participles in that paper. (6) The categories of tense/mood-marking and agreement only apply to active stems in Hungarian. 6 Huba Bartos 2.1.2. Laczkó (2000) – two -vA suffixes Laczkó, primarily on the basis of his critique of Komlósy’s account, devel- ops an analysis that makes use of two -vA affixes in the lexicon, such that -vA involves a feature of passivization in itself, hence it primarily at- 2 taches to passivizable stems (transitives), but since this ‘passivization’ suboperation is about the promotion of an internal argument, it also applies (in a sense vacuously) to unaccusatives. -vA , on the other hand, involves no 1 passivization, and occurs with active transitives and unergatives. This solu- tion establishes a pattern of ‘active_transitive/unergative’ vs. ‘pas- sive_transitive/unaccusative’ grouping, which (Laczkó claims) matches the traditional ‘manner’ vs. ‘state’ distinction in interpretation. In fact, the chief motivation5 for him to combine unaccusatives with -vA (when, in principle, 2 unaccusatives would fit the structural description of the -vA rule, too) is 1 precisely the possibility of correlating the -vA ~ -vA opposition with the 1 2 ‘manner’ ~ ‘state’ opposition. (7) passive reading -vA state összekötözve feküdt 2 PV-tie-vA lay(3sg) transitive ‘she was lying tied up’ active reading -vA manner a csokrot összekötözve leült 1 the bouquet-ACC PV-tie-vA down-sat(3sg) ‘having tied the bouquet she sat down’ unerg. base -vA manner kiabálva rohangált 1 shout-vA run-FREQ-PAST(3sg) intransitive ‘she was running around shouting’ unacc. base -vA state kimelegedve rohangált 2 out-warm-ADVPRT run-FREQ-PAST ‘she was running around sweating hot’ The particulars of Laczkó’s analysis are summarized in the following:6 The Syntax of Hungarian -vA Adverbial Participles 7 (8) a. -vA V + [-vA ] → [V + -vA ] 1 1 PRT 1 PRT (i) the highest argument with a non-objectlike role becomes the subject (ii) subject is realized as PRO b. -vA V + [-vA ] → [V + -vA ] 2 2 PRT 2 PRT (i) the object argument becomes the subject (ii) subject is realized as PRO Apart from the neatness of match between -vA and the state reading, there is 2 some further motivation for assuming that unaccusatives take -vA , rather 2 than -vA : this makes it possible to establish the generalization that it is 1 the -vA participles that occur predicatively in a copular construction: 2 (9) a. * Laci mosolyogva van. – unergative (-vA ) 1 Laci smile-vA is ‘Laci is smiling.’ b. * Laci meg van írva a levelet. – active transitive (-vA ) 1 Laci PERF is write-vA the letter-ACC * ‘Laci is written the letter.’ c. A levél meg van írva. – passive transitive (-vA ) 2 the letter PERF is write-vA ‘The letter is written.’ d. A festék meg van száradva. – unaccusative (-vA ) 2 the paint PERF is dry-vA ‘The paint is(= has) dried.’ One weakness of the analysis is that it says nothing about the identifica- tion of the PRO subjects of these participles. A more serious problem, though, is that this account sees too much into the ‘manner’ ~ ‘state’ distinc- tion, and would therefore need considerable augmentation to cater for (i) other, non-manner non-state adverbial readings; and (ii) interesting grammaticality effects unrelated to the argument structural factors referred to in the lexical rules, such as the following contrast type (my take on which will be presented in section 4.2. below): 8 Huba Bartos (10) TEMPORAL ANTERIOR vs. STATE: a. Szépen/Gyorsan felöltözve(,) elindult munkába. neatly/quickly up-dress-vA away-start work-into ‘Dressed up neatly/quickly, he left for work.’ b. Szépen/*Gyorsan felöltözve találtuk / ült a szobájában. neatly / quickly up-dress-vA found-1pl/sat(3sg) the room-3sg-in ‘We found him / He was sitting neatly/*quickly dressed in his room.’ (11) ‘PURE’ STATE vs. REASON/STATE: a. (?*Orvul) hátbadöfve feküdt a földön. sneakily back-into-stab-vA lay(3sg) the ground-on ‘He was lying on the ground, stabbed in the back (*sneakily).’ b. Orvul hátbadöfve(,) a földön feküdt. sneakily back-into-stab-vA the ground-on lay(3sg) ‘Stabbed sneakily in the back, he lying on the ground.’ (12) ‘PURE’ STATE vs. TEMPORAL/REASON: a. (*Véletlenül) kiborulva találták a levest az asztalon. accidentally out-spill-vA found-3pl the soup-ACC the table-on ‘They found the soup spilt over the table (*accidentally).’ b. Véletlenül kiborulva(,) a leves az egész asztalt accidentally out-spill-vA the soup the whole table-ACC elborította. away-covered(3sg) ‘Having spilt out accidentally, the soup covered the whole table.’ In closing this section, mention must be made of Németh’s (2007) recent contribution to the topic, which does not offer any precise technical analysis, but suggests that at least in the copular predicative ‘V-vA van’ construction, a lexical solution where -vA affixation is sensitive to aspectual verb classes could account for a wide variety of semantic/pragmatic limitations on the availability of -vA participles. The Syntax of Hungarian -vA Adverbial Participles 9 2.2. Syntactic solutions There is another line of research that has attempted to devise some syntactic account for deriving the adverbial participles and their behavior. Some of these (e.g., É. Kiss 1998, Bene 2005) have essentially followed the lead of the lexical analyses, reorchestrating them in the domain of (narrow) syntax, while others (e.g., Kenesei 2000) have tried to look upon the issue from a completely different angle, arguing that participles are special inflectional forms of verbs, thus they naturally project full clauses, and have no deriva- tional properties, so that all of their peculiarities must have a syntactic (rather than lexical, argument structural) explanation. But the most thorough and successful syntactic analysis to date is Tóth’s (2000), which recognizes that -vA affixation can target different syntactic domains (VP, VoiceP, TP), giving rise to units of different status and function. As will be clear in later sections, my present proposal essentially follows her lead, reproducing its key insights in a different model. 2.2.1. Bene (2005) – Two -vA suffixes, transposing Laczkó’s key ideas to syntax Bene analyses the -vA (and, irrelevantly for us, also the -vÁn) participles as cases of V → Adv derivation in syntax, in an articulated VP shell structure, positing two distinct -vA suffixes, explicitly calling them ‘manner -vA’ and ‘state -vA’. Manner -vA always attaches to the largest extension of the VP (i.e., vP if present, VP otherwise) — this applies to unergatives and active transitives (at vP, (13a)), as well as unaccusatives (at VP, this being the larg- est V-projection, (13b)). (13) a. AdvP b. AdvP 3 ei Adv vP Adv VP ! 3 ! ri -vA PRO v’ -vA V DP manner manner 3 ! v VP PRO ACT 6 Note that she recognizes that unaccusatives can host the ‘manner’ type suffix, too, as in the following examples (Bene 2005: 83): 10 Huba Bartos (14) a. A labda [ pattogva] gurult. the ball bounce-vA rolled(3sg) ‘The ball was rolling bouncing.’ b. [Peregve] hullik a falról a vakolat. trickle-vA fall(3sg) the wall-from the plaster ‘The plaster is falling trickling from the wall.’ While the presence of v can take care of the accusative case of some in- ternal argument, the external argument in spec,vP remains without nomina- tive (presumably because the AdvP projected by the participial suffix cannot merge with a T, and the whole AdvP is an island), so its only option is to be represented as a PRO, which can then be identified with some argument in the matrix domain. ‘State -vA’, on the other hand, attaches to VPs whose highest argument is a PATIENT, and this highest argument is realized as a PRO, again for case reasons (no v present to check/assign accusative): (15) AdvP 3 Adv VP ! 3 -vA V PRO state This precludes combining it with unergatives (no PATIENT at all), and forces it to apply to the core VP in transitives, before the vP layer would be built, since at that point their highest argument is still a PATIENT. Whether this is a welcome consequence depends on how you define ‘state adverbial’ — but the following examples suggest that finding the appropriate definition for ‘state adverbial’ may not be an easy job ((4a) is repeated here as (16a)): (16) a. [Kezében puskát tartva] bukkant fel a vadász a hand-3sg-in gun-ACC hold-vA emerged(3sg) up the hunter the bozótból. bush-from ‘The hunter emerged from the bush holding a gun in his hand.’ b. ?[A maciját szorongatva] találtak rá az. the teddy.bear-3sg-ACC clutch-vA found-3pl onto the