119 Susanne Wurmbrand * UDK 811.163'367.625:81'362 University of Vienna, Austria DOI: 10.4312/linguistica.60.1.119-137 Iva Kovač ** University of Vienna, Austria Magdalena Lohninger *** University of Vienna, Austria Caroline Pajančič **** University of Vienna, Austria Neda Todorović ***** University of British Columbia, Canada FINITENESS IN SOUTH SLA VIC COMPLEMENT CLAUSES: EVIDENCE FOR AN IMPLICATIONAL FINITENESS UNIVERSAL 1 INTRODUCTION The notion of finiteness is a much-debated topic in syntax, morphology, and seman- tics, as many fundamental questions have not been answered yet. Due to the vast cross-linguistic variation in the distribution of finiteness, many works have concluded that there is no single morpho-syntactic definition of finiteness, nor a single semantic function associated with it (see e.g., Cristofaro 2007; Bisang 2007; Nikolaeva 2007). Among the morpho-syntactic categories that have been suggested to reflect finiteness are tense, aspect, mood, illocutionary force, person marking, politeness, special forms not used in independent clauses, and/or nominal morphology on the verb (see the works in Nikolaeva 2007). Semantically, finiteness is often associated with clausal independ- ence, specifically, the possibility of a sentence to occur as a free-standing (declarative) main clause (Maas 2004; Bisang 2007; Givón 1990), or, particularly in the Government and Binding and Minimalism traditions, with independent tense or an anchoring to the logophoric center of a clause (Bianchi 2003; Adger 2007). Despite the (theoretical and empirical) variation, various distributional tendencies and patterns have been observed (see Givón 1990; Cristofaro 2007), which indicate that the distribution is not arbitrary but follows certain implicational relations. In this paper, we will look at variation in the distribution of finiteness in a well- defined empirical domain – the South Slavic languages (SSL), and show that it falls * susanne.wurmbrand@univie.ac.at ** iva.kovac@univie.ac.at *** magdalena.lohninger@univie.ac.at **** caroline.pajancic@univie.ac.at ***** neda.todorovic@ubc.ca Linguistica_2020_1_FINAL.indd 119 Linguistica_2020_1_FINAL.indd 119 22. 12. 2020 15:07:38 22. 12. 2020 15:07:38 120 along two dimensions: language and type of complement. To illustrate, while Serbian (Sr) allows finite and non-finite complements of verbs like try (1a), Slovenian (Sl) only allows non-finite complements in this context (1b). 1 On the other hand, when the matrix verb is a speech verb like claim in (1c), the distribution in Slovenian is exactly the op- posite from (1b) – only a finite complement is possible. (1) a. Pokušala sam {da čitam / čitati} ovu knjigu. (Sr) tried.sg .f aux .1sg {da read.1sg / read.inf .ipfv } this book ‘I tried to read this book.’ b. Poskusila sem {*da berem / brati } to knjigo. (Sl) tried.sg .f aux .1sg {*da read.1sg / read.inf .ipfv } this book ‘I tried to read this book.’ (Adrian Stegovec, p.c.) c. Trdim, {da berem / *brati} to knjigo. (Sl) claim.1sg {da read.1sg / *read.inf .ipfv } this book ‘I claim that I am reading/to be reading this book.’ (Adrian Stegovec, p.c.) The property in (1c) holds for all SSL and we propose that the uniformity in the choice of finite forms in these types of complements is due to a grammatical constraint shared by all SSL. We also show that in addition to the two types of complements in (1), there is a third class, illustrated by complements to verbs like decide in (2), which shows flexibility regarding finiteness in most, but not all SSL languages. (2) Odločila sem se {brati /da bom brala} (Sl) decided.sg .f aux .1sg refl {read.inf .ipfv /da will.1sg read.sg .f } to knjigo. this book ‘I decided to read this book.’ (Adrian Stegovec, p.c.) Comparing constructions like (1) and (2) across the SSL, we argue that complement clauses form a semantic hierarchy, and that the distribution of (non-)finiteness in the SSL reflects an implicational scale along this hierarchy. We suggest that the implica- tional nature of the hierarchy (and the distribution of finiteness) is derived via contain- ment relations of clausal domains. Lastly, we propose that the variation is the result of different distributions of the features triggering finiteness, specifically in SSL, agree- ment features. Our findings thus support the existence of implicational hierarchies, defined in semantic terms, on which morphological coding, such as manifestations of finiteness, operates. More broadly, we will conclude that: i) there is no (universal) semantic correlate of (non-)finiteness, since cross-linguistically, all types of complements can be realized as 1 Since we will argue that da does not always act as a complementizer, we gloss it as da in the examples. Linguistica_2020_1_FINAL.indd 120 Linguistica_2020_1_FINAL.indd 120 22. 12. 2020 15:07:38 22. 12. 2020 15:07:38 121 finite or non-finite in at least some language; ii) finiteness is not confined to a particular domain in the clause, but can be distributed over all clausal domains (see also Adger 2007); but iii) there are systematic implicational relations which hold among different types of complements and which, despite possibly arbitrary specific finiteness settings, allow us to predict certain properties of finiteness. 2 IMPLICATIONAL COMPLEMENTATION HIERARCHY (ICH) Languages exhibit a variety of different types of complementation, which can be di- vided into different classes based on their semantic and/or their morphosyntactic prop- erties, such as finite/non-finite, subjunctive, nominalization (see among others Givón 1980; Pesetsky 1992; Horie 2001; Cristofaro 2005; Dixon 2006; Noonan 2007 for dif- ferent approaches). This paper follows Givón’s basic insight that classes are defined semantically, forming an implicational complementation hierarchy (ICH), which we define below. Syntactic and morphological distinctions (such as finiteness), if present in a language, operate along that scale. While morphosyntactic effects may be neu- tralized in different languages, the semantic scale is observable cross-linguistically, and morphosyntactic distinctions can never go against the hierarchy. Wurmbrand and Lohninger (to appear) propose that in addition to the fine-grained semantic scale given in Givón (1980), languages bundle categories into three super-sets, which we refer to as Proposition, Situation, and Event, adopting the terminology and definitions in Ram- chand and Svenonius (2014) (a similar classification has been proposed in Rochette 1988, 1990, although with different terminology). Complements of the type Proposi- tion involve speech, epistemic, and factive contexts. These types of complements are temporally independent, have no pre-specified tense value, are anchored in an utterance or embedding context, and may involve speaker-oriented parameters. Complements of the type Situation involve emotive and irrealis contexts. These types of complements are elaborate eventualities without speaker- and utterance-oriented properties, but with time and world parameters, allowing them to refer to a specific, possibly pre-deter- mined, time. The most common type of Situation complements are forward expanded unrealized events where the time of the complement must be after the time of the ma- trix verb (Abusch 2004; Wurmbrand 2014b). Complements of the type Event include implicative and strong attempt contexts. These types of complements are semantic Properties (Chierchia 1984; Wurmbrand 2002) in that they lack speaker- and utter- ance-oriented, as well as time and world parameters; they are tenseless, may involve actuality entailments (Hacquard 2006), and may have reduced argument structure and/ or event properties. Using this classification, Wurmbrand (2014a, 2015), and Wurm- brand and Lohninger (to appear) establish the implicational nature of the ICH as in Table 1. Independence refers to properties such as the presence and/or interpretation of an independent subject in the complement clause; transparency characterizes whether the embedded clause is permeable for certain cross-clausal operations or dependencies; and integration refers to the degree to which the embedded predicate is incorporated into the matrix predicate. Linguistica_2020_1_FINAL.indd 121 Linguistica_2020_1_FINAL.indd 121 22. 12. 2020 15:07:38 22. 12. 2020 15:07:38 122 Table 1: Implicational complementation hierarchy (ICH) most independent least transparent least integrated Proposition ≫ Situation ≫ Event least independent most transparent most integrated ICH hierarchy effects have originally been discussed predominantly for infini- tives, where clause union or restructuring effects are easily detectable. However, in Todorović and Wurmbrand (2020) and Wurmbrand and Lohninger (to appear), it is observed that even in languages with no infinitives, complements show (in)dependence and transparency effects that track the ICH. Moreover, the semantic classes identified above are typologically robust, whereas morphosyntactic properties differ significantly across languages. If, as we suggest, the basic complementation classes are defined se- mantically by the ICH, we expect to see the different (semantic) types of complements to display variable properties, even when morphosyntactic finiteness does not distin- guish between them. That the ICH applies to finite and non-finite complements alike is illustrated here by the distribution of clause-introducers in complement clauses in Bul- garian and Macedonian, both languages that do not have infinitives. As shown in (3a), Proposition complements must occur with če in Bulgarian and cannot be introduced by da. Situation complements as in (3b) can occur with either če or da. Lastly, Event complements can only occur with da, as in (3c). (3) a. Lea tvărdi {če / *da} čete kniga. Lea claim.prs .3sg {that / *da} read.prs .3sg book ‘Lea claims that she is reading a book.’ (Marchela Oleinikova, p.c.) b. Lea reši {če *(šte) / da } čete kniga. Lea decided.prf .3sg {that *(will) / da } read.prs .3sg book ‘Lea decided to read/that she will read a book.’ c. Lea se opitvaše {*če / da} čete kniga. Lea refl try.prf .3sg {*that / da} read.prs .3sg book ‘Lea tried to read a book.’ The same restrictions hold for Macedonian as illustrated in (4). (4) a. Lea tvărdi {deka / *da} čita kniga. Lea claim.prs .3sg {that / *da} read.prs .3sg book ‘Lea claims that she is reading a book.’ (Sandra Jakimovska, p.c.) b. Lea se rešila {deka *(e) / da } čita kniga. Lea refl decided.prf .3sg {that *(will) / da } read.prs .3sg book ‘Lea decided to read/that she will read a book.’ Linguistica_2020_1_FINAL.indd 122 Linguistica_2020_1_FINAL.indd 122 22. 12. 2020 15:07:38 22. 12. 2020 15:07:38 123 c. Lea probala {*deka / da} čita kniga. Lea try.prf .3sg {*that / da} read.prs .3sg book ‘Lea tried to read a book.’ As shown in (3b), while da complements can encode future directly, če/deka com- plements require an overt future element to be interpreted as a Situation complement (see section 4.1 for an explanation of this difference). In Bulgarian, some speakers may also allow a decide complement introduced by če and no overt future. In this case, the configuration typically cannot receive a Situation interpretation, but is instead shifted to a Proposition context such as a performative use where the matrix subject evaluates or assigns truth to the embedded proposition (similar to cases like I decided that he is a nice person; but other attitude meanings are possible for some speakers as well). Thus, predicates may sometimes undergo a class change, and it is therefore essential to pair the morphosyntactic properties with the interpretation when evaluating ICH ef- fects. If a Situation predicate can be changed or coerced into a Proposition predicate, it follows from the ICH (which is defined semantically) that the predicate then takes on the morphosyntactic properties of the Proposition class (such as the complementizer če). The distribution of complement clauses in Bulgarian (Bu) and Macedonian (Ma) is summarized in Table 2. Table 2: Clause introducers in Bu, Ma Proposition Situation Event če/deka ✓ ✓ (with future) * da * ✓ ✓ As shown in Table 2, Bulgarian and Macedonian exhibit what we refer to as an ICH signature effect: the two classes at the opposite ends of the ICH scale show the opposite properties, while the class in the middle is ‘in-between’ in that it is compatible with both properties (in this case). In the next section, we will show that the distribution of finiteness in SSL shows ICH signature effects and follows the ICH in an interesting way. In section 4, we then suggest a direction for deriving the ICH and the finiteness distribution. Section 5 concludes the paper. 3 FINITENESS HIERARCHY 3.1 A Finiteness ICH Signature Effect We now turn to the SSL that have infinitives, i.e., Bosnian, Croatian, Slovenian, and Serbian. To illustrate that SSL show ICH signature effects regarding (non-)finiteness, we begin by summarizing the finiteness property of the three types of complements in Slovenian, as shown in (5) (repeated from (1) and (2); see below for further explana- tions regarding (5b)). Linguistica_2020_1_FINAL.indd 123 Linguistica_2020_1_FINAL.indd 123 22. 12. 2020 15:07:38 22. 12. 2020 15:07:38 124 (5) a. Trdim, {da berem / *brati} to knjigo. (Sl) claim.1sg {da read.1sg / *read.inf .ipfv } this book ‘I claim that I am reading/to be reading this book.’ b. Odločila sem se {brati / da bom brala / decided.sg .f aux .1sg refl {read.inf .ipfv / da will.1sg read.sg .f / (*) da berem} to knjigo. (*) da read.1.sg} this book ‘I decided to read this book.’ c. Poskusila sem {brati / *da berem / *da bom tried.sg .f aux .1sg {read.inf .ipfv / *da read.1sg / *da will.1sg brala} to knjigo. read.sg .f } this book ‘I tried to read this book.’ As in all SSL, Proposition complements can only be finite (5a). Situation comple- ments can be either non-finite or finite. In Slovenian (and many other languages), finite complements usually require an overt future tense to convey the same meaning as the infinitive in (5b), unless the language and context allows future interpretations to be ex- pressed by present tense (e.g., as in planned or scheduled events). If a present-for-future use is not possible, for whatever reasons, a finite form without overt future (i.e., the third option in (5b)) is excluded when the intended reading is the same as the reading of the corresponding infinitival construction. Similarly to Bulgarian, however, this form may still be rendered acceptable if the matrix verb is shifted to a performative (i.e., a Proposition) interpretation (e.g., (5b) could be used in a situation where the subject is making a decision about which book to read in a play and then declares which book they decided on). Finally, Slovenian Event complements, as shown in (5c), can only occur as non-finite. These observations are summarized in Table 3. Table 3: Finiteness in Slovenian complements Proposition Situation Event finite ✓ ✓ (with future) * non-finite * ✓ ✓ Comparing the distribution of Bulgarian and Macedonian clause introducers with the finiteness distribution in Slovenian, we see an obvious parallel. Both cases display a clear ICH signature effect: the Proposition and Event classes show opposing properties, whereas the Situation class allows both (albeit with certain restrictions). Linguistica_2020_1_FINAL.indd 124 Linguistica_2020_1_FINAL.indd 124 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 125 3.2 A Possible Finiteness Universal In the broader context of South Slavic, different languages show different patterns of availability of infinitives and finite forms; the distribution of these forms seems to be largely dependent on the geographical location and language contact. As can be seen in Figure 1, Bulgarian and Macedonian do not allow infinitives at all, Croatian is the most infinitive-friendly language, and Bosnian, Slovenian, and Serbian occupy the middle of the scale; Slovenian inclines more towards the non-finite, Serbian towards the finite extreme of the scale, and Bosnian is in the middle (judgements for Bosnian differ, thus illustrating the language contact situation quite well). Figure 1: Infinitives in the SSL Since Bulgarian and Macedonian do not use infinitives, we concentrate here on the grammatical patterns of the other four SSL, Bosnian (Bo), Croatian (Cr), Slovenian (Sl), and Serbian (Sr). The distribution is given in (6). Note that due to the contact situation of these languages, language/dialect affiliation is not always clear-cut. Since categorical judgments may not always be possible, the marks should be understood as preferences. The data are given in Serbian in (6) (see Vrzić 1996), but the judgments are to be understood as applying to the translations of these examples into the different languages. (6) a. Tvrdim {da čitam / čitati} ovu knjigu. claim.1sg {da read.1sg / read.inf .ipfv } this book ‘I claim that I am reading this book.’ (finite) ✓Sr, ✓Bo, ✓Sl, ✓Cr ‘I claim to be reading this book.’ (non-finite) *Sr, *Bo, *Sl, *Cr b. Odlučila sam {da čitam / da ću čitati / čitati} decided.sg .f aux .1sg {da read.1sg / da will.1sg read / read.inf .ipfv } ovu knjigu. this book ‘I decided that I will read this book.’ (finite) ✓Sr, ✓Bo, ✓Sl, */?Cr ‘I decided to read this book.’ (non-finite) ?Sr, ✓Bo, ✓Sl, ✓Cr c. Pokušala sam { da čitam / čitati} ovu knjigu. tried.sg .f aux .1sg { da read.1sg / read.inf .ipfv } this book ‘I tried that I am reading/will read this book.’ (finite) ✓Sr, ?Bo, *Sl, *Cr ‘I tried to read this book.’ (non-finite) ✓Sr, ✓Bo, ✓Sl, ✓Cr As shown in (6a), the SSL are uniform with Proposition complements in that they disallow infinitives across the board. The Situation class in (6b) allows infinitives in Linguistica_2020_1_FINAL.indd 125 Linguistica_2020_1_FINAL.indd 125 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 126 all SSL (even though they are dispreferred in many Serbian varieties), but otherwise exhibits variation. Serbian (and possibly Slovenian) 2 can express the future meaning with a finite present tense form. Bosnian, Slovenian, and Serbian allow infinitives or finite overt future forms, but Croatian strongly disprefers any finite form. 3 Lastly, the Event class in (6c) permits infinitives in all SSL, with it being the only possible form in Croatian and Slovenian. A finite complement clause is allowed in Serbian and possibly in Bosnian, but judgements in Bosnian differ and we have not been able to conclu- sively determine the distribution. The distribution of finite vs. non-finite complements is summarized in Table 4 (due to the variation within Bosnian, we have not been able to conclusively allocate it to a category, and we therefore list it in two places). The distri- bution clearly shows that there is a variation according to the two parameters: language and type of complement. Table 4: Finiteness in South Slavic Proposition Situation Event Bulgarian, Macedonian finite finite finite Serbian, Bosnian? finite (non-)finite (non-)finite Slovenian, Bosnian? finite (non-)finite non-finite Croatian finite non-finite non-finite The tendencies observed in the SSL reveal a clear finiteness scale which follows the ICH: a type of complement can never be ‘more’ finite than the type of complement to its left. On the basis of this distribution we propose the (hypothetical) finiteness universal in (7). (7) (Hypothetical) Finiteness Universal If a language {allows/requires} finiteness in a type of complement, all types of complements further to the left on ICH also {allow/require} finiteness. 4 A SKETCH OF AN ACCOUNT 4.1 Towards Deriving the ICH While the ICH in Givón 1980 is defined functionally, Wurmbrand and Lohninger (to appear) propose a grammatical approach. To address the question of how the ordering and implicational nature of the ICH arise, we start with the mapping of the semantic sorts Proposition, Situation, and Event to syntax. Following Ramchand and Svenonius (2014), Propositions, Situations, and Events are semantic sorts expressing conceptual 2 An anonymous reviewer suggested that this may be possible in Slovenian as well, but we have not investigated yet whether this is restricted to present-for-future contexts or possible for future statements in general. 3 This has been confirmed by speaker judgements, preliminary corpus searches on Google, and a query on hrWaC (the Croatian web corpus). Linguistica_2020_1_FINAL.indd 126 Linguistica_2020_1_FINAL.indd 126 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 127 primitives which are in a coherent containment relation – Situations are elaborations of Events, Propositions are elaborations of Situations. More specifically, Situations are created by combining time/world parameters with an existentially closed Event, and Propositions combine speaker-oriented/discourse-linking parameters with an existen- tially closed Situation. In other words, these semantic sorts are computed in a predict- able way by combining the verb with its arguments (creating an Event), by relating an (existentially closed) Event to a time through T or other temporal elements (creating a Situation), and by anchoring a Situation to a context through an element of the opera- tor domain (e.g., C, creating a Proposition). Note that while the broad distinction into clausal domains is considered a general property of phrase structure (see also Grohm- ann 2003), the detailed internal organization of these domains (e.g., CP, TMA, and vP) may vary cross-linguistically. Figure 2 illustrates the containment relations among clausal domains and their semantic correspondences. Figure 2: Clausal domains Since, as we propose, complementation is also defined by the semantic sorts Propo- sition, Situation, and Event, the same containment implications arise. As specified in Table 5, the three types of complements have different minimal requirements. Propo- sition complements are cognition and utterance complements with independent and not predetermined tense interpretations (I said that he left/will leave/is leaving). There is still a tense dependency in the sense that the tense in every complement clause is interpreted relative to the event time of the matrix predicate, but, crucially, the choice of the embedded tense value is free. Following Kratzer (2006) and Moulton (2009a,b), aspects of the meaning of an attitude configuration are situated in the operator domain of the complement clause. The operator domain also separates the matrix predicate and the embedded temporal domain, leading to the tense value independence of these complements. Situation complements involve an independent temporal domain in that the matrix and embedded temporal interpretation can differ (e.g., I decided yesterday to leave tomorrow). However, the tense value is predetermined – the embedded clause, whether infinitive or finite, should be situated after the event time of the matrix predi- cate. There is thus a closer/tighter connection between the matrix predicate and the embedded temporal domain. Lastly, Event complements do not involve an independ- ent tense – they are always interpreted as simultaneous with the matrix predicate. The ranking and implicational nature of the ICH can then be seen as a reflex of the resulting semantic complexity scale. Linguistica_2020_1_FINAL.indd 127 Linguistica_2020_1_FINAL.indd 127 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 128 Table 5: Clausal domains and complexity Proposition Situation Event Minimal requirement Operator domain TMA domain Theta domain TMA domain Theta domain Theta domain Complexity most complex intermediate least complex In this paper, we will pursue the hypothesis that the different types of complements can vary in minimal size as in (8). This does not mean that Event complements are nec- essarily always just theta domains (e.g., vPs). Structures could vary across languages, however, our main claim is that they can never go against the hierarchy. For instance, there could not be a language where Proposition complements are always less complex than Event or Situation complements. (8) a. Proposition b. Situation c. Event 3 3 3 believe Op domain decide TMA try 3 3 … TMA … 3 … Returning to the distribution of clause introducers in Bulgarian, the system outlined here can derive the ICH signature effect as well as the restriction noted about overt future. The assumption we make is that če is a true complementizer (i.e., a head in the operator domain), whereas da is a lower clausal head. Since Proposition complements require the operator domain, it follows that they always occur with če, as in (9a). Situa- tion complements can occur without the operator domain, in which case the embedded clause, a TMA domain, directly combines with the matrix verb. The irrealis interpreta- tion arises through a covert future modal woll (Wurmbrand 2014b; Todorović 2015), which, following Todorović and Wurmbrand (2020), needs to be identified within the syntactic context. One way to license woll is via Merge with a Situation verb as in (9b) (we leave open whether this is a selectional or featural relation). If, on the other hand, a Situation verb combines with an operator domain, (9c), the matrix verb and woll are too far apart, and only an independent overt future is possible (in which case woll is licensed by Tense (Abusch 1988)). The resulting future statement, although structurally more complex than a simple woll -clause, still satisfies the semantic requirement of a Situation verb, which only demands that the complement refers to a forward expanded unrealized event. Linguistica_2020_1_FINAL.indd 128 Linguistica_2020_1_FINAL.indd 128 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 129 (9) a. Proposition b. Situation c. Situation 3 3 3 believe Op domain decide WOLLP decide Op domain 3 3 3 če TMA WOLL če *WOLLP … da ✓FUT In the next section, we show how this implementation of the ICH derives the finite- ness universal in (7) (repeated in (10)). (10) (Hypothetical) Finiteness Universal If a language {allows/requires} finiteness in a type of complement, all types of complements further to the left on ICH also {allow/require} finiteness. 4.2 Finiteness in the SSL An account of the distribution of finiteness in the SSL in Table 4, repeated as Table 6, needs to derive the following properties: i) Proposition complements are always finite in the SSL (but not cross-linguistically). ii) Situation and Event complements can be finite or non-finite in the SSL, with language-specific settings restricting the options. iii) The distribution of finiteness follows the implicational universal in (7). Table 6: Finiteness in South Slavic Proposition Situation Event Bulgarian, Macedonian finite finite finite Serbian, Bosnian? finite (non-)finite (non-)finite Slovenian, Bosnian? finite (non-)finite non-finite Croatian finite non-finite non-finite Before answering these questions, we need to address one basic, yet very difficult question, namely what finiteness is. 4.2.1 What is Finiteness? Typically, finiteness is associated with a property of the temporal domain (but see Cris- tofaro 2007; Bisang 2007 for other options). Comparing the distribution of finiteness in Table 6 with the temporal properties of these complements in Table 7, it becomes clear that there is no general semantic tense property that can be mapped to a morphosyntac- tic category finite or non-finite in SSL. Linguistica_2020_1_FINAL.indd 129 Linguistica_2020_1_FINAL.indd 129 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 130 Table 7: Embedded tense properties Proposition Situation Event Embedded temporal domain yes yes no Predetermined tense value no yes N/A While it may be tempting to treat the uniformity of finiteness in Proposition comple- ments as a reflex of a semantic property (e.g., independent, non-predetermined tense; see also below), this would raise the question of why Situation complements can (Ser- bian, Bosnian, Slovenian) or must (Bulgarian, Macedonian) also be finite, and why even Event complements can (Serbian, Bosnian) or must (Bulgarian, Macedonian) be finite. It would also be difficult to extend such an account to languages outside the SSL, where Proposition complements can also be non-finite (e.g., English She claims to have won). Similarly, although Event complements tend to be non-finite, this cannot be the result of mapping the semantic lack of tense to a morphosyntactic category non-finite, since it is only a trend – Serbian and Bosnian allow, and Bulgarian, Macedonian require finite Event complements, where, importantly, the interpretation is the same as in Event con- texts in the other languages. Lastly, Situation complements also show the entire spectrum from obligatory finite (Bulgarian, Macedonian), optionally finite (Serbian, Bosnian, Slo- venian), to non-finite (Croatian), where, once again, the interpretation of the complement in all languages is the same, namely that of a forward expanded unrealized event. While we conclude that it is not possible to directly derive the distribution of finite- ness from semantic properties, we have seen in section 4.1 that the dependencies in Table 7 nevertheless play an important role in the distribution of complementation it- self. The different temporal properties reflect a scale of independence of the embedded clauses, which we have suggested is structurally implemented via different syntactic complexities. This approach allows us to derive the implicational nature of the ICH, and it will also provide an answer for the question of why the distribution of finiteness follows the implicational universal in (7). Returning to the question of what finiteness is, we follow Cristofaro (2007) and Bisang (2007), who argue that there is no universal category of finiteness, but that lan- guages can differ in what properties they utilize to express finiteness. We suggest that in the SSL, finiteness corresponds to agreement (whereas in other languages it could be tense, the combination thereof, or even other properties). More specifically, we follow Adger (2007), where it is suggested that features related to finiteness are not confined to a particular syntactic position (such as a Fin head in the CP), but can also occur on lower clausal heads. For instance, Adger suggests that subject licensing in Scottish Gaelic (a finiteness property) requires the (uninterpretable) features [T] and [Agr], but these features do not necessitate a semantic T or C head – they can also be inserted (somewhat parasitically) on other heads or project independent AgrPs in any domain of the clause. Although the details of finiteness in Scottish Gaelic and the SSL differ, we adopt this proposal in spirit and assume that in the SSL, finiteness is the spell-out of agreement features, which can occur on v, T, or C. Linguistica_2020_1_FINAL.indd 130 Linguistica_2020_1_FINAL.indd 130 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 131 4.2.2 Proposition Complements We are now able to derive the distribution of finiteness in complement clauses in SSL, starting with Proposition complements. We have observed that, whenever the operator domain must be projected, i.e., in Proposition complements (see Table 5), finiteness is obligatory (and this holds regardless of the finiteness settings of Event and Situation complements). We propose that in Proposition complements, the locus of finiteness is the CP in SSL. More specifically, as illustrated in (11), if the operator domain is projected in a complement clause, agreement features are obligatory and they occur on C or a split CP with a separate agreement projection (we do not split Agr in the trees). (11) Proposition 3 believe Op domain 3 C TMA Proposition [Agr] Finiteness in the CP is, of course, what one would expect following cartographic approaches. Our approach differs, however, in several respects. First, as we will see momentarily, finiteness is not restricted to the CP domain. In particular, we argue, following Adger (2007), that finiteness cannot entail the presence of a CP. Adger pro- poses that finiteness comes in two versions – a semantic notion of finiteness, associated with the CP, and occurrences of finiteness in the lower clausal domains which are not associated with a semantic function of finiteness. While we adopt the second part of Adger’s proposal, we submit that finiteness in complement clauses is never associated with a semantic property. Following Bianchi (2003), Adger suggests that a semantic finiteness head in the CP is responsible for identifying the embedded event/reference time [E/R] with the speech time [S] and/or the relation of participants to the external logophoric center, the external speech event. While this is a possible approach for main clauses, it does not carry over to embedded clauses. Bianchi (2003: 7) already qualifies the claim that “A finite verb form can encode the relation of E/R to S” with “at least in main clauses.” In complement clauses, the embedded tense is not related to the speech time, but always evaluated with respect to the matrix tense. Thus, even in Proposi- tion complements, which have their own temporal domain and no predetermined tense value, there is a tense dependency with the matrix clause and not the speech event. Furthermore, according to Bianchi (2003: 7), “A non-finite form does not encode any relation to S”. Since this is true for all types of complement clauses (none of them involve deictic tense), it would not allow us to distinguish between the different types of complements, and it therefore offers no way to approach the distribution in Table 6. The only option to tie finiteness in Proposition complements in the SSL to a seman- tic property is to associate it with the attitude property itself (following Kratzer 2006; Linguistica_2020_1_FINAL.indd 131 Linguistica_2020_1_FINAL.indd 131 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 132 Moulton 2009a,b). But since Proposition complements are not finite universally (see e.g., English, German), it once again cannot be a necessary connection between a se- mantic and morphosyntactic property. Our approach captures this – finiteness is not semantic but purely morphosyntactic. However, the distribution of this morphosyntac- tic property, implemented by the projection of agreement features, is sensitive to the structure, which in turn is determined by semantic properties. For the SSL we thus have the language-specific property in (12). (12) C [Proposition]: +Agr The last point to note is how an Agr head/feature in the CP triggers finiteness on the verb. There are various technical ways to derive this. For the sake of simplicity, we as- sume that the clausal heads C, T, v+V enter a dependency with each other (this could be V/T-movement, Agree, or feature lowering) and that the highest verbal element must realize the Agr feature. 4.2.3 Situation and Event Complements Since Situation and Event complements can lack the operator domain (see Progovac 1993a,b, 1994, 1996; Stjepanović 2004; Todorović and Wurmbrand 2020 for detailed motivation for clause reduction despite finiteness), under a cartographic approach where finiteness is located (solely or partially) in the CP, the question arising is how clauses without CP can be finite. Our account can successfully capture the presence of finiteness in the absence of CP as well as the distribution in Table 6. Following Adger (2007), we assume that agreement features, which are responsible for finiteness in the SSL, can be located in the other clausal domains, i.e., on heads of the TMA domain and even on heads of the theta domain. The structure of a finite Situation complement is given in (13a), the one of a finite Event complement in (13b). (13) a. Situation b. Event 3 3 decide wollP try vP 3 3 W O L L ... v ... [Agr] [Agr] The distribution of agreement features in the TMA and theta domains must be re- stricted by language-specific settings, which are spelled out in Table 8 for the SSL. Linguistica_2020_1_FINAL.indd 132 Linguistica_2020_1_FINAL.indd 132 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 133 Table 8: Agr distribution CP domain TMA domain Theta domain Bulgarian, Macedonian Agr Agr Agr Serbian, Bosnian? Agr optional Agr optional Agr Slovenian, Bosnian? Agr optional Agr no Agr Croatian Agr no/marginal Agr no Agr The approach thus captures the fact that Situation and Event complements can be finite or non-finite, even though Proposition complements must always be finite in the SSL. Since the finiteness of Proposition complements comes from the CP, and the CP can be missing in non-Proposition complements, it is correctly predicted that finiteness is only obligatory (in the SSL) when the operator domain is required. In all other types of complements, non-finite forms could be possible due to different settings of the lower clausal heads, which may or may not come with an Agr head. The approach thus allows us to derive the cross-linguistic variation – languages could differ in the inven- tory of Agr associated with heads of the different domains. The last advantage of this approach is that, despite the language-specific settings in Table 8, there are predictions it makes, which we turn to in the last subsection. 4.2.4 Finiteness Universal Recall that the distribution of finiteness in different types of complements is systematic and follows the finiteness universal – if a complement is allowed/required to be finite in a language, all the complements to the left of it allow/require finiteness. Thus, for instance, there is no language where Event complements are obligatorily finite and Situation and/or Proposition complements optionally non-finite; or no language where Event complements are optionally finite but Situation and/or Proposition obligatorily non-finite. Note that this does not mean that it is not possible for a language to realize finite Event complements and non-finite Situation complements at the same time – but what is impossible is that these be the only options in the language. Our approach derives these implicational relations in the distribution of finiteness. Since clausal domains are in a containment configuration (see Figure 2), it follows that settings in a lower domain affect all clauses that include that domain, i.e. also clauses with additional higher domains, since higher domains necessarily include the lower ones. To be more concrete, if in a language the theta domain is specified as (obligato- rily) finite (Bulgarian, Macedonian), all types of clauses will be realized as finite since the theta domain is included in all clause types. 4 In other words, if Event complements can or must be finite in a language, all other types of complements in the language can 4 The specification for Agr in Table 8 would thus not be necessary for the TMA and CP domains in these languages, since the Agr feature in the theta domain is sufficient to trigger finiteness in all types of clauses. We included it in the table for expository purposes, but also to leave open the option of double finiteness marking (see Todorović and Wurmbrand 2020). Linguistica_2020_1_FINAL.indd 133 Linguistica_2020_1_FINAL.indd 133 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 134 or must be finite as well. This is illustrated in (14) – finiteness as in (14a) entails finite- ness in configurations (14b,c). Although the Agr specifications for heads can differ in our system, it is not possible to derive a configuration in which, for instance, the theta domain contains Agr but clauses that include higher domains do not. Even if one were to posit Agr on v, but no Agr specifications on TMA or CP heads, the containment rela- tions predict that lower Agr specifications can never be ‘undone’. (14) a. vP b. TMA c. Op domain 3 3 3 v VP … vP … TMA [Agr] 3 3 v VP … vP [Agr] 3 v VP [Agr] 5 CONCLUSION We have shown in this paper that the distribution of (non-)finiteness in the SSL reflects an implicational scale along an independently attested semantic hierarchy. We have suggested that in the SSL, finiteness is triggered by clausal agreement features associ- ated with different syntactic heads. Building on a complexity approach to the comple- mentation hierarchy, cross-linguistic variation in finiteness, as well as variation across different types of complements is derived as the result of language-specific differences in the distribution of agreement features. More broadly, we have concluded that there is no (universal) semantic correlate of (non-)finiteness and, contra cartographic ap- proaches, that finiteness is not confined to a particular domain in the clause, but can be distributed over all clausal domains. Acknowledgements We thank the audience of the Workshop on the Morphology of South Slavic languages and an anonymous reviewer for invaluable feedback. This work has been supported by the Austrian Science Fund (FWF): M 2332-G30 (Universals and variation in clausal complementation) and P34012-G (Implicational hierarchies in clausal complementation). References ABUSCH, Dorit (1988) “Sequence of Tense, intensionality and scope.” In: H. Borer (ed), The Proceedings of the 7 th West Coast Conference on Formal Linguistics. Standford: CSLI Publications, 1–14. ABUSCH, Dorit (2004) “On the temporal compostion of infinitives.” In: J. Guéron/ J. Lecarme (eds), The syntax of time. Cambridge, MA: Massachusetts Institute of Technology, 27–53. Linguistica_2020_1_FINAL.indd 134 Linguistica_2020_1_FINAL.indd 134 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 135 ADGER, David (2007) “Three domains of finiteness: A minimalist perspective.” In: I. Nikolaeva (ed), 23–58. BIANCHI, Valentina (2003) “On finiteness as logophoric anchoring.” In: J. Guéron/L. Tasmovski (eds), Temps et point de vue/Tense and Point of View. Paris: Université Paris X. 213–246. BISANG, Walter (2007) “Categories that make finiteness: Discreteness from a func- tional perspective and some of its repercussions.” In: I. Nikolaeva (ed), 115–137. CHIERCHIA, Gennaro (1984) Topics in the syntax and semantics of infinitives and gerunds. Doctoral dissertation. Amherst: University of Massachusetts. CRISTOFARO, Sonia (2005) Subordination. Oxford: Oxford University Press. CRISTOFARO, Sonia (2007) “Deconstructing finiteness: Finiteness in a functional- typological perspective.” In: I. Nikolaeva (ed), 91–114. DIXON, Robert M. W (2006) “Complement clauses and complementation strategies in typological perspective.” In: R. M. W Dixon/A. Y. Aikhenvald (eds), Complemen- tation: A cross-linguistic typology. Oxford: Oxford University Press, 1–48. GIVÓN, Talmy (1980) “The binding hierarchy and the typology of complements.” Studies in Language 4/3, 333–377. https://doi.org/10.1075/sl.4.3.03giv GIVÓN, Talmy (1990) Syntax: A Functional-Typological Introduction, Vol. II. Am- sterdam: J. Benjamins. GROHMANN, Kleanthes K. (2003) Prolific domains: On the anti-locality of move- ment dependencies. Amsterdam/Philadelphia: John Benjamins. HACQUARD, Valentine (2006) Aspects of modality. Doctoral dissertation. Cam- bridge, MA: Massachusetts Institute of Technology. HORIE, Kaoru (2001) “Complement clauses.” Language Typology and Language Uni- versals. An International Handbook 2, 979–993. KRATZER, Angelika (2006) “Decomposing Proposition verbs.” Talk given in honor of Anita Mittwoch. The Hebrew University of Jerusalem. MAAS, Utz (2004) “‘Finite’ and ‘nonfinite’ from a typological perspective.” Linguis- tics 42/2, 359–385. https://doi.org/10.1515/ling.2004.012 MOULTON, Keir (2009a) “Clausal complementation and the wager-class.” In: A. Schardl/M. Walkow/M. Abdurrahman (eds), Proceedings of the North Eastern Linguistics Society Annual Meeting 38. Amherst: University of Massachusetts, GLSA,165–178. MOULTON, Keir (2009b) Natural selection and the syntax of clausal complementa- tion. Doctoral dissertation. Amherst: University of Massachusetts. NIKOLAEV A, Irina (2007a) “Constructional economy and nonfinite independent clauses.” In: I. Nikolaeva (ed), 138–180. NIKOLAEV A, Irina (ed) (2007b) Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press. NIKOLAEV A, Irina (2007c) “Introduction.” In: I. Nikolaeva (ed), 1–19. NOONAN, Michael (2007) “Complementation.” In: T. Shopen (ed), Language typol- ogy and syntactic description: Complex Constructions (Volume II). Cambridge: Cambridge University Press, 52–150. Linguistica_2020_1_FINAL.indd 135 Linguistica_2020_1_FINAL.indd 135 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 136 PESETSKY, David (1992) Zero syntax II: An essay on infinitives. Cambridge, MA: MIT. PROGOV AC, Ljiljana (1993a) “Locality and subjunctive-like complements in Serbo- Croatian.” Journal of Slavic Linguistics 1, 116–144. PROGOV AC, Ljiljana (1993b) “Subjunctive: The (mis)behavior of anaphora and negative polarity.” The Linguistic Review 10, 37–59. https://doi.org/10.1515/ tlir.1993.10.1.37 PROGOV AC, Ljiljana (1994) Negative and positive polarity: A binding approach. Cambridge: Cambridge University Press. PROGOV AC, Ljiljana (1996) “Clitics in Serbian/Croatian: Comp as the second posi- tion.” In: A. Halpern/A. Zwicky (eds), Approaching second: Second position clitics and related phenomena. Stanford: CSLI Publications, 441–428. RAMCHAND, Gillian/Peter SVENONIUS (2014) “Deriving the functional hierarchy.” Language sciences 46, 152–174. https://doi.org/10.1016/j.langsci.2014.06.013 ROCHETTE, Anne (1988) Semantic and syntactic aspects of Romance sentential com- plementation. Doctoral dissertation. Cambridge, MA: Massachusetts Institute of Technology. ROCHETTE, Anne (1990) “On the restructuring classes of verbs in Romance.” In: A. Di Sciullo/A. Rochette (eds), Binding in Romance: Essays in honour of Judith McA’Nulty. Ottawa: Canadian Linguistic Association, 96 –128. STJEPANOVIĆ, Sandra (2004) “Clitic climbing and restructuring with ‘finite clause’ and infinitive complements.” Journal of Slavic Linguistics 12/1–2, 173–212. TODOROVIĆ, Neda (2015) “Tense and aspect (in)compatibility in Serbian ma- trix and subordinate clauses.” Lingua 167, 82–111. https://doi.org/10.1016/j. lingua.2015.04.007 TODOROVIĆ, Neda/Susi WURMBRAND (2020) “Finiteness across domains.” In: T. Radeva-Bork/P. Kosta (eds), Current Developments in Slavic Linguistics. Twenty Years After (based on selected papers from FDSL 11). Frankfurt am Main: Peter Lang, 47 –66. VRZIĆ, Zvjezdana (1996) “Categorial status of the Serbo-Croatian ‘modal’ da.” In: J. Toman (ed), Formal Approaches to Slavic Linguistics: The College Park Meeting 1994. Ann Arbor: Michigan Slavic Publications, 291–312. WURMBRAND, Susi (2002) “Syntactic versus semantic control.” In: C. J. Zwart/W. Abraham (eds), Studies in comparative Germanic syntax. Amsterdam: John Benja- mins, 95–129. WURMBRAND, Susi (2014a) “Restructuring across the world.” In: L. Veselovská/M. Janebová (eds), Complex Visibles Out There. Proceedings of the Olomouc Linguis- tics Colloquium 2014: Language Use and Linguistic Structure. Olomouc: Palacký University, 275–294. WURMBRAND, Susi (2014b) “Tense and aspect in English infinitives.” Linguistic Inquiry 45/3, 403–447. https://doi.org/10.1162/LING_a_00161 WURMBRAND, Susi (2015) “Restructuring cross-linguistically.” In: T. Bui/D. Özyıldız (eds), Proceedings of the North Eastern Linguistics Society Annual Meet- ing 45. Amherst: University of Massachusetts, GLSA, 227–240. Linguistica_2020_1_FINAL.indd 136 Linguistica_2020_1_FINAL.indd 136 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39 137 WURMBRAND, Susi/Magdalena LOHNINGER (to appear) “An implicational uni- versal in complementation – Theoretical insights and empirical progress.” In: J. M. Hartmann/A. Wöllstein (eds), Propositional Arguments in Cross-Linguistic Re- search: Theoretical and Empirical Issues. Berlin: Mouton de Gruyter. Abstract FINITENESS IN SOUTH SLA VIC COMPLEMENT CLAUSES: EVIDENCE FOR AN IMPLICATIONAL FINITENESS UNIVERSAL This paper shows that the distribution of (non-)finiteness in the South Slavic lan- guages reflects an implicational scale along an independently attested semantic comple- mentation hierarchy (e.g., Givón 1980). We suggest that in the South Slavic languages, finiteness is triggered by clausal agreement features associated with different syntactic heads. Building on a complexity approach to the complementation hierarchy, we pro- pose that cross-linguistic variation in finiteness and variation across different types of complements are the result of language-specific differences in the distribution of agreement features. More broadly, we conclude that there is no (universal) semantic correlate of (non-)finiteness and, contra cartographic approaches, that finiteness is not confined to a particular domain in the clause. Following Adger (2007), we argue that finiteness can be distributed over all clausal domains. Keywords: syntax, morphology, language variation, implicational hierarchies, com- plementation, infinitives, finiteness, South Slavic Povzetek OSEBNE IN NEOSEBNE GLAGOLSKE OBLIKE V JUŽNOSLOV ANSKIH DOPOLNILIH: DOKAZI ZA IMPLIKACIJSKO UNIVERZALIJO V članku pokažemo, da porazdelitev osebnih in neosebnih glagolskih oblik v južno- slovanskih jezikih odraža implikacijsko lestvico po neodvisno potrjeni semantični hie- rarhiji dopolnil (npr. Givón 1980). Zagovarjamo trditev, da v južnoslovanskih jezikih pojavnost (ne)osebnih glagolskih oblik sprožajo stavčne ujemalne oznake, povezane z različnimi skladenjskimi jedri. Na podlagi pristopa kompleksnosti k hierarhiji dopolnil predlagamo, da so medjezikovne razlike v pojavnosti (ne)osebnih glagolskih oblik in razlike med vrstami dopolnil rezultat jezikovno specifičnih razlik v porazdelitvi uje- malnih oznak. V splošnem zaključujemo, da ne obstaja (univerzalni) pomenski korelat pojavnosti (ne)osebnih glagolskih oblik in da, v nasprotju s kartografskim pristopom, ta lastnost ni omejena na specifično domeno v stavku. Podobno kot Adger (2007) trdimo, da se lastnost (ne)osebnih glagolskih oblik lahko porazdeli po vseh stavčnih domenah. Ključne besede: skladnja, morfologija, jezikovna variacija, implikacijske hierarhije, dopolnila, nedoločniki, (ne)osebne glagolske oblike, južnoslovanski jeziki Linguistica_2020_1_FINAL.indd 137 Linguistica_2020_1_FINAL.indd 137 22. 12. 2020 15:07:39 22. 12. 2020 15:07:39