3.1 The role of the state goal

In addition to being characterized by speaker-oriented deixis, BFSI constructions with speaker-oriented verbs of inherently directed motion would seem to be incompatible with a sentence-final locative adverbial (Benincà 1988; Calabrese 1992; Sheehan 2006: 148). In (19a–b), the sentence-final locative expression forces the preverbal position of the subject.

To be sure, the acceptability of (19a) improves considerably if the postverbal DP is indefinite.

The Definiteness Effect in (19a) is not found in BFSI constructions with other verbs of change. In fact, (21a) is the only option as an out of the blue announcement, whereas (21b) would be felicitous in a context that presupposes the preverbal subject or if the sentence-final location is in narrow focus (e.g. for contrastive purposes).

Assuming that the locative-g is realized as a null speaker-oriented preverbal argument in (19a), the Definiteness Effect observed in this example could ultimately be the consequence of the θ-criterion, i.e., the condition by which the same thematic role (in this case, location) cannot be assigned twice. The same problem does not arise in (20) or (21a), where, we propose, BFSI is not licensed by a null locative.

Importantly, an overt locative phrase can follow the DP in a construction with narrow focus on the definite DP and a dislocated locative phrase (cf. (22B)), or with narrow focus on the PP itself (cf. (23B)). These are not BFSI constructions, and, hence, they do not require a preverbal SoP. (We indicate narrow focus with small caps; the comma separating the focal DP from the following PP in (22B) indicates a prosodic boundary.)

The hypothesis that BFSI is licensed in different ways with different verbs of change is supported by evidence from the dialect of Borgomanero (Tortora 1997; 2014), which was mentioned above. In this dialect, a locative clitic sequence ngh… gghi occurs in BFSI constructions with speaker-oriented verbs of inherently directed motion (cf. (24a)). The ngh… gghi sequence is claimed by Tortora (1997: 58) to be in a clitic-doubling relation with a covert preverbal locative goal argument. There is no counterpart of this clitic sequence in VS constructions with other verbs of change (cf. (24b vs. 24c)), which must be licensed differently.

BFSI constructions are to some extent sensitive to sentential aspect (Bentley 2006: 378; Leonetti 2017: 896), in that they are hardly compatible with the simple present tense (cf. (25a–c)). However, this restriction does not apply when the verb is a speaker-oriented verb of inherently directed motion (cf. (26a–c)).

The difference between (25a–c) and (26a–c) is evidenced in genuinely out-of-the blue contexts, for example if a speaker describes or announces what s/he sees at the time of the utterance. In neither case should there be any understood presupposition. For most native speakers, in these contexts the simple present tense in the examples in (26) has a progressive aspectual value.⁵ By contrast, the same aspectual value can hardly be assigned to the examples in (25) because, we claim, too much must be inferred: not only the progressive aspect, but also the situation that allows BFSI with these verbs. Overt marking for progressive aspect with the examples in (25a–c) is in principle acceptable, as shown in (27), but these sentences are barely compatible with the out-of-the-blue contexts typically licensing BFSI, since in the progressive these examples describe ongoing eventualities, rather than bounded eventualities predicated of an inferred situation. For this reason, SV order would be more natural in the examples in (27).

The above facts are reminiscent of Bianchi’s (1993) claim that an eventive sentence describes a specific event, as opposed to an atemporal situation as in (25a–c) (see §2). In addition, if, as we tentatively assumed, the state-g could be projected as an argument and serve as SoP in BFSI, the contrast between (25a–c) and (26a–c) would remain unexplained, since a specific state-g is entailed in the event structure of verbs of quantized change regardless of sentential aspect. In other words, there would be no principled explanation for the failure of the state-g to serve as SoP in (25a–c). The contrast between (25a–c) and (26a–c), therefore, would seem to suggest that only with verbs of inherently directed motion can BFSI be licensed by an argument of the verb, which serves as SoP regardless of sentential aspect or any other contextual clues.

On the basis of the contrast between the location-g, which can be realized in syntax as a thematic argument selected by the verb, and the state-g, which cannot (cf. (18a–b)), and in light of the further evidence adduced in this section, we revise our initial hypothesis on the role of the state-g in BFSI. In particular, we propose that, with verbs of quantized change lacking a thematic goal argument, BFSI is not licensed by a component of the argument structure of the verb, but rather by a situational argument, which arises in all-new discourse contexts, when a specific bounded eventuality is predicated. The acceptability of (20) suggests that BFSI can also be licensed by this argument with verbs of inherently directed motion.

This argument provides the situation about which the bounded eventuality is predicated and against which the truth-value of the proposition is evaluated. It is thus comparable with Klein’s (2008) topic situation, although it must be understood that the situational argument not given, but rather introduced with the utterance itself. Following Bianchi (1993), we represent (28a) as (28b), where e is the situational argument.

To flesh out our hypothesis that a situational argument can arise with the utterance, we start from the observation that we are dealing with structures that are not articulated as subject/predicate (or topic/comment) oppositions, but rather as fully predicative – or thetic – sentences (see Kuroda 1972; Sasse 1987; Ladusaw 1994; a. o., and, for the related notion of presentational construction, Calabrese 1992; Lambrecht 1994; 2000). Together with the verb, the theme argument is part of the predication (Fuchs 1980: 449; cited in Sasse 2006: 258; Bianchi 1993; Zeller 2013; Bentley 2018). In such structures, the starting point of the predication – or the argument which the predication is about – must be identified (recall Bianchi’s Principle of Nonvacuous Predication).⁶ If a goal argument is available in the thematic grid of the verb, it can provide the information that is needed as the starting point of the predication (§3.3). If a thematic goal argument is not available, the situation about which the event is predicated has to be inferred. The situational argument thus arises inferentially with the utterance.

Given that the state-g does not serve as SoP, the question that arises at this point is whether the entailment of a specific state-g in event structure is at all relevant to the licensing of BFSI. In other words, could a situational argument be inferred with any verb of change? Bianchi (1993) rules out the structure in (28b) for state predicates that are not stage-level states. Of the two stative predications in (29a–b), only (29b) is a stage-level state and, indeed, it exhibits subject inversion in broad focus.

On the other hand, Bianchi does not compare different types of verb of change. However, even a cursory look at verbs of non-quantized change (Beavers 2011) – of which degree achievements (Dowty 1979: 88–90) or gradual-completion verbs (Bertinetto & Squartini 1995: 23) are a subclass – suggests that the entailment of a specific state-g is relevant to the licensing of BFSI. Indeed, the utterances in (30a–b) would normally only be felicitous in a discourse context where the subject alone is in focus.

Observe in passing that the definiteness of the postverbal DP is irrelevant to the acceptability of these examples: even if this is indefinite, the examples are not more easily interpretable as broad-focus structures. What improves their acceptability is the addition of a previous context (this year the kids have grown…, at the party the kids got bored…). In other words, these examples need the addition of an overt or understood aboutness topic to be acceptable.

Degree achievements are compatible with both in and for temporal adverbials (Dowty 1979: 89–90).

Other diagnostics to identify these verbs are their being entailed by their progressive forms (cf. (32a–b vs. 17a–b)) and their compatibility with adverbials describing the extent of the change (cf. (33a–b)).

Beavers (2011) claims that verbs of non-quantized change only entail that a goal exists. Unlike verbs of quantized change, though, they do not entail a specific final goal on the relevant scale of change. Given that verbs of quantized and non-quantized change contrast with respect to their interpretation in VS order, we conclude that the specific state-g entailed by the former class does play a role in the licensing of BFSI. We claim that the situation about which the event is predicated can only be inferred if the structure of the event is bounded. With verbs of quantized change, the boundedness requirement is satisfied by the attainment of the state-g. In the next section, we shall discuss how BFSI is licensed with some verbs of non-quantized change, contrary to initial expectations.

3.2 Inferred state-goal

A subclass of degree achievements turns out to be as compatible with BFSI as verbs of quantized change.

These verbs test out as degree achievements in terms of their compatibility with adverbials describing the extent of the change (cf. (35a–b)). However, their perfect is not entailed by the progressive form (cf. (36a–b)).

Hay, Kennedy & Levin (1999) claim that, in this class, a bounded difference value is inferred on the basis of the adjectival base of the verbs. They call difference value the measure of the amount to which an argument of the verb changes with respect to a gradable property. The adjectival base is a closed-range adjective, i.e., an adjective associated with a property that has a maximum value, as witness the compatibility of English empty (adj.) with completely. The same compatibility characterizes the verb empty.

The bounded difference value that is inferred with the verb empty explains its telic behaviour (cf. (38)). Importantly, the bounded measure of the change is not entailed, but rather only inferred on the basis of a conversational implicature, which can be cancelled (cf. (39)).

On a par with English empty (v.), the verbs in (35a–c) and (36a–c) are compatible with both completely and not completely.

By contrast, verbs of quantized change do not combine with the same adverbials. Since the attainment of a stage goal is entailed, it is not an inference and it cannot be cancelled by not completely. On the other hand, the combination with completely is odd because it is redundant (Hay, Kenney & Levin 1999).

The contrast between (40a–c) and (41a–d) indicates that the verbs in the former group are not verbs of quantized change (cf. also (35a–c)). The facts in (40a–c), however, suggest that these verbs can be telic by virtue of an inference of bounded change. We thus propose that, even though these verbs do not entail a specific goal on a scale of change, a state goal (state-g) can be inferred on the basis of the scalar structure of a gradable property with which they are associated by means of their adjectival base. The key feature of the said scalar structure is that it is compatible with the inference of a maximum value. The inferred state-g satisfies the requirement that the eventuality be bounded for a situational argument to be inferred and BFSI to be licensed (cf. (34a–c)).

With degree achievements formed from open-range adjectives, i.e., adjectives associated with a property for which it is not possible to identify a maximum value on a scale, a goal is never actually reached, but it is approximated in an asymptotic way (Cruse 1986: 206). According to Hay, Kennedy & Levin (1999), the incompatibility with completely indicates, in this case, that the scalar structure of the relevant gradable property (cf. (42a–c)), and the related scale of change (cf. (43a–c)), are incompatible with a maximum value.

Nonetheless, Hay, Kenney & Levin (1999) note that a maximum value may be associated with a conventional property of the entity denoted by the theme argument: the maximum/minimum conventional length of a skirt, width of a road, etc.

Alternatively, there may be other contextual clues that give rise to the conversational implicature of a maximum value on a scale of change; for example, the maximum width of a pothole over which a car can pass with no damage to the wheels.

Therefore, the degree achievements under discussion can also have an inferred state-g, even though this arises from contextual clues. If there is such an inference, the event is bounded and a situational argument can be inferred, with BFSI being licensed as a result.

We note in passing that there are at least two non-deadjectival degree achievement verbs which admit an inference of bounded change on the basis of contextual clues: aumentare (intr.) ‘increase’ and diminuire (intr.) ‘decrease’.⁷ We suggest that BFSI (cf. (47c, 48c)) is licensed with these verbs if the said inference arises and thus a situational argument is also inferred.

Turning now to the degree achievements that are hard to find in BFSI (cf. (9a–b, 30a–b, 49a–b)), observe that they do not combine with completely (cf. (50a–b)), which suggests they are not associated with a scale of change, or a property, with a maximum value.

In this respect, the degree achievements in question behave like those based on open-range adjectives (cf. (43a–b)). In the absence of an inference of bounded change, BFSI is awkward. The question that arises is why an inference of bounded change does not arise on the basis of contextual clues. Our answer to this question is that such an inference can in principle arise, but the constant in the event of change described by these verbs, and hence the types of theme argument that such verbs select, do not lend themselves easily to it.

The conventional length of a skirt may be determined by a fashion, or a particular style, or the occasion in which it is worn. The width of a road may depend on the amount of traffic it has to cater for. Similar maximum values are not normally associated with the changes described by psych-verbs and verbs describing physical change, or with their experiencer/theme arguments. However, if, contrary to expectations, a maximum value is associated with the relevant event of change, or with a property of the argument, even these verbs allow the inference of a goal and, as a result, they can occur in BFSI. Imagine that a conventional height must be reached for a child to be admitted to a basketball training club. In this case, the BFSI structure in (51) would not be odd, at least for some speakers, although the counterpart with SV order would also be felicitous.⁸

Similarly, in particular situations, a maximum value may be associated with a psychological state. Imagine that a family with little children has been invited to someone’s house and the parents are planning to leave as soon as the children seem bored enough to start behaving badly. In this context, the utterance in (52) would be felicitous.

The examples in (51) and (52) are announcements, stating that a conventional or predetermined goal has been reached in an event of growth and of psychological change, respectively. This bounded event structure is compatible with the inference of a situational argument and, hence, with BFSI. The addition of sentence-initial adverbs such as finalmente ‘finally/eventually’ or adesso ‘now’, highlighting the fact that the long-awaited goal has been reached, facilitates the bounded interpretation, although in the presence of such adverbs in sentence-initial position the relevant structures do not exemplify bare subject inversion.

3.3 Thematic Subject of Predication

We now return to verbs of inherently directed motion. Recall that these are verbs of quantized change, entailing a specific goal location on a scale of change (Beavers 2011).

The lexical-aspectual properties of this class of verbs lend themselves to the inference of a situational argument. However, on the basis of the default deictic interpretation of BFSI (cf. (54a)), as well the Definiteness Effect illustrated in (54b) (cf. (19a vs. 20)), we suggested that the goal location of these verbs can provide the starting point of the predication, or what the predication is about, i.e., in our terms, the SoP. In agreement with Pinto (1997), we suggest that the deixis of the goal location points to the centre of the utterance situation (Vanelli 1972), namely the speaker. We represent (54a) as in (54a′).

The proposal in (54a′) does not suit the verbs of inherently directed motion which, as was mentioned in passing at the beginning of Section 3, are lexically specified for movement away from a location: partire ‘leave’, uscire ‘go/come out’, etc. A key difference between these verbs and those that we labelled speaker-oriented is that they can exhibit a source, not a goal, location (cf. (55a)). In addition, BFSI constructions with these verbs are compatible with an overt sentence final location (cf. (55b)) and sensitive to sentential aspect (cf. (55c)). Therefore, these verbs behave like the verbs of quantized change discussed in Section 3.1.⁹

If the locative argument of the verb in (55b) served as SoP, we would expect this structure to be incompatible with an overt sentence final locative phrase. In turn, the simple present tense should be acceptable in (55c). Therefore, we assume that, with the verbs that are lexically specified for movement away from a location, the predication is not about a location, but rather about a situational argument.¹⁰ As is the case with the verbs of quantized change discussed in Section 3.1, this argument can be inferred because these verbs are telic, that is, they include a specific state-g in their event structure. Relevant evidence is provided in (56a–b), indicating that the perfect of partire ‘leave’ and uscire ‘go out’ is not entailed by the progressive forms.

The aspectual property tested in (56) is also shared by speaker-oriented verbs of inherently directed motion (e.g. arrivare ‘arrive’, venire ‘come’, cf. (17a)). The key difference between the two verb classes, therefore, depends on whether the location is a goal or a source. The reason why a source argument, differently from a goal argument, does not lend itself to the role of silent SoP must lie in the different roles that these arguments play in predicate composition. The goal is the final state in the predication, whereas this is not the case with the source. This difference, in turn, raises the issue of the status of the goal as part of the constant, or a templatic component of the predicate, or an actant argument. Given that the goal can be realized as an argument in syntax (cf. (18b, 53a–b)), we have considered it to be an argument. This important issue, however, requires further investigation independently of the study of BFSI.

Incidentally, our hypothesis on non-speaker oriented verbs of inherently directed motion is supported by the lack of the clitic sequence ngh… gghi in Borgomanererese BFSI constructions with these verbs, in contrast with their counterparts with the speaker-oriented verbs (cf. (24)).

To sum up, we have proposed that, with verbs of inherently directed motion which are lexically specified for movement away from a location, BFSI is licensed by a situational argument SoP, which can be inferred because these verbs are telic.

We conclude this section with a discussion of the few activity verbs that can be found in genuine BFSI, i.e., in VS constructions that neither exhibit nor presuppose a topic. Relevant examples are given here.

The conventional tests for activities include the compatibility with temporal for-adverbials and the incompatibility with temporal in-adverbials.

Since Dowty (1979: 60–62) it has been known, however, that activities can have accomplishment readings, i.e., readings as bounded events, which are not lexically coded, but rather arise from contextual knowledge or clues. To give but one example, phone can describe the bounded event of making a phone call. In the bounded reading, conveyed below by the addition of a sua madre, lit. ‘to his mother’, telefonare does combine with the temporal in-phrase, as well as with the periphrasis has finished V-ing, where finish indicates completion, rather than interruption.

Similarly to phone, other activities can behave as accomplishments. Thus, the examples in (61a) describe unbounded eventualities, whereas those in (61b–c), which would at first be judged to be ungrammatical (cf. (59)), can be read as bounded events by virtue of knowledge that is not encoded in the clause. For example, in the case of nuotare ‘swim’, the bounded reading only makes sense if one knows that Gianni is in the habit of swimming a certain length a day.

Importantly, the verbs in (61a–c) are not nearly as compatible with genuine BFSI as those in (58a–c). Unless a spatial or temporal topic is overtly or covertly added (At the party…, Last night…), the only natural reading of the sentences in (62a–c) involves narrow focus on the postverbal DP.

The contrast between (58a–c) and (62a–c) suggests that the possibility of an accomplishment reading cannot alone explain the occurrence of some activities in BFSI.

With specific respect to bussare ‘knock’ and chiamare ‘call’, we note that they are not mere activities, but rather activity-based semelfactives, describing repeated events of knocking and calling (Smith 1991: 55–57; Van Valin 2005: 42–45). Semelfactives are known to have punctual readings (Beavers 2013: 682), which should in principle make them compatible with the inference of a situational argument SoP. This expectation is not fully borne out, though, as we show here.

A distinctive property of all the verbs in (58a–c) is that they have a locative or benefactive goal argument in their thematic grid. (The second argument of chiamare ‘call’ can be realized as a transitive object, a point which is irrelevant here, see Note 11 and §4.3).

In fact, the deixis of the BFSI constructions in (65a–c) (cf. (58a–c)) is speaker-oriented, and these examples are fairly incompatible with an overt locative phrase in sentence-final position. The same Definiteness Effect applies as with speaker-oriented verbs of inherently directed motion (cf. (19a vs. 20)).

The obvious explanation for the results in (65a–c) is that, similarly to the BFSI constructions with speaker-oriented verbs of inherently directed motion, the constructions in (58a–c) and (65a–c) have a silent locative SoP, which licenses BFSI. The silent SoP is the locative or benefactive goal argument of the relevant verbs, which takes speaker-oriented deixis in the utterance situation. The constructions can thus be represented as follows.

The constructions in (66a–c) can, instead, be assumed to be licensed by a situational argument SoP, which is inferred by virtue of the bounded reading admitted by the verbs under examination and conveyed here by the overt locative goal arguments.

The generalization that can be drawn on the basis of our findings is that BFSI is not admitted with activities, which are unbounded, unless they also select a locative or benefactive goal argument, which plays a key role in the inference of a bounded reading and can serve as SoP. The presence of this goal explains the difference between the verbs in (58a–c) and other activities (cf. (62a–c, 63a–b)).¹¹

3.4 The Subject of Predication of bounded states

We now return to the predicates upon which Bianchi’s (1993) event-argument approach is based, namely stage-level states (see §2). To begin with, we note that most adjectives describing stage-level states also allow individual-level readings, which require the argument DP to have wide scope and occur preverbally (Bianchi 1993; Bianchi & Chesi 2014: 533). For this reason the VS construction with a stage-level state is only grammatical when it is clear that the postverbal DP takes low scope (cf. (67a)), like the postcopular DP of an existential there sentence (Heim 1987). The adverb solo ‘only’ conveys precisely this reading to the construction (Beaver et al. 2005) and this is why (67a) is grammatical. Unambiguous individual-level states are rejected in VS order (cf. (67b)).

Stage-level states are also found in the coda of existential there-sentences (Milsark 1974; 1979), which deserve some discussion here.

Existential there-sentences are structures with non-canonical morphosyntax, which express a proposition about existence or presence in a context (McNally 2011; Bentley, Ciconte & Cruschina 2015). Traditionally, they have been related to locatives (Lyons 1967; Clark 1978; Creissels 2014). In fact, according to Freeze (1992), they have the same initial underlying structure as locative predications; the VS order emerges whenever the locative predicate, in the form of either a PP or a proform, moves to the subject position (see also Moro 1997). This view has been challenged by those claiming that the existential predicate is neither the proform nor an optional locative phrase, but rather the postcopular DP, conventionally referred to as the pivot (Williams 1994; Hazout 2004; Francez 2007; Cornilescu 2009). In support of this analysis, Francez (2007) notes that the pivot is the only obligatory component of the existential construction cross-linguistically.

Assuming that the pivot is the predicate, the question must be addressed of what the subject or the argument of the existential predicate is. An answer to this question would be to consider the proform itself to be an expletive subject (see Burzio 1986; Williams 1994; Hazout 2004). However, this idea may be tenable for English, where the proform there figures in subject position, and inverts with the finite copula in interrogatives, but it does not suit other languages, where the proform cooccurs with an invariant subject pronoun (e.g. French y, which cooccurs with il, see Bentley & Ciconte 2016; Bentley & Cruschina 2016).

An alternative proposal comes from Francez (2007; 2010), who pursues the hypothesis that the pivot is the predicate of a contextual domain variable, which he calls an implicit argument. Key evidence for this variable comes from the truth conditions on existential sentences. Consider the examples below.

The context-dependent existential there-sentence in (69a) clearly means something different from the context-independent sentence in (69b). Crucially, the truth-value of the former depends on contextual information. This supports the view that a defining component of this construction is a contextual domain. As was mentioned, the existential pivot takes narrow scope and has no presupposition of existence. This is not true of the subject of (69b), il caffè ‘coffee’, which has wide scope and is presupposed to exist.

Importantly, existential constructions range from propositions expressing presence in a specific, albeit often understood, location (cf. (69a, 70a)) to propositions expressing existence in a much broader sense (cf. (70b)). Koch (2012) calls the former type existentials of bounded existence and the latter existentials of generic existence.

Francez’s (2007) contextual domain variable must, therefore, be understood to be an abstract argument, which can – but need not – be enriched by an overt modifier contributing to its identification (Francez 2009). The phrase in Africa in (70a) is a case in point.

Following Cruschina (2012), we claim that the existential proform (ci) is the spell-out of the contextual domain variable. The locative phrase in (70a), on the other hand, can be considered to be the SoP of the construction. Thus, in (70a) the presence or existence of white lions is situated in Africa. In the absence of a locative phrase, existentials could be claimed to have a silent situational SoP, which provides the situation about which they are predicated. While being entirely consistent with the semantics of existentials, which are states bounded by a contextual domain, this hypothesis is not economical. More to the point, the contextual domain variable is sufficient to provide the situational coordinates of the predication. Therefore, we opt for the hypothesis that the variable identifying or restricting the contextual domain serves itself as SoP. A key difference between the BFSI constructions discussed in previous sections, on the one hand, and existentials, on the other, is that the contextual domain variable is an obligatory component of the existential construction. In fact, existentials have obligatory VS order and do not allow SV counterparts (see Bentley, Ciconte & Cruschina 2015: 139 for some apparent exceptions). Thus, strictly speaking, they are not subject inversion constructions, even though the pivot may exhibit patterns of subject behaviour (see Bentley 2013; 2017 for an in-depth discussions of this point).

Returning now to stage-level states, the existential there sentence constitutes a syntactic environment in which the DP takes narrow scope and the adjective in the coda can be read as a stage-level state, as is shown in (68).¹²

3.5 Typology and predictions

We have constrained our analysis of subject inversion to out of the blue announcements (i.e., BFSI), where the verb and the postverbal DP are discourse new, and there is no overt or covert topic. In these constructions, the need arises to identify the starting point of the predication or the situation about which the event is predicated (Klein 2008). We have proposed that if a thematic goal argument is available in the thematic grid of the verb, it can provide the information that is needed as the starting point of the predication. In this case, the aboutness requirement is satisfied by a thematic SoP. If a thematic goal argument is not available, the situation about which the event is predicated must be inferred. A non-thematic argument thus arises inferentially with the utterance, providing the situation about which the event is predicated. We have called this a situational SoP.

We have identified two verb classes which can take a thematic SoP in BFSI: speaker-oriented verbs of inherently directed motion (arrive, enter, come in, return, etc.) and a small group of activity verbs and activity-based semelfactives (phone, knock, call). Apart from the locative goal of these verbs, we have not identified any other thematic arguments that serve as silent SoPs in BFSI. Although the search for such arguments remains an issue for future research, we have noted that the two subtypes identified in our analysis contribute to the boundedness of the event. Source arguments do not play this role in event structure and we have claimed that they cannot serve as SoPs. We thus propose that thematic SoPs must be goal arguments.

The situational SoP is not part of the thematic grid of the verb. In fact, it is not selected by the verb, but rather inferred in the utterance situation. However, this SoP is sensitive to the event structure of the predicate. Specifically, it is only inferred when a bounded eventuality is predicated. The bounded eventualities that we have identified are events of quantized change (see §3.1) and events of non-quantized change for which a bound is inferred (see 3.2). In addition, we have claimed that existential constructions are bounded stative constructions, where an implicit contextual domain variable serves as SoP. The predicate classes identified and the relevant SoPs are illustrated and exemplified in Table 1.

We can now spell out our predictions on the acceptability of BFSI. Aside from existential constructions, which have obligatory VS order and an obligatory contextual domain variable serving as SoP, if a language allows BFSI at all, it will allow it with verbs of quantized change. These include verbs with a locative goal, which can be projected as an expressed argument in syntax, or take on the role of silent SoP, and verbs with a state goal, but no goal argument. A situational SoP is compatible with the latter type because of their inherent telicity. With verbs of non-quantized change, a bound must be inferred in order for an SoP to be added situationally. Therefore, those among these verbs that are based on closed-range adjectives lend themselves more readily to BFSI than those that are based on open-range adjectives. Nonetheless, a number of contextual clues can arise from, say, properties of the theme DP and provide the necessary bound. As a result, a situational argument is inferred and BFSI is licensed. In general terms, activities and states are not found in genuine BFSI because they are inherently non-scalar and unbounded and, thus, they are not readily compatible with the inference of a situational SoP. However, the activities (and semelfactives) that have a locative or benefactive goal argument can be admitted in BFSI. We have suggested that the goal plays a key role in an inference of boundedness and hence in the licensing of BFSI.

It is our contention that these predictions are valid in those SVO languages in which VS order is not entirely free. In his study of word order in Romance, Leonetti (2017) claims that subject inversion is not free in all null-subject languages. Rather, null-subject languages exhibit different degrees of flexibility in allowing broad-focus interpretations of verb initial structures. Thus, Italian and Catalan only allow broad focus in VS sentences, but not in VOS or VSO sentences, whereas Spanish admits broad focus in VS and VOS/VSO alike. From this point of view, Spanish is closer to Latin than Italian and Catalan. As a result of diachronic change, the last two languages exhibit stricter associations between word order and specific information structure patterns.¹³ Our study indicates that the information-structure restrictions discussed by Leonetti are inextricably related to event structure constraints. Thus, BFSI is not limited to intransitive (VS) structures as such, but rather to the verb initial structures that allow the Principle of Nonvacuous Predication to be satisfied. The fulfilment of this Principle depends on the relative suitability of the predicate to the types of SoP identified in this paper. The picture that emerges, therefore, is one where a distinction must be drawn between two types of broad focus: the Italian one, which is only licensed by the SoP discussed above, and a more permissive one. We will be able to flesh out this insight after we analyse the BFSI constructions syntactically.

4.1 The Subject of Predication and the thetic/categorical opposition

In line with recent claims, we maintain that all subjects originate within the vP. We further assume, following Bianchi & Chesi (2014), that the subject of a thetic sentence stays within the vP, while the subject of a categorical structure must raise from its thematic position to a higher position. Semantically, the postverbal subject of thetic sentences is interpreted as part of the description of an event within the predicative nucleus of the clause. By contrast, the subjects of categorical statements are compositionally external to the predicative part of the syntactic tree (Kuroda 1972; 2005; Ladusaw 1994). A consequence of this distinction is that, regardless of definiteness, the postverbal DP of the thetic structure is not presuppositional, in the sense that it is part of an eventuality whose existence is being asserted. By contrast, a categorical structure first posits the existence of the preverbal DP and then predicates something about it. In this sense, the preverbal DP of a categorical judgement is presuppostional (Bianchi & Chesi 2014: 532):

With this distinction in the background, the crucial questions that we now address is the following. Is SubjP completely inert or inactive in thetic sentences with a postverbal DP? In such sentences, what is the predication about?

4.2 The silent Subject of Predication in SubjP

According to Cardinaletti (2004: 151), “it is fair to assume that in thetic sentences, no XP in the clause checks the subject-of-predication feature”. Our analysis diverges from this position. In agreement with Bianchi & Chesi (2014), we assume that the distinction between categorical and thetic structures is dependent on the position of the thematic DP: raising a constituent from its thematic position to SubjP yields a categorical judgement, while in a thetic configuration the subject will stay in its base position (see below for some observations on non null-subject languages). However, for us, lack of movement of the DP to SubjP, and the impossibility to retrieve and identify the SoP anaphorically, on the basis of the previous discourse, do not entail that SubjP is inert. Rather, we claim that the silent thematic or situational SoP activate and occupy SubjP.¹⁴ The goal SoP is a thematic argument and hence it moves to SubjP from its thematic position (cf. (72a)). The situational argument SoP, on the other hand, is merged directly in SubjP (cf. (72b)). In existential sentences, the pivot predicate is within a Small Clause and the pro-argumental proform moves to Infl due to its clitic status (Cruschina 2012; Bentley & Cruschina 2016; Bentley 2017; to appear). As a spell-out of the contextual domain variable, it is in a doubling relation with the SoP in SubjP (cf. (72c)).

This analysis has the advantage of unifying the conditions that license BFSI with different predicate types. In this account, the silent SoP plays a fundamental role: it allows the argument DP to remain in its thematic position. This is the basic configuration that yields BFSI in intransitive predications.

Two aspects of our proposal deserve further clarification: first, the characterization of the SoP with respect to the notion of topic, and, second, the status of the SoP vis-à-vis the EPP and other syntactic requirements associated with the notion of subject. Our notion of SoP undoubtedly resembles similar concepts that make reference to the notion of topic (e.g. Erteschik-Shir’s 1997 stage topic, Klein’s 2008 notion of the topic situation) and have been considered to be crucial to an understanding of subject inversion in Romance (Lahousse 2007; 2011; Giurgea 2017; 2018; Leonetti 2017). Although our analysis builds upon these approaches, we must differentiate here between SoPs and topics proper. To being with, topics are characterized by a connection to the previous discourse that is a priori excluded with the SoP of BFSI (in the sense of bare BFSI, see §1). Drawing on Pesetsky (1987), Rizzi (2005) defines the latter property in terms of D(iscourse)-linking. Being D-linked, topics, but not SoPs, are infelicitous in out-of-the-blue contexts. On the other hand, SoPs share with topics the aboutness property, namely, their being ‘what the sentence is about’, which in SVO languages guarantees a prominent – typically, initial – position within the sentence. Therefore, while aboutness and D-linking are necessary properties of topics, only aboutness characterizes SoPs, as shown in (73) (from Rizzi 2005: 212).

Observe that Erteschik-Shir (1999: 124) does point out that the stage topic of an utterance can be overt or implicit. It is the latter type of stage topic that resembles our silent SoP. Similarly to what we propose for the situational SoP, in presentational sentences the implicit stage topic provides the spatio-temporal coordinates of the described event (Erteschik-Shir 1997: 26–27).

Along the same lines, Klein (2008) proposes that all utterances are evaluated with respect to a specific situation, which he labels topic situation. The topic situation can be identified externally (i.e., pragmatically) or linguistically (i.e., by means of overt linguistic material). As noted by Klein (2008: 290), his notion of topic situation is distinct from the traditional notion of topic, generally defined in terms of givenness or aboutness: “Under [his] approach, the ‘topic’ is not a person or an object – it is always a SITUATION. This makes it different from all other approaches […]. The topic is the situation about which the sentence says something.” In this view, the spatio-temporal coordinates provided by the SoP in BFSI will suffice to identify the topic situation of the utterance.

It is also important to disentangle the requirement of a SubjP from other syntactic requisites typically associated with subjects. For Cardinaletti (2004), movement of the logical subject to SubjP takes place in order to check the subject-of-predication feature. The same requirement was subsequently formulated in terms of Subject Criterion, which must be satisfied at the syntax-semantics interface (see Rizzi 2005; 2006; Bianchi & Chesi 2014 for details). The Subject Criterion can be viewed as a restatement in current theory of Bianchi’s (1993) Principle of Nonvacuous Predication, and we thus assume that the SoP in SubjP satisfies the Subject Criterion.

The presence of a SoP in SubjP is in principle orthogonal to the satisfaction of the EPP or the need to check Case and φ-features. In current theory these are associated with lower subject positions, and, in null-subject languages, can be checked covertly by a pro (Rizzi 1986 and subsequent work) or in a long distance fashion (Cardinaletti 2004: 151–152; Quarezemin & Cardinaletti 2017). Evidence in support of the claim that the SoP in SubjP is independent of agreement comes from the observation that not all overt constituents sitting in SubjP control agreement on the verb, as is clearly demonstrated by the experiencer of some unaccusatives (cf. (13b, 13b′)). An anonymous reviewer suggests that, in BFSI, the EPP is satisfied by the SoP itself, otherwise the fact that the construction is not available with all verbs remains unexplained. In our account, some verbs are not found in BFSI because of their lexical-aspectual properties, which are incompatible with the thetic structure with a silent SoP in SpecSubjP. At the same time, we propose that in the constructions analysed in this paper the EPP may be dispensed with or substituted by the Subject Criterion.¹⁵ This proposal raises the question of how the subject requirements are satisfied in narrow focus. Postverbal subjects licensed under narrow focus are outside the scope of our discussion (see Pinto 1997; Belletti 2004). We note, however, that such structures are not constrained to particular lexical-semantic verb classes, which suggests that they are not thetic in the sense discussed above (cf. (72a–c)): the subject is not in situ and there is no SoP in SubjP. Rather, we assume with Cardinaletti (2004: 151) that the subject of predication is the one of the preceding context. As for the EPP, it can be replaced by the Subject Criterion, as is the case with BFSI.

In non-null-subject languages, for example English, subjects tend to move to a preverbal position independently of the categorical/thetic opposition. For these languages, we can assume that subjects generally move to a preverbal subject position for independent syntactic reasons, such as the EPP requirement or the checking of Case and φ-features. In particular, Cardinaletti (2004) and Quarezemin & Cardinaletti (2017: 394) suggest that the subject may check Case and φ-features in SpecTP, and then the Subject Criterion in SpecSubjP.¹⁶,¹⁷

4.3 Thetic and non-thetic broad focus subject inversion

In Section 3.5, we tentatively suggested that it was necessary to draw a distinction between two types of broad focus subject inversion: the Italian one, which is licensed by an SoP, and a more permissive one, which is found in Spanish and other Romance languages. We now propose that these two types are revealing of a distinction between thetic and, respectively, non-thetic broad focus. Italian BFSI is thetic, and hence it is subject to the event structure constraints that have been discussed. The intransitives that do not abide by these constraints can occur in broad focus, but the relevant structures exhibit SV order because they are not thetic, i.e., they do not have a silent SoP in SpecSubjP and VS as the predication in the vP. As for transitives, we explained earlier why unbounded ones, as well as transitives with an incremental theme, are incompatible with the kind of SoP that we have discussed in this work (see Note 11). We have not identified any thetic transitive predications in Italian. We thus assume that in broad focus, transitives have the subject DP is in SubjP. The case of Spanish and the other Romance languages that allow VOS/VSO order in broad focus requires careful examination of the interplay of event structure with syntax, which has to be left to future research. Thetic structures ought to be subject to event structure restrictions in these languages, as is the case with Italian, following the predictions made in Section 3.5. The grammaticality of VOS/VSO order in broad focus, on the other hand, must depend on independent aspects of the syntax or the discourse-syntax interface in these languages. Our main point remains that the languages that only admit thetic BFSI constrain this structure to particular predicate types.