V1 clauses and EPP in Old Italian

This work discusses some aspects of the so-called “relaxed verb-second hypothesis” for Old Romance languages focusing on Old Italian data. Although the idea that Medieval Romance varieties displayed some kind of verb-second grammar is often accepted in the literature, careful consideration of the data and the predictions casts some doubts on this hypothesis and suggests pursuing a different road to account for word order phenomena in these varieties. The focus here is on verb-first main clauses, which result from merging a null element in the left periphery according to Wolfe’s (2015) influential work. If this approach is adopted, problems arise for the definition of the null categories that can occupy the left periphery of the clause: there is no motivation to postulate such elements unless a rule of obligatory pre-field occupation is independently justified, and the data speaks against such a rule. The question then arises whether a further weakened version of the relaxed V2 hypothesis is still preferable to a non-V2 analysis of Old Italian Grammar.

The purpose of this paper is to contribute to the discussion about Old Romance V2 on the basis of V1 main declarative clauses in Old Italian.While the literature both in favor and against V2 in Old Romance is quite extensive, these two positions have primarily been constructed on distinct arguments and observations.In this paper, I delve into the assumptions and the predictions of one of the major contributions that interprets V1 as evidence of a V2 syntax, namely the work by Wolfe (2015); I assess the desirability of these assumptions and explore what counterexamples there could be.Although the focus is on Old Italian, future research may broaden the scope of this discussion to include other Old Romance varieties.
In this section, an overview of the relaxed V2 hypothesis and the current account of V1 clauses in Old Romance is provided.Section 2 discusses the problems that arise if the account by Wolfe (2015) is adopted, and the evidence against obligatory pre-field occupation.The data in Section 3 suggests that other generalizations that the V2 hypothesis is supposed to explain might be questioned as well.Section 4 concludes.

Assumptions for a relaxed V2 grammar
The discussion in Holmberg (2015) shows that a strict, linear-order-based definition of a V2 grammar, that requires the finite verb in second position in all main clauses, cannot be maintained on a cross-linguistic level.All languages that show phenomena calling for a V2-based explanation also allow for certain deviations from the canonical pattern under certain conditions.
As a consequence, another interpretation of V2 was proposed, reported in (1), which considers two abstract syntactic rules and tries to avoid commitment to a restriction on the linear order of words or phrases.
A functional head in the left periphery attracts the finite verb.b.
This functional head requires a constituent in its specifier position.
The syntactic rule in (1a) derives from den Besten's (1983) original intuition that V2 is a root phenomenon and that the finite verb competes for the same position as the complementizer, which is higher in the structure than the canonical subject position.The rule in (1b) derives the obligatory presence of a constituent on the left of the finite verb (obligatory pre-field occupation, as it is named in Blümel, 2017), i.e., the ungrammaticality (or high markedness) of V1 orders.Promiscuity (any type of phrase) and uniqueness (only one phrase) of the pre-field are not derived by necessity, and this is a desideratum of the approach; these features appear in their strictest form only in a handful of V2 languages, such as German or Dutch, but need to be parametrized to account for cross-linguistic variation.
In order to derive a constraint on the number of phrases in the pre-field, the socalled "Bottleneck effect" (Roberts, 2004) is commonly assumed.According to this assumption, the phrase that satisfies the rule in (1b) acts as an intervener to the movement of any other phrase.
"XP-movement to [Spec,CP] in full V2 clauses is movement caused only by [C]'s EPP-feature […].The moved XP is thus of no particular type in terms of the typology of potential interveners, and so is able to block any type of movement."(Roberts, 2004, p. 316) On the other hand, external merge of phrases on the left of the intervener is still allowed.This strategy is used to derive "V3+" orders in Old Romance, where multiple topical elements can occur on the left of the inflected verb (Benincà, 2013;Wolfe, 2016).
The definition in (1) does not assume any cluster of phenomena to be necessary or sufficient for a language to be V2.Instead, any grammar whose phenomena can be analyzed as the result of those rules can be called a V2 grammar.Elaborating on this idea and on the observation of data from Romance varieties, V2 was proposed to be the result of a "conspiracy" of phenomena (Poletto, 2019;Wolfe, 2021): no single phenomenon is sufficient to make a grammar V2, but a combination of relevant phenomena can suggest a V2 analysis of a language, or trigger the acquisition of a V2 grammar by a child.I will focus on V1 main declarative clauses below; other potentially V2-related phenomena will be considered in Section 3.

V1 in Old Romance
An influential paper by Wolfe (2015) shows that V1 main declarative clauses are generally attested in most Old Romance texts.However, following ideas already put forth in the literature, Wolfe proposes that such V1 clauses are no counterexample to the V2 hypothesis, but either result from merging a null (topical) element in first position (Benincà, 2013;Poletto, 2014;Roberts, 1993;Salvi, 2012) or they are a marked syntactic option akin to what is also attested in Germanic V2 languages (see for instance Catasso, 2019;Önnerfors, 1997).
To support his proposal, Wolfe tries to show empirically that V1 declarative clauses are restricted to the three functional types reported below. 1 The first two types presumably involve a null topical element, while the third one is analyzed assuming the presence of a special discourse operator in first position (definitions from Wolfe, 2015).
1 Wolfe (2015) also defines the category of Deictic V1, only useful for Old Venetian, where the null topic is highly salient in the discourse and anchored to the "here and now."This category is then conflated with Topic Continuity in Wolfe (2016).
Topic Continuity V1: These sentences feature a Null Topic, which is always co-referent with a preceding nominal expression.

Thetic V1:
The whole sentence is informationally new and serves to introduce not-yet-activated referents into a discourse.
Narrative V1: These sentences occur exclusively with verbs of saying and express/recount a proposition that is true, rather than explicitly asserting its truth value.
Given the frequencies of V1 declarative clauses in the different varieties considered and the distribution of the above categories (reported in Table 1), it is possible to observe that Topic Continuity V1 is invoked for those languages that also display the highest frequency of V1 overall.The three functions are exemplified in (2) with data from Old Italian, which is a V2 language according to Benincà (2013) and Poletto (2014).Here, "Old Italian" indicates the Tuscan variety attested in texts written between the 13th and the 14th century.All the data used in this work comes from prose texts of Corpus MIDIA, 2 period 1200-1375.However, what all the examples considered by Wolfe and all the possible candidates for Topic Continuity V1 in Old Italian data 3 have in common is that the subject is dropped.Instead, no association between other null arguments and V1 is suggested by the data or by the literature.Object drop in Old Italian is reportedly very rare, and generally possible only with coordination (Egerland, 2002;Pontbriand, 2022) or in non-finite clauses (Salvi, 2010).Instead, topical direct objects in V1 main clauses regularly appear as clitic pronouns, as in (4).
(4) (Boccaccio, Esposizioni) Commendalo qui l' autore dell' amplitudine della praise-3.SG-him here the author of-the breadth of-the sua facundia.his eloquence 'Here, the author praises him for his abundant eloquence.'Given that Old Romance varieties allow for subject drop at least in main clauses, the question arises whether Topic Continuity V1 and canonical pro-drop are two distinct phenomena.According to Wolfe (2015) and following work, they are derived by means of two distinct null elements that differ with respect to their featural specification.
In what follows, these two null elements will be referred to as pro and proTop.In principle, the former derives canonical null subjects in modern Romance languages, whereas the latter is licensed after interacting with a functional head in the left periphery of the clause.

Topic Continuity and Thetic V1
This section discusses the obligatory pre-field occupation in Old Italian.The focus is on Topic Continuity and Thetic V1 as defined by Wolfe (2015), where it is assumed that a null topic satisfies the V2 requirement.A discussion of the possible ways of implementing Topic Continuity V1 shows that either incorrect predictions are made, or the requirement needs independent justification.The data from V1 sentences with overt arguments provides conclusive evidence against this requirement.

3
Here, a sample of 471 V1 main clauses is considered.

V1 with null subjects
Following Wolfe (2015), Topic Continuity V1 is derived by means of the null element proTop, which differs from canonical pro inasmuch as it requires further interaction with a left-peripheral head to be licensed.
This interaction is triggered or required by a different featural specification (an unvalued Topic feature, according to Wolfe, 2015, who adopts the framework by Chomsky, 2000).However, this difference is not clearly supported by a distinct contribution to the information structure of the clause: pro is itself a topical element (Frascarelli, 2007), and a different type of topicality between Germanic topic-drop and Romance pro-drop is arguably not principled (Sigurðsson, 2011).Moreover, no doubling effect in the sense of Koeneman and Zeijlstra (2014) can be observed, given that both proTop and the left-peripheral head it interacts with are phonologically null: this makes the difference invisible for acquisition.This potential problem will not be considered in this paper.
Given the considerations so far, pro and proTop are defined as minimally distinct elements: both have no phonological component, are associated with nominative case, have a referential denotation, and are topical from the point of view of information structure.The only difference is associated with the locus of syntactic licensing, which is below TP for pro and in SpecCP for proTop.The discussion below will therefore focus on the potential syntactic consequences of maintaining this distinction to account for V1 clauses in Old Romance.
The hypothesis that both pro and proTop are in principle available to an Italian variety can be interpreted in two ways.The two null elements can be either both available to the grammar, or proTop can replace pro in Old Italian.Therefore, the two hypotheses in (5) need to be considered, and both will be discussed in what follows.
In Old Italian, proTop is acquired instead of pro.b.
In Old Italian, both proTop and pro are available.
The hypothesis in (5a) will now be examined.If this hypothesis is adopted, the two corollaries in (6) follow.
Every null subject is derived by means of the same proTop.b.
Every proTop must be generated below TP, to be licensed as an argument, before interacting with a left-peripheral head.
To formalize the interaction of proTop with a left peripheral head and its consequent licensing, it is possible to introduce a specific operation in the grammar that licenses null elements, such as C/edge-linking in Sigurðsson (2011): this operation can only take place locally and requires the null element to move across C/Fin to interact with a Top head.This is represented in Figure 1.Alternatively, an Agree operation with a Topic head can check the [uTopic] feature on proTop.Even if movement is not strictly required, Agree also needs to obey locality, and no intervener can occur between the probe and the goal (this option is also preferred by Wolfe, 2015, 2019, following Chomsky, 2000and Roberts, 2010).This is represented in Figure 2. A third possibility is to introduce a criterial [Top] feature, in the spirit of Rizzi (2000), that must be carried by both a left-peripheral head and its specifier.This assumption again requires that proTop moves across Fin to SpecTopP to create the correct spechead configuration.This is represented in Figure 3.In all cases described, given (6), the movement of the null category or the Agree operation must cross the Fin head, whose specifier acts as an intervener given the Bottleneck effect.For the purpose of this conclusion in a framework that assumes an Agree operation as in Chomsky (2000), it is crucial that phrasal movement is always a consequence of an Agree operation, and that interveners to movement, as in the case of the Bottleneck effect, are actually interveners to Agree.Therefore, given that the Bottleneck effect allows for movement of at most one phrase (or one Agree operation) across Fin, all these approaches predict the interaction in (7) between proTop and other overt left-peripheral constituents. (7) Null subjects are not licensed if a constituent has already been moved to the left periphery.
This prediction has never been reported in the literature, and the data shows that it is not borne out.Evidence for this is that non-topical items (in the sense of Martins, 2019), which cannot be first-merged in the left periphery, can be fronted when the subject is null in the same clause.This is exemplified below with a focused quantifier (8a) and a low aspectual adverb (8b Given the prediction and the counterexamples above, the hypothesis that all null subject clauses in Old Italian are derived by means of the same proTop must be rejected. The second hypothesis, stated in (5b), considers that both pro and proTop are available to the grammar and can enter the derivation.Again, two possibilities arise, listed in (9).( 9) a.
Both pro and proTop enter the derivation of Topic Continuity V1. b.
Either pro or proTop is selected for the derivation.
If the derivation of Topic Continuity V1 requires both canonical pro in the clausal domain (below TP) and proTop in the left periphery, as in (9a), their distribution mirrors what has been said for resumptive clitics and topicalized phrases in Old Romance (Benincà, 2013): the resumptive element is sufficient to make the sentence grammatical, and the topicalized phrase is optionally first-merged in the left periphery.The crucial difference with clitic left dislocation is that in this case, both the topicalized phrase and the resumptive element are null.
As a consequence, every sentence with a null subject and at least one non-null constituent on the left of the verb can be parsed in two different ways.In one parse, only pro is present; in the other, in addition to pro, proTop is also merged in the left periphery.This is exemplified in (10).
(10) (Giamboni, Libro de' vizi) E (proTop) poi mi puose pro la mano in and then to_me put-3.SG the hand in sul petto.on_the chest 'And then she put her hand on my chest.' The only situation where only one parse is available is when no overt phrase occupies the preverbal position.In all the other cases, proTop is reduced to an element that, apart from being phonologically null and language-specific, is also syntactically inert and vacuous for semantics and information structure: it triggers no syntactic operation, and the reference and topicality are already independently provided by pro during the derivation of the same clause.
As a consequence, postulating the availability of proTop serves the only purpose of satisfying the rule of obligatory pre-field occupation.There is no separate reason to assume such a null category if the rule does not find independent justification.
Instead, the option in (9b) suggests that the derivation conveniently chooses between a null element that can move to and occupy the left periphery when needed and a null element that does not need to interact with the left periphery when this interaction is blocked.
This formulation presents an apparent look-ahead problem.However, it is possible to conflate this problem into the general possibility of moving topical constituents or subjects to SpecCP in V2 languages.The formulation in (11) does this while achieving empirical equivalence to (9b).
The hypothesis in (11) would be supported on a cross-linguistic level if all V2 languages that allow for pro-drop also allowed for V1 main clauses with null subjects.Kashmiri (Bhatt, 1999), Old Icelandic (Sigurðsson, 1990) and Estonian (Lindström, 2001;Vihman & Walkden, 2021) seem to be such cases, but further research is needed to show whether argument drop with and without V1 has different properties in these languages.If this interpretation of ( 11) is adopted, the V2 requirement is trivially satisfied whenever a sentence has a null subject.
However, for Old French, it has been claimed that Topic Continuity V1 (and V1 in general) is very restricted or absent, given its low frequency in corpora (Klaevik-Pettersen, 2019; Wolfe, 2015Wolfe, , 2019)).If this difference between Old French and Old Italian is to be maintained, a language-specific null category that behaves as in (9b) needs to be stipulated on the basis of the rule of obligatory pre-field occupation, which in turn needs a separate justification again.
To conclude this section, although both roads in (5b) seem in principle to be viable to some extent, the rule of obligatory pre-field occupation needs independent justification.The options allow for trivial satisfaction of the requirement, or alternatively, build on an ad hoc stipulation of a null element that has no other function in the grammar and is otherwise unmotivated.In the next section, V1 sentences with overt subjects will be discussed in order to consider whether any separate motivation for the obligatoriness of a constituent in first position can be found.

V1 with overt subjects
According to the hypothesis developed by Wolfe (2015Wolfe ( , 2016Wolfe ( , 2019) ) for Old Romance varieties, once sentences with verbs of saying have been excluded, all V1 main clauses with overt arguments 4 are thetic.This requirement on information structure imposes that all the arguments are new information, i.e., the entire sentence is in broad focus.
Therefore, V1 clauses with overt arguments are expected to be a minority in the overall amount of V1 clauses: a V1 clause with a null argument can have a null topical subject (which is an unmarked option) and a variety of different information structures for the rest of the material; from the discussion in the previous section, null subjects are as unconstrained with V1 as in general. 5Instead, in Wolfe's theory, a V1 clause with overt arguments has the specific requirement of theticity.Given that no such requirement on information structure exists for sentences with overt arguments in general (i.e., V2 sentences with overt arguments do not have to be thetic), V2+ sentences can be used as a baseline for comparison.
A collection of V1 main clauses from a corpus of Old Italian suggests that overt subjects in V1 are not rarer than overt subjects with V2+.For V1, 471 sentences were collected from Corpus MIDIA: only sentences with a null referential subject or an overt DP subject were considered, excluding verbs of saying, sentences introduced by a conjunction, verbs that do not assign a θ-role, impersonal sentences, and CP subjects. 6As a baseline for comparison with V2+, a sample of 210 main clauses from the same corpus was collected.As shown in Figure 4, the proportions of overt subjects in the two samples are comparable, 7 licensing the question of whether all V1 clauses 4 Therefore excluding the null topics of Topic Continuity V1 already discussed in the previous section.

5
An interesting discussion of other unexpectedly high frequencies of V1 clauses in Old Spanish can be found in Sitaridou (2011Sitaridou ( , 2019)): apart from being very frequent in general in both main and embedded clauses, it also displays features that are unattested in V1 clauses of Old Germanic languages.The present paper provides additional evidence for treating V1 clauses as counterexamples to the V2 hypothesis.6 Although some of these categories are potentially relevant to the discussion, they were not considered because no prediction about them is explicitly made in the literature discussed here.Conjunction is analyzed as a topic marker in Poletto (2014) and is reported to allow exceptional main clause V1 in historical varieties of German (Coniglio & Schlachter, 2013 Here and in all the plots below, the confidence intervals are not an indication of variance in the data, but are instead intended to convey the amount of certainty for the (invariable) measure I obtained from the corpus.They were derived using the "prop.test"function in R (R Core Team, 2022).
with overt subjects are really thetic.A qualitative look at the data may give an answer to this question.To derive Thetic V1, Wolfe (2015) assumes another type of null topic pronoun, "that keeps all verbal arguments postverbal and thus unambiguously interpreted as informationally new."Therefore, two conditions need to be satisfied: all the arguments need to be new information, and a null topic must be available.
However, among the sentences collected, some feature subjects that are already given in the discourse.Two examples are reported below.In the target sentence of (12a), the subject "the king" is at least given, and possibly topical, because it is already the subject of the previous sentence.Similarly, example (12b) has a single discourse topic, that is realized as a postverbal pronoun in the target V1 sentence.
( A good candidate for a null topic in preverbal position is the null stage topic as defined in Erteschik-Shir (1997, 2013).This proposal assumes that a topic is necessary for truth value assessment and that thetic sentences are interpreted as being about a spatio-temporal discourse anchor.A null stage topic is unavailable with individuallevel predicates, that are regularly interpreted as properties of the subject.Supporting this proposal is the unavailability of individual-level predicates with singular nonspecific indefinite subjects, which cannot be topics.The sentence in ( 13) is infelicitous because no good candidate for topicality is available: neither the subject nor a null stage topic.
Going back to the Old Italian data, the prediction is that, if a V1 sentence features an individual-level predicate, it needs to promote one referent denoted by an overt constituent (generally the subject) to the role of topic; therefore, the sentence itself cannot be thetic.Such sentences are attested in the data sample for V1, as exemplified in ( 14). ( 14 Therefore, evidence from sentences with individual-level predicates suggests that some V1 sentences with overt subjects do not qualify as thetic and have no null topic available to occupy the pre-field.This is paired with the fact that some postverbal subjects in V1 clauses are clearly non-new.The surprisingly high frequency of V1 clauses with overt subjects (comparable to the frequency of V1 clauses with null subjects) is now supported by the non-exceptional status of such subjects in terms of information structure.
To conclude, not only did the data provide no separate evidence for a rule of obligatory pre-field occupation, but it showed instead that counterexamples of such a rule exist.Moreover, the assumptions that come with it are theoretically problematic, as shown in Section 2.1, and they generate a grammar that is too restrictive vis-à-vis the empirical evidence.The present proposal is therefore to abandon the requirement of pre-field occupation for the Old Italian grammar.

Other V2-related phenomena
The discussion so far has suggested that the V2 hypothesis needs to be substantially revised or abandoned for Old Italian.A formulation of a relaxed V2 grammar that follows the syntactic rules stated in Holmberg (2015) (see (1) above) cannot be maintained because no left-peripheral functional head can be found that regularly requires a constituent in its specifier position.
The question then arises whether a revision of the relaxed V2 hypothesis comes at the expense of the explanation of the other V2-related phenomena that it tried to capture.In this section, I will briefly present some data suggesting that some major explananda might turn out not to be well defined or accounted for under a relaxed-V2based analysis of Old Italian grammar.I will consider three phenomena: the distribution of pro-drop in embedded clauses, the distribution of enclisis and proclisis, and clitic doubling of fronted direct objects.
Much work on Old Italian assumes that a strong indicator of an underlying V2 grammar in Old Romance is the asymmetry between main and embedded clauses with respect to certain syntactic phenomena, which include the distribution of pro-drop and the position of pronouns that cliticize on verbs in first position.
Building on the intuition by Adams (1987) for Old French that pro is licensed only if the verb moves to C, Benincà (1994), Poletto (2014), and related literature maintain that null subjects are highly constrained in Old Italian embedded clauses and that this is an indicator of an asymmetry between clause types and a different position of the verbal head: T in embedded clauses and C in main clauses.According to Poletto (2020), verb movement to C licenses a null topic that then licenses pro in the clausal domain; where verb movement does not apply, null subjects cannot be licensed.
Given this proposal, the expectation is that null subjects are frequent in main clauses, but very rare or absent in embedded clauses.However, the data below suggest that a different generalization is also possible.In a sample of 1157 sentences, 8 null 8 These sentences come from nine prose texts of corpus MIDIA: anon., Capitoli della Compagnia di San Gilio, Statuto degli oliandoli di Firenze, Novellino; Alighieri, Convivio, Vita Nuova; Giamboni, Fiore di rettorica, Libro de' vizi; Latini, Fiore di filosofi, Rettorica.The following clause types were considered: main declarative and interrogative clauses, complement clauses embedded under dire "say", relative clauses introduced by the pronoun quale "which", causal and temporal adverbial clauses introduced by imperciò che "because" and quando "when", antecedents of conditionals introduced by se "if", and interrogative subjects seem to be robustly attested across different clause types, as shown in Figure 5.The generalization that null subjects are highly constrained in embedded clauses is not straightforwardly supported by corpus data.Independently of potentially significant differences between clause types or possible explanations for specific sentences, 9 an overview of the attested null subjects as the one above casts doubts on a naïve version of asymmetric pro-drop and challenges the explanatory power of the relaxed V2 hypothesis: the hypothesis alone is not enough to account for the data, and it can be reconsidered in light of the discussion in the previous sections.
A second argument in favor of an asymmetry between main and embedded clauses is related to the position of clitic pronouns when the verb is in first position.According to Benincà (2006), enclisis occurs when the inflected verb has moved to C together with a proclitic pronoun and no constituent is focus-fronted; this configuration triggers further movement of the verb to a higher functional head and stranding of the clitic pronoun, that becomes enclitic.This formulation predicts enclisis in verb-and topic-initial main clauses and proclisis elsewhere.As Benincà (2013) further notes, enclisis should never occur in embedded clauses, thus being a strong indicator of a main-embedded asymmetry.
However, a careful consideration of the data reveals that enclisis never occurs in embedded clauses after a complementizer, but can be often found in embedded clauses embedded under (a)domandare "ask", sapere "know" or vedere "see".The italicized words were used for the queries.9 Using a different corpus and different methods, Cognola & Walkden (2019) also find a number of null subjects in Old Italian embedded clauses.Yet they claim that these null subjects might be qualitatively different from what can be found in a canonical pro-drop language such as modern Italian.
clauses after coordination.In a sample of clauses introduced by the coordinator e 'and' 10 , enclisis is the preferred option both in main and embedded environment. 11Despite a potential difference between main and embedded clauses, the claim that enclisis is ruled out in embedded environment is clearly falsified by the data.One of the counterexamples is reported in ( 15 A third argument concerns clitic doubling of fronted direct objects.Old Romance languages have been claimed to display a V2 grammar also because of their remarkable fronting possibilities.In particular, object fronting without clitic resumption is mentioned as evidence for a V2 syntax by Wolfe (2015Wolfe ( , 2016)).The phenomenon was discussed by Benincà with the following generalization: "A fronted direct object can occur without clitic resumption only if it is adjacent to the verb" (Benincà 2013: 74).The structural explanation for this fact is that direct objects that 10 See Sitaridou (2019) for a thorough discussion of the hypotheses around V1 clauses introduced by e/y in Old Spanish and why none of them is coherently compatible with a V2 grammar.
11 All instances of the conjunction e 'and' were collected from corpus MIDIA for the same texts as footnote 5 when e was followed by a word either tagged as a verb with an enclitic pronoun or as a proclitic pronoun.The 156 sentences used for Figure 6 are all the relevant ones in the sample.
are syntactically left-adjacent to the inflected verb bear focus and are moved because of a focus operator, which is never associated with clitic resumption.Instead, a topicalized direct object does not need to be adjacent to the verb and is always resumed by a clitic pronoun.
However, it is possible to find counterexamples of this generalization in Old Italian texts.In the examples in (15) below, the fronted direct object is separated from the inflected verb by another argument of the predicate (a subject and a beneficiary 12 ); nonetheless, no resumptive pronoun is necessary.
( Although the data just presented are not conclusive about the phenomena they exemplify, they show that abandoning the V2 hypothesis for Old Italian, as proposed 12 An anonymous reviewer raises the doubt that gran male al comune e a' cittadini could be syntactically ambiguous and be treated as a single constituent.Although it is difficult to rule out this hypothesis completely, notice that fare male is a (possibly idiomatic) predicate that regularly takes a beneficiary argument, which can also be realized as a dative clitic pronoun to the exclusion of the nominal male, as is the case for gli in (ii) below.
(ii) che io non gli farò male alcuno.that I NEG to_him do-fut.1.sgbad any 'that I will not cause him any harm.'(Boccaccio, Decameron) here, does not necessarily come with the loss of an explanation for a number of syntactic phenomena.Instead, the analysis of these phenomena might have been biased in the literature by a hypothesis that does not find strong theoretical or empirical support.

Conclusions
This paper has considered the proposal, put forth by Wolfe (2015), that V1 main declarative clauses do not represent a counterexample to the hypothesis that Old Romance varieties have a V2 grammar that requires the pre-field to be occupied.According to the proposal, all V1 clauses can be reduced to a limited number of functions that correspond to just as many null elements that can fill the preverbal position.
It was shown that the category of Topic Continuity V1 poses problems to the derivation.It either makes incorrect predictions, if it is assumed that only one null pronoun is available to the grammar, or it finds no independent justification if two null pronouns need to be assumed.Moreover, the data about V1 clauses with overt arguments shows that not all these sentences can be thetic, thus providing conclusive counterexamples to the rule of pre-field occupation.
This discussion reaches the conclusion that such a rule must not be assumed for Old Italian grammar.This modification of the hypothesis returns a straightforward way of capturing the data discussed above.
V1 sentences with null subjects are immediately accounted for, and no new assumptions are required.There is only one null element, i.e. the same canonical pro as modern Romance null subject languages.No movement of a null element to the left periphery needs to be postulated to satisfy the requirements of the grammar; in fact, nothing is forced to be merged on the left of the verb.This is in line with the discussion in Sitaridou (2019) for Old Spanish that it is not possible to identify a category of phrases that are associated with verb movement when fronted.This follows naturally from the conclusion of this paper (in agreement with Sitaridou) that no rule of obligatory pre-field occupation can be assumed.Fronting is the result of a mapping between word order and information structure (thoroughly investigated by Batllori & Sitaridou, 2020 for Old Spanish) that is independent of a V2 grammar.
From the diachronic point of view, another advantage emerges: no unmotivated change in the pro-drop system from Old Italian to Modern Italian needs to be postulated.In the absence of evidence for a different distribution of the phenomenon, a better hypothesis accounts for pro-drop in Old and Modern Italian in the same way.Independently of what ultimately licenses pro in these languages, Wolfe's (2015) hypothesis introduces variation that is not clearly reflected by the data, whereas the current proposal is compatible with the absence of variation.
The existence of V1 sentences with overt arguments can also receive a straightforward explanation if the obligatory pre-field occupation rule is abandoned.In thetic sentences, arguments may be postverbal as a result of a general crosslinguistic tendency for focused material to occur later or for thetic sentences to be verbinitial (see Matras & Sasse, 1995).Once the role of information structure is acknowledged, the predictive power of a V2-related rule of pre-field occupation is considerably diminished.
The data also suggests that a postverbal subject can be the topic of the sentence.The availability of postverbal non-focused subjects has already been observed by Ciconte (2018), who proposes that they are generally continuing topics.If this is an independently motivated possibility of the Old Italian grammar, V1 sentences that are not thetic are indeed expected to occur under the new hypothesis.
The apparent shortcoming of this proposal is that it introduces the possibility of a V2 language without obligatory pre-field occupation, which was not explicitly considered in the typology of Old Romance (Wolfe, 2019) or V2 in general (Holmberg, 2015).Given the fact that the V2 hypothesis for Old Romance has already received criticisms in the literature (Alboiu, Hill & Sitaridou 2014;Batllori & Sitaridou, 2020;Kaiser, 2002;Martins, 2019;Rinke, 2009;Rinke & Elsig, 2010;Rinke & Meisel, 2009;Sitaridou, 2011Sitaridou, , 2012Sitaridou, , 2019)), primarily building on the fact that the verb often occurs later than the second position, the possibility endorsed here is to abandon the hypothesis completely and maintain that Old Italian is not a V2 language at all.The data presented in Section 3 suggests that some major generalizations associated with the V2 hypothesis are not straightforwardly supported by Old Italian data.It is therefore desirable to reconsider the hypothesis at least for Old Italian, with potential consequences for a general analysis of Old Romance verb syntax.

Figure 5 .
Figure 5. Frequency of null subjects across clause types

Figure 6 .
Figure 6.Frequency of enclisis after coordination e 'and'

Table 1 .
Frequency of V1 and distribution of V1 types in Wolfe's corpus ).
Sitaridou, 2019dou, 2019against the applicability to Old Spanish).Out of 56 V1 clauses without a θ-marked DP subject, a strikingly frequent pattern (31 sentences) is the construct a(v)venne/ad(d)ivenne/intervenne che 'it happened that'.This is exemplified in (i) below.No overt (postverbal) expletive pronouns are attested in the sample.
They are made of the same substance as Him, and they love what God loves, and they hate what God hates.And their fortitude benefits so much from this that nothing can harm them.They behave like true knights, who never see a storm bad enough to worry about.' (Caterina da Siena, Lettere) ), where subjunctive mood on rechi 'bring' is only licensed under embedding.