From Latin to Modern Italian: Some Notes on Negation

: This article aims at investigating some diachronic aspects of the Italian negative system, considering a time span ranging from Old Latin to Modern Italian. Most of the negative polarity phenomena populating the Modern Italian system are consequences of a crucial change that occurred in Old Latin: The Latin negative morpheme n¯on (“not”), which initially displayed a maximal projection status, and became a syntactic (negative) head. This change caused the shift from a double negation system to a negative concord one, which affects many Romance languages (and their dialects). It also determines the availability of the expletive reading of negation in Italian, as well as in other Romance languages (ex. French), calling for a new generalization: only languages (and structures) displaying a negative head allow the expletive interpretation of negation, languages displaying a maximal projection status do not.


Introduction
This article aims at investigating some diachronic aspects of the Italian negative system, considering a time span ranging from Old Latin to Modern Italian.
Negation characterizes all and only human languages (Horn 1989;Speranza and Horn 2012;Greco 2020b) and it represents a one-place operator reversing the truth-value conditions of the sentence in which it occurs. Consider, for example, the following declarative sentences: (1) a.
The situation is under control. b.
The situation is not under control.
The sentence in (1b) is true if and only if the sentence in (1a) is false, and vice versa. However, this definition just holds for some types of clauses, such as declarative clauses, since they have truth-value conditions, but not, for example, for interrogatives and exclamatives that do not have them. For this reason, I will consider negation as a complement-set operator (see Delfitto and Fiorin 2014 and the references therein) in order to take into account the complexity of the data and other kinds of sentential negation (among others, see Krifka 2010;Delfitto 2013).
Crucially, Latin and Italian are often taken as instances of two very different negative systems-although Italian comes from Latin-respectively, a double negation system (1a)where the co-occurrence of two negative elements generates an affirmative meaning-and a negative concord one (1b)-where the co-occurrence of two, or more, negative elements constitute a single instance of negation: (2) a. nemo nōn videt (Cic., Lael. 99.6) nobody not sees 'Everyone sees' b.
Non vede nessuno not sees nobody 'Nobody sees' In line with part of the literature (see below), I will show that this classification is incorrect, since Latin also displays some examples of negative concord constructions. I will propose that this phenomenon depends on a crucial change that occurred in the syntactic status of the Latin negative morpheme nōn ("not"), which initially displayed a maximal projection status, but it became a syntactic (negative) head over time. This was the result of changes that responded to generalizations, such as those formalized as Jespersen's Cycle and Spec-to-head principle/Head Preference Principle. Crucially, this change occurred in the earliest attestations of Latin (I-III century BC), particularly in texts reporting examples of "lower class" Latin-a stylistic variant typical of the not-educated classes, such as the letters of the soldier Claudius Terentianus (1st century BC) and in texts reporting examples of colloquial Latin-a register that educated classes used in non-formal contexts and was well represented by Cicero's letters and Plautian's comedies.
The shift from a maximal projection status to a head one had some crucial consequences, such as the availability of the expletive reading of negation in Old and Modern Italian (3a-b), as well as in other Romance languages (ex. French): ( Both in (3a) and (3b) the temporal subordinate clause should be negative due to the occurrence of the Italian negative morpheme non ('not'), but it is affirmative, instantiating a case of vacuous-or expletive-negation. Crucially, the availability of this phenomenon seems to rely on the syntactic status of the negative morpheme that is involved, calling for a new generalization: (4) Generalization: only languages (and structures) displaying a negative head allow the expletive interpretation of negation, languages displaying a maximal projection status do not.
Moreover, I will show that expletive negation-which is commonly considered to be a unitary phenomenon cross-linguistically codified-consists of distinct subtypes and I will discuss a twofold partition between weak and strong expletive negation sentences based on the behavior with some neg-words and negative polarity items. This paper is organized as follows: in Section 2 I will first address some fundamental questions on the Latin negative system, focusing on the most widespread negative morpheme nōn ("not") (Section 2.1). I will then show the occurrence of some negative concord structures (Section 2.2), proposing that the change in the nōn syntactic status (Section 2.3)from a maximal projection status to a head one-caused them. In Section 3 I will offer a dissertation on negation in Old Italian (Section 3.1), focusing on the Tuscan language (Section 3.1), from which Modern Italian was born. I will also consider some of the first attestation of expletive negation (Section 3.2). Finally, I will discuss the negative system in Modern Italian, starting from some phenomena inherited from Old Italian (Section 4.1). I will then show some innovation in Modern Italian, at least with respect to expletive negation (Section 4.2), discussing a twofold partition within it and some consequences (Section 4.3). I will then finish the paper with some concluding remarks in Section 5.

Negation in the Latin System
The constellation of negative phenomena affecting human languages is very rich and ample. For example, negation can range over the entire sentence, as well as over a singular constituent: (5) a. Non erit profecto tibi, quod scribo, hoc novum (Cic. De Orat. 3.1) 'What I write will certainly not be new to you' b. Agri reliquit ei non-magnum modum . . .
(Plaut. Aul. 13-14) 'He lefted him a not-big piece of land' In this work, greater attention will be paid to the typology in (a), the sentential negation, and different syntactic structures will be considered.

A Complex Negative System
Latin displayed two sentential negative morphemes, nōn and nē 1 (Ernout and Thomas 1953, pp. 148-49); the former was originally reserved to constative sentences and the latter to the non-constative ones, such as performative, prohibitive, optative clauses, etc. The morpheme nē also instantiated a case of complementizer in clauses introduced by verba timendi (ex. timeo ne) 2 . A further negative element was represented by haud 3 . However, its attestation was very limited compared to the previous forms and it was restricted to some syntactic contexts, for example, with adverbs and adjectives (ex. haud facile, let. 'not easy') and, only rarely, with verbs (Plaut. Amph. 185: Facit ille quod volgo haud solent, Eng. 'He does what people usually do not') 4 . However, haud and nē were replaced by nōn already in the early stages of Latin, becoming the most widespread negative marker. This is the reason why I will focus on it in this paper, leaving aside the other negative markers 5 .
The morpheme nōn is a complex element deriving from nē ('not') + oinom (lat.ūnus, 'one') (Ernout and Meillet 1959, p. 444;de Vaan 2008;Fruyt 2011), still recognizable in the ancient form noenum (Plaut. Aul. 67). Its derivation perfectly fits into the common evolution of negative elements, as proposed by Jespersen (1917) ("Jespersen's Cycle") 6 : the original negative morpheme nē has been joined to a second element with a reinforcing value, in this case a numeral pronoun. At first it was exclusively dedicated to the indicative and the infinitive mode 7 , however, since the Imperial age, it has progressively replaced the form nē, so much so as to lead some grammarians to report the mistake (Quint. inst. 1, 5, 50: Qui tamen dicat pro illo 'nē feceris' 'non feceris', in idem incidat vitium, Eng. 'However, anyone who says 'non feceris', instead of 'nē feceris' 'don't do it', would fall into the same mistake'). It is worth knowing that nōn tended to build compounds, such as nōn-ne, an interrogative particle presupposing an affirmative answer. More examples are in nōnnihil, nōn-numquam, and nōn-nemo, where two negative items give an affirmative meaning, respectively "something", "sometimes" and "someone". This particular propensity of nōn to form compounds will be very important for its definitive syntactic classification, and it will be addressed in Section 2.3.
Another form of sentential negation is the use in isolation of negative pronouns, adjectives, and indefinite adverbs, such as nemo ('nobody'), nusquam ('nowhere'), nullus ('not any'), ecc. (es. Enn. trag. 22 R3.: Nemo est tam firmo ingenio "No one is so resolute in spirit"). These negative objects do not require any kind of restriction in the choice of the verbal mode. Again, they became a form of sentential negation through a series of transformations that are well represented by Jespersen's Cycle. Quoting Ernout and Thomas's (1953, p. 153) words, «Il arrivait à ces formes de perdre leur sens propre pour servir de négations fortes, surtout dans la langue parlée et en poesie». Consider, for instance, the derivation of nullus, which is formed by nȇ and ullus (<oinolom; Eng. 'any') (Orlandini and Poccetti 2012): this process is often at the basis of the creation of negative polarity items (NPI), which are conserved in many Romance languages. We already saw that Latin shows these kinds of changes, as in nōn, but this passage has remained visible in the case of the negative indefinites, as some ancient texts show: The so-called minimizers represented a similar form of sentential negation. Those are lexical items denoting small measures and quantities, that were used to strengthen the negative interpretation of a sentence. Consider, for instance, nihil ('noting'), where the negative morpheme nȇ is associated with an element indicating a minimum quantity, hīlum, technically the "thread" of green beans (de Vaan 2008; Orlandini and Poccetti 2012) (ex. "nihil est quod timeas", 'you have nothing to fear' Plaut. Amph. 1132).
Crucially, the negative elements-regardless of their category, i.e., negative morpheme, NPI, etc.-showed two constant behaviors: they usually preceded the verb in the word sequence 8 and they instantiated a case of double negation-two negation yield an affirmation-when they occur in the same sentence: (7)  According to Ernout and Thomas (1953), the order of the negative elements also determined their logical interpretation: if the indefinite preceded the negation (nemo non), there was a universal interpretation corresponding to 'all' ((ibid., p. 154) 'affirmation total ou renforcée'), if the negative morpheme preceded the indefinite (non nullam) there was an existential interpretation corresponding to 'something' ((ibid, p. 153) 'affirmation partielle ou restreinte') 10 . However, the double negation phenomenon was stable only at the regulatory level, whereas it was not stable at the "language of use" level, thus opening the doors to those syntactic changes that flowed into the Romance languages.

From Double Negation to Multiple Negation
The scenario described above presents a fairly stable and codified linguistic situation, in which the formation of negative sentences followed well-defined strategies. However, in the earliest attestations of Latin new strategies emerged in the use of the language, imposing changes that the written norm rejected. The linguistic data that will be discussed in this section are, therefore, selected by adopting a precise choice: (i) texts reporting examples of "lower class" Latin-a stylistic variant typical of the not-educated classes, such as the letters of the soldier Claudius Terentianus (1st century BC); (ii) texts reporting examples of colloquial Latin-a register that educated classes used in non-formal contexts and well represented by Cicero's letters and Plautian's comedies.
The most important phenomenon emerging in the analysis of "lower class" and colloquial Latin is the phenomenon known as negative concord: the occurrence of multiple negative particles within the same sentence constitutes a single instance of negation 11 : These examples are taken from Plautus, one of the most copious authors in the use of this construct (Molinelli 1988), which can reasonably be attributed to the mimesis of the spoken language-typical of the comedy register. As the sentences (8a) and (8b) show, the negative concord phenomenon was not subject to any kind of restriction associated with the nature of the negative elements: both negative morphemes and indefinites can appear together, as well as conjunctions and complementizers 12 .
The negative concord phenomenon also occurs in some very old texts, both Italic texts (5a) (3rd century BC)-such as the Aes Rapinum (Pulgram 1978, p. 145)-and Latin texts (5b) (2nd-3rd century BC)-such as a little fragment that has been attributed to Marcius Vates 13 : Other examples were found in Ennius, Lucilius and Varro, showing that this phenomenon was already common in Old Latin and throughout the time span from the first to the third century B.C. (see Molinelli 1988). Consider now Claudius Terentianus's letter: (10) hic a[ut]em sene aer[e ni]hil fiet neque epistulae commandaticiae nihil valunt nesi si qui sibi aiutaveret (CLaSSES, CEL-I-142-259) 'Here nothing will be accomplished without money, and letters of recommendation have no value unless a man helps himself' In (10) the written language differed very little from the spoken one, showing some new grammatical constructions that were not accepted in literature. Crucially, the Classical Latin was also not immune to the negative concord phenomenon as witnessed by Cicero and Tibullus texts: The fact that Cicero used the negative concord construction was particularly important, since it could not represent a mimesis of the spoken language-as it was for Plautus-but, rather, a rhetorical emphasis that should not sound strange to the audience.
The incidence of the negative concord phenomenon significantly increased during the fourth century A.D., so much so that the grammarian Diomedes overtly criticized it: (12) modus soloecismi fit per geminationem abnuendi, ut si dicas 'numquam nihil peccavi' cum debeat dici 'numquam peccavi', quoniam duae abnutivae unam confirmationem faciunt. 'a type of solecism occurs with the negation doubling, that is, it is said "numquam nihil peccavi" intead of "numquam peccavi", since two negations yield an affirmative meaning' That a grammarian condemns this mistake testifies its widespread diffusion. Negative concord phenomenon tends to stabilize as a new normative construction, and not just as merely a phenomenon restricted to the language of use. Moreover, other grammarians reported the same kind of mistake, such as Nonius Marcello and Augustine, who attributed it to the Greek influence (see Rönsch 1965, p. 447;Ernout and Thomas 1953, p. 154). An explosion of sources testifying the negative concord diffusion starts from the fifth century AD, where the phenomenon became so pervasive that it lost the status of grammatical mistake (Molinelli 1988).
In summary, the shift from a double negation language to a negative concord one already occurred in Old and Classical Latin, even if it became pervasive only in the first centuries AD. In the next section, I will address a possible explanation for this shift by discussing the syntactic nature of the morpheme nōn, which, as seen above, it is undoubtedly the predominant form of negation.

The Head Status of the Morpheme Nōn
We saw in the previous sections that the possible interactions of the negative elements within the same sentence are two: (i) double negation constructions, that the grammatical norm prescribed; (ii) negative concord constructions, that the grammatical norm refused, but that the language of use and the colloquial style adopted. In recent literature these different outcomes have been traced back to the syntactic nature of the negative morpheme that is involved. In fact, simplifying the discussion in Zeijlstra (2004Zeijlstra ( , 2008, it can be assumed that if there is a syntactic negative head-which projects the structure of the negative phrase NegP-then negative concord construction occurs 14 ; on the other hand, if there is not a syntactic negative head-and, consequently, there is no a negative phrase NegP-then double negation construction occurs. Clearly, this leads to a kind of contradiction in the Latin system: nōn should be, at the same time, a negative head instantiating a case of negative concord construction, and a maximum projection, instantiating a case of double negation construction. To account for this anomaly, I will assume Gianollo's (2016) analysis, according to which the negative morpheme has only acquired a head status over time, as an effect of Jespersen's Cycle (the following scheme is adapted from Gianollo 2016): This hypothesis is coherent with a well-known tendency in the diachronic evolution of a language: elements classified as maximum projections often become lexical heads (van Gelderen 2004: Spec-to-head principle/Head Preference Principle) 15 . This change, and the more general reorganization of the syntactic constructions-such as the transition from the OV order to the VO one (Ledgeway 2012)-probably pushed Latin from being a double negation language, to being a negative concord one. Coherently, this shift in the negative paradigm should have first appeared in the colloquial contexts, while the linguistic norm must have endured much longer. This is precisely what the data reported in Section 2.2 seem to tell us. Crucially, this shift began in a distant period, when the language was still in an old form, as witnessed by the earliest attestations of the third century B.C. (see Section 5).
Another confirmation of the syntactic head status of nōn is its ability to form compounds, as witnessed by non-ne, an interrogative particle presupposing an affirmative answer: Nonne hac noctu nostra navis huc ex portu persico venit? (Plaut. Amph. 404) 'Didn't our ship arrive tonight from Port Persicus? b.
Nonne his vestigiis ad caput malefici perveniri solet? (Cic. S. Rosc. 73, 6) 'Is it not the case that one generally arrives to the starting point of a crime by following these traces?' The adverb nonne derived from the negative morpheme nōn and the clitic -ne, which also appears in other forms, such as num-ne, is an interrogative particle presupposing a negative answer. Assuming the standard hypothesis that clitics are heads and that they can only be joined to other heads (Kayne 1989), it follows that the negative morpheme nōn is a head too. Again, not surprisingly the occurrence of nonne was already attested in classical Latin, mostly in texts reproducing the spoken language-for example, comedies (14a)-and in texts where the rhetorical emphasis could legitimize strategies adopted in colloquial contexts (14b) (Ernout and Meillet 1959).
However, this shift in the syntactic status of nōn should have led to the progressive disappearance of all those phenomena typically associated with the double negation construction. This disappearance actually took place, but only with a certain degree of slowness and, more importantly, some data seem to be incoherent with that. The first discrepancy is the occurrence of nōn in negative questions such as why + negation? According to Merchant (2001), I will assume that only negative elements with a maximum projection status can appear in this kind of question, while negative elements with a head status are excluded 16 . Consider, for example, Modern English where the negative morpheme 'not' can appear in these constructs (Why not?) since it is an adverb (maximum projection), while in Modern Greek the negative morpheme dhen cannot, since it is a negative head (*Giati dhen?, let. "Why not?") 17 .
Coming back to the Latin negative morpheme nōn, it actually appeared within the interrogative 'why' questions, for example, those introduced by quōr, the original form of cur ("why"): (15) Quor non? (Plaut. Pseud. 318; Ter. Andr. 384) 'Why not?' Sentences such as (15) suggest, at first sight, a contrary conclusion for the head status of the morpheme nōn. However, it should be emphasized that the small number of attestations (five in the whole corpus represented in the Classical Latin Texts. A Resource Prepared by The Packard Humanities Institute) and their occurrence in only two authors do not seem to represent a definitive argument in this debate 18 . Crucially, this could register that oscillation between stage two and three of the Jespersen's Cycle, when nōn passed from an adverbial reinforcing element (maximum projection) to a head element.
To sum up, one of the most important shifts in the Latin negative system was the transition from the double negation system to the negative concord one. This was a consequence of the change in the syntactic status of the morpheme nōn, which acquired the status of a negative head as an effect of Jespersen's Cycle (van Gelderen 2004). The consequences of this change first appeared in the texts reproducing the colloquial language, then they spread over the Latin system. I will show in the next section that the head status of the negative morpheme is what Italian inherited since its origin.

Negation in the Old Italian System
The first trend in the diachronic process leading to the Old Italian is the relative simplification of the sentential negative system. In fact, only one negative morpheme survives, becoming ubiquitous in all grammatical structures, i.e., non. This morphological simplification did not reduce the plethora of the negative constructions populating the very rich system of Italian and ancient Romansh dialects 19 . In this paper I will just focus on the sentential negation, with particular attention to the Tuscan area, from which Modern Italian was born 20 .

Sentential Negation in the Tuscan Area
Texts from the Tuscan area are fundamental in order to collect all the linguistic changes that led to Modern Italian 21 . The centrality of non-and its phono-syntactic variants (ex. nonn in front of a vowel)-is the first fact emerging from the study of the ancient Tuscan documents. This clearly came from Latin, in which nōn substituted all the other negative markers (Section 1). Tuscan dialects also maintained the preverbal position of the negative morpheme, whereas some other Old Italian languages did not 22 . Consider, among many other documents, one of the earliest attestations of the vernacular Tuscan (twelfth century), i.e., the Laurentian Rhythm: Interestingly, Old Tuscan also inherited the negative concord system from Latin, where the negative morpheme non co-occurs with negative indefinites 23 . These indefinites can realize both adjective and pronominal functions; some examples are "neuno"/"niuno", (nobody/nothing), "nullo"/"nulla" (no/nothing) and "niente"/"neiente"/"neente" (nothing): (17)  Cases of negative concord constructions were very common and certainly they were not limited to any stylistic contexts, as it was in Latin. In fact, they constituted the 'neutral' grammatical norm. Crucially, the occurrence of negative concord phenomena also suggests that the Latin nōn has been transmitted to ancient Italian preserving its syntactic head status, a necessary condition in order to achieve this phenomenon. Moreover, the negative indefinites could also anticipate the negative morpheme, reproducing the same alternation affecting Latin: Whereas the co-occurrence of negative elements in Latin gives rise to double negation constructions-where the different order of the elements imposes a different semantic interpretation (see Section 3)-in old Tuscan it gives rise to a negative concord phenomenon and the position of the negative indefinites does not affect their semantic interpretation. Moreover, negative indefinite pronouns, adjectives and adverbs could also occur by themselves in a preverbal position, instantiating a case of sentential negation: The sentences (17-19) show a kind of continuousness in the occurrence of negative elements in old Tuscan: negative indefinites can occur either in isolation or with the negative marker non, both anticipating it and following it. This seems to challenge the clear partition between strict and non-strict negative concord languages (Giannakidou 1997(Giannakidou , 2000; see the note 14) 24 .
However, negative indefinites could also occur in a postverbal position with no negative marker anticipating them, maintaining an existential interpretation. Consider, for example, the interrogative sentences of the yes/no type (Zanuttini 2010, pp. 576-77)  Another phenomenon that Old Tuscan inherited from Latin is the use of words indicating small measures and quantities (the so-called minimizers) to reinforce the negative interpretation of the sentence (see the evolution of nihil in Section 1). One of the most widespread words was undoubtedly mica (etym. "crumb") that was already attested in Latin (mica (m)) (de Vaan 2008, p. 378). Quoting the words of Parry (2013, p. 80), the use of these objects has a "quantificational interpretation by being used idiomatically to express the lowest point on a pragmatic scale, that is, "not even a crumb" 26 (see Haspelmath 1997 for the original discussion on the pragmatics of minimizers): Thing that used to be bad and still are bad will not be forever'

An Emerging Phenomenon in Old Tuscan
Finally, old Tuscan developed a peculiar case of negation, i.e., Expletive Negation (EN) (see, among many others, Jespersen 1917;Horn 1989Horn , 2010Yoon 2011;Makri 2013;Greco 2021b), in which the negative marker non does not deny the propositional content of a sentence. According to Zanuttini (2010), there are two contexts in which EN occurs in old Italian: (i) temporal (22a) and comparative (22b) sentences and (ii) subordinate sentences depending on some verbal classes, such as fear, doubt, prevent, forbid, and deny. The negative marker in these contexts is absolutely optional and, in fact, the same sentences could omit it without changing the propositional meaning (22a'-b'): Interestingly enough, EN was available in Latin as well, but in a very limited number of contexts (Mari and Tahar 2020), such as the sentences introduced by verba timendi, and it was mainly realized by the morpheme ne (see Section 1): Crucially, only Old Tuscan developed a complex system of EN and it represents an innovation in comparison to Latin as well as to other Old Italian languages (Greco 2021b). Even though it is not clear the reason why EN develops in a language, I propose that the expletive use of negation depends on the syntactic status of the negative element: only syntactic heads implement the EN phenomenon and, crucially, Tuscan 'non' is a head-as witnessed by the rich occurrence of negative concord phenomena.
In order to evaluate this idea, consider, for instance, the case of the Modern French (Muller 1978;Makri 2013) and of the Late Middle English (van der Wurff 1999). Both languages display two negative elements, one with the head status and one with the maximal projection status, but only the former instantiates the EN structure: As it is well known (Kayne 1989;Pollock 1989;Zanuttini 1997), the French morphemes ne and pas constitute a single instance of negation by being generated in the same NegP: pas in (Spec, NegP) and ne in Neg 0 . Crucially, EN in the subordinate clause 'je η'ai ètè bien reçu' only displays the negative head ne, excluding the element with the maximal projection status pas. Similarly, the Late Middle English sentence displays two negative markers syntactically different: the adverb not with a maximal projection status, and the negative marker ne with a head status. Again, only ne realizes EN. All these facts seem to suggest that the syntactic nature of a negative operator is the key feature that allows EN, suggesting a new generalization: (26) only languages (and structures) displaying a negative head allow the expletive interpretation of negation, languages without negative heads do not.
This generalization moves in the same direction of Zeijlstra's (2011) observation that "there is no language without Negative Concord that exhibits a negative marker that is a syntactic head" (p. 136). From this point of view, two apparently distinct phenomena, i.e., negative concord and EN, seem to be the reflex of a single parameter: the syntactic nature of a negative element. Crucially, old Tuscan inherited the negative marker non from Latin, which displays a head status, and it shows both the phenomena: the negative concord and the expletive negation constructions. In Section 4.3, I will attempt a possible explanation for the EN phenomenon.
To summarize, we saw in this section that many phenomena affecting the sentential negation in Old Italian are inherited from Latin: the head status of non and its (fixed) preverbal position, the negative concord phenomena-that was not limited to any stylistic contexts, as it was in Latin, the use of negative indefinites in isolation; and the use of minimizers to reinforce the negative interpretation of a sentence. On the other hand, some other phenomena were peculiar with Old Tuscan, for instance, the expletive use of negation and the disappearance of all the double negation phenomena. I show that these two facts are just the consequence of the head status of non, which started in Latin, but it reached its conclusion in Old Italian.

Negation in the Modern Italian System
We have now reached the last step in the historical evolution of negation. Modern Italian preserves many phenomena coming from the old Tuscan; however, it also displays some innovations. Some of them are consistent with the canonical evolution described by Jespersen's Cycle, some others are not.

The Old Italian Inheritance
The first phenomenon that Modern Italian inherited from the old forms is the exclusive use of the negative morpheme non. As in the Old Tuscan, negation always occurs before the inflected verb and no element, but clitics can occur between them 27 : According to Zanuttini (1997), I assume that the position occupied by non is the highest one that a negative morpheme can occupy within a sentence in Romance languages 28 .
Another phenomenon that Modern Italian inherited from Old Italian is the negative concord constructions, which represent the linguistic norm. This definitively confirms the syntactic nature of non as a negative head 29 : As the sentence (b) shows, a second strategy to deny a sentence is by means of the negative indefinite pronouns, adjectives, and adverbs. Again, this possibility depends on the position of the negative indefinites 30 , which can deny the sentence only if it precedes the inflected verb, as the Old Tuscan did: Modern Italian can therefore be considered a negative concord language of the nonstrict type (Giannakidou 2000; see note 14).
The similarities with the Old Italian system do not stop there. Even in Modern Italian, negative indefinites can appear in a postverbal position, with no negative element anticipating them and maintaining-as in the first documents of Tuscan area-an existential interpretation, such as in yes/no questions 32 : Moreover, Modern Italian also replicates the behavior of Old Italian when it reinforces the negative meaning of a sentence thanks to minimizers and polarity items. The most common expressions are affatto ('at all'), mica (etym. 'crumb'), idiomatic expressions such as alzare un dito ('lift a finger'), and some profanities such as cavolo/cazzo (lit. 'cabbage'/'dick'): I recall that mica was already used in Latin (Orlandini and Poccetti 2012) and it spans all the centuries up to modern use, in which it can even appear in a preverbal position, negating a sentence by itself (Mica l'ha visto! Let. mica CL.it.has seen; 'S/he did not see at all!'). This kind of reorganization from a reinforced element to a sentential negative one is compatible with stage three of Jespersen' Cycle. However, it should be noted that mica also adds a pragmatic nuance that lacks in the use of non and, therefore, these two elements are not completely interchangeable 33 . Moreover, mica seems unable to license neg-words in its scope-contrary to what standard negation usually does-challenging the regular outcome of the stage three of Jespersen' Cycle 34 : (33) a. *Mica ha mangiato niente mica has eaten nothing b.
Non ha mangiato niente neg has eaten noting 'S/he has not eaten enithing'

A Notable Innovation
Finally, Modern Italian also inherited another use of negation, i.e., expletive negation. As in Old Italian, temporal and comparative sentences host EN. However, Modern Italian also shows many new contexts in which EN can appear. They can be both subordinate and root clauses (see Greco (2019Greco ( , 2020b and the references therein for the grammaticality judgments). Starting from the subordinate ones, EN can be introduced (i) by a verb as in interrogatives (34a), (ii) by conjunctions such as "piuttosto che" ("rather than") (34b), "finché" ("until") (34c), "a meno che" ("unless") (34d) and "prima che" ("before") (34e), (iii) by "chissà" ("who knows") (34f), and (iv) by a comparative 35 (34g) In all these structures negation is expletive and, therefore, it does not reverse the polarity of the proposition (as the English translation shows). Consider now EN in root clauses; it can occur in (v) negative exclamatives (35a); in (vi) negative rhetorical questions (35b); in (vii) not-that clauses (35c); and in (viii) surprise negation sentences (Greco 2020a As already anticipated in Section 2.2., it is not clear the reason why a language develops the expletive interpretation of negation, but this surely depends on the head status of the negative morpheme. Crucially, even though the EN in matrix and subordinate clauses realizes a unique phenomenon in which the negative marker does not deny the sentence in which it occurs 36 , Italian distinguishes two different EN classes showing different syntactic behaviors (Greco 2019(Greco , 2020b. Consider the case in which EN co-occurs with a reinforcing element-such as alzare un dito ("lift a finger"; see sentence 32) and with neg-words-as nessuno ('n-body') 37 -in temporal clauses and in exclamatives: Temporal clauses allow the occurrence of both the reinforcing element (technically, a weak negative polarity item) and the neg-word, whereas exclamatives do not. Starting from distributional and syntactic differences such as this, Greco (2019Greco ( , 2020b proposes that EN structures can be either weak or strong depending on whether they maintain some features typically associated with standard negation (for example, allowing weak-NPIs) or not. Applying this label to our example, it follows that until-clauses fall into the weak EN class and exclamatives into the strong one. Crucially, Greco (2019) tested all the EN clauses seen above with regard to several polarity-sensitive elements (weak/strong-NPIs, not-also conjunction, and Neg-words) confirming the twofold classification of ENs (Table 1).

A Small Remark on the Head Status of Non
We saw in Section 2.2 that only languages (and structures) displaying a negative head allow the expletive interpretation of negation, languages without negative heads do not. We also saw that the same kind of head-requirement is mandatory in the distribution of the negative concord phenomenon (Zeijlstra 2011). Interestingly enough, the head status of non also allows it to select different kind of arguments. Of course, the negative head non can select the tense phrase (Belletti 1990;Zanuttini 1997;Poletto 2008), as we saw in (21). Moreover, according to Greco (2020aGreco ( , 2021a, non can also be merged in the CP-domain (à la Laka 1990)-when the v*P-phase has already been closed-instantiating a case of strong EN 40 . Consider, among many other examples, the case of exclamatives. Exclamatives show a twofold interpretation: one in which negation is expletive (37a) and one in which it is standard (37b). In Greco's (2021a) work 41 the former was labeled "Expletive Negation Exclamative" (ENE), and the latter "Negative Exclamative" (NE): The two structures differ grammatically. According to Grimshaw (1979) and Zanuttini and Portner (2003), exclamatives are factive and, therefore, they can only be embedded under factive predicates. However, focusing on a specific sub-class of factive predicates, i.e., to know-verbs, only the NE interpretation is possible, and the ENEs one is ruled out: The expletive reading of negation in (38b) is completely ruled out, whereas the standard one is preserved. According to Greco (2021a), a possible way to take into consideration the differences between NEs and ENEs is to assume a twofold derivation of negation: when the negative marker not is merged in the TP-domain, it gives the standard negation reading, as in a negative exclamative; when it is merged in a higher position, i.e., the CP-domain, it gives the expletive negation reading as in expletive negative exclamatives-(phases are underlined) 42 . Crucially, the high position of negation in ENEs can also explain why they cannot occur under factive predicates, as with to know verbs in (38b). More specifically, it has been proposed (cfr. Grewendorf 2002;Haegeman 2004Haegeman , 2012) that some factive verbs select a reduced CP, leaving no space for several functional phrases, including, arguably, negation. If this is true, that means that the only available option for negation in exclamatives under to-know verbs is to be in the TP-domain, realizing the standard value of negation as (31b) shows.
Finally, the syntactic schema in (39) also takes into account some differences between weak and strong EN clauses, such as the difference between temporal and exclamative clauses in allowing NPIs and neg-words. I recall that EN in temporal clauses allows them, whereas EN in exclamatives does not (see sentences in (36)). According to (39b), when non is merged in the CP-field in exclamatives, the domain of the vP is impenetrable and, therefore, negation cannot see inside it (see Chomsky 2004Chomsky , 2008Chomsky , 2013. From this configuration it follows that ENEs cannot host NPIs and neg-words since, according to Giannakidou (1997) and Zeijlstra (2004), a negative operator must bind all the free variables in the vP domain in order to allow them, and negation in ENEs loses this possibility because it belongs to another phase, namely the CP-phase.

Concluding Remarks
This article aims at investigating some diachronic aspects of the Italian negative system, considering a time span ranging from Old Latin to Modern Italian. Most of the negative polarity phenomena populating the Modern Italian system are consequences of a crucial change that occurred in Old Latin: The Latin negative morpheme nōn ("not"), which initially displayed a maximal projection status, became a syntactic (negative) head. This change occurred in the earliest attestations of Latin (I-III century BC), particularly in texts reporting examples of "lower class" Latin-a stylistic variant typical of the not-educated classes, such as the letters of the soldier Claudius Terentianus (1st century BC) and in texts reporting examples of colloquial Latin-a register that educated classes used in non-formal contexts and well represented by Cicero's letters and Plautian's comedies. Crucially, the shift from a maximal projection status to a head status also caused the shift from a double negation system-where the co-occurrence of two negative elements generates an affirmative meaning-to a negative concord one-where the co-occurrence of two, or more, negative elements constituted a single instance of negation, which affects many Romance languages (and their dialects). This was the result of those changes that responded to generalizations, such as those formalized as Jespersen's Cycle and Spec-to-head principle/Head Preference Principle. It also determined the availability of the expletive reading of negation in Old and Modern Italian, as well as in other Romance languages (ex. French), calling for a new generalization: only languages (and structures) displaying a negative head allow the expletive interpretation of negation, languages displaying a maximal projection status do not. Modern Italian also develops two classes of expletive negation sentences, respectively the strong and the weak class. It has also been proposed that the head status of non also allows it to select different kind of arguments, making some predictions on the status of exclamative clauses: when the negative head not is merged in the TP-domain, it gives the standard negation reading as in negative exclamative; when it is merged in a higher position, i.e., the CP-domain, it gives the expletive negation reading as in expletive negative exclamatives.

Conflicts of Interest:
The authors declare no conflict of interest.
Notes 1 According to Ernout and Thomas (1953, p. 148), there were some reinforced forms of nē, i.e., nec and nī. I will not take theme into consideration in this paper since they are not essential for the present discussion. See Orlandini and Poccetti (2008) for a discussion on the origin of nec and its development in ancient Italian. 2 In subordinate clauses introduced by factual verbs is used ut non (e.g., facio ut non), and in those clauses introduced by some verbs with a negative meaning, quin is used (e.g., non dubito quin). Oniga (2014, pp. 272-73) describes the system. The conjunction quīn derives from *quī-ne. 3 See Hackstein (2016) for the etymological derivation of haud.

4
Consider that Plautus used haud to deny simple sentences as well (cfr. Lindsay 1907, pp. 130-31). 5 See Pinkster (2015, chp. 8) for a detailed review of all negative constituents and their uses in Latin. For example, in this article I do not choose to mention the negative morpheme nȇ, which has been replaced by nōn, leaving the only traces in compounds such as nihil (nȇ + hilum), nullus (nȇ + ullus), etc. Moreover, I will not address the cases of intrinsically negative verbs as well, such as nȇ + scio (cfr. Pinkster 2015), just focusing on the negative sentential constructs. 6 In this chapter I will often refer to Jespersen's Circle and I will discuss it in a more detailed way in the next sections. However, addressing a full discussion is beyond the goal of this paper. For a detailed discussion on the effect of Jespersen's Cycle in Latin see, among many others, Ernout and Thomas (1953), de Vaan (2008), Fruyt (2011), Orlandini and Poccetti (2012, and Gianollo (2016). 7 Some (rare) uses of nōn with the subjunctive mode are attested; for example, Rhet. Her. 2, 41: Si ad exercitum non uenisset (lit. if to army.Acc neg come.Subj. pluperfect.3rdSg; "If he were not come to the army . . . ."). 8 It may be possible to consider as exceptional some interrogative clauses introduced by the clitic particle -ne: Vidisti-ne fratrem Chaeream? 'Have you seen Chaeream?' (Ter. Eun. 713). It has been argued (de Vaan 2008, p. 403) that the clitic particle originated from a negative root, although it lost its semantics. If this were true, it would represent an idiosyncrasy of the Latin SOV system. As Ledgeway noted (2012, p. 221), "typological investigations have revealed that, whether as a prefix or an independent word, SVO languages most typically display preverbal negation, whereas SOV languages commonly show postverbal negation. Within this perspective, the preverbal position of Latin negation, whether as an independent word or as an incorporated prefix, thus proves entirely consistent with a head-initial typology." For a detailed discussion on the transition from Latin to the neo-Latin languages, see also Tagliavini (1969); for a general discussion on the position of negation within a sentence, see Bernini and Ramat (1996), Zanuttini (1997), and Poletto (2008. 9 See Oniga (2014, p. 264) for the affirmative interpretation of this sentence due to the interaction of the two negative items. 10 Technically, one element scopes over the other. The negation scope indicates the portion of the sentence on which it operates and depends on many factors, including the phrasal structure. For a detailed discussion see Chierchia and McConnell-Ginet (2000). 11 For a detailed discussion on the negative concord phenomenon, see, among many others, Mathesius (1937), Molinelli (1988), Zeijlstra (2004Zeijlstra ( , 2008, Torrego (2009), Pinkster (2015, and van der Auwera and Alsenoy (2016). 12 According to Pinkster (2015, chp. 8), the negative concord phenomenon could either be pleonastic or it could strengthen the negative interpretation of a sentence. Both these cases occur in (8). Crucially, only the pleonastic function survives in the Romance languages. 13 See Molinelli (1988) for a discussion on the interpretation of these sentences. 14 Negative concord constructions can be either strict or non-strict (Giannakidou 1997(Giannakidou , 2000Zeijlstra 2004). In the former case, the negative morpheme is mandatory, in the latter case, it is the opposite. For example, among Romance languages, Romanian falls into the strict type, since the negative morpheme nu must appear with the indefinite negatives, not allowing them to realize the sentential negation (ia-a') on their own; Italian falls into the non-strict type, since the negative morpheme cannot occur with negative indefinites if they are in a preverbal position, where they negate the sentence by themselves (Section 3): An interesting parallelism can be found in the history of negation in French: from the original form ne (stage 1), it moved to a reinforced form ne pas (stage 2), to then arrive at the new simplified form pas, common in spoken language (see Kayne 1975). According to some important works in the field (see, among many others, Kayne 1989;Pollock 1989;Zanuttini 1997), the French morphemes ne and pas constitute a single instance of negation by being generated in the same NegP: pas in (Spec, NegP) and ne in Neg 0 . See Section 3.2 for a more detailed discussion. 16 See the original work for the detailed discussion behind this proposal. 17 In these cases, languages display a maximum-projection negative element that can appear in such contexts. In the case of Modern Greek it is oxi: Giati oxi? (let. "Why not?") (cf. Merchant 2001). 18 In order to provide comparative data, the interrogative adverb Quidni? ("Why not?") occurs 43 times and 8 times in the disjoint form Quid ni? within the same corpus. 19 See Parry (1996), Zanuttini (1997) and Manzini and Savoia (2005) for a very detailed discussion on negation in Italian and Romansh dialects. 20 See Greco (2021b) for a detailed discussion on negation in Old Italian dialects. 21 In this paragraph I will often present data from two sources: the entry on negation written by Raffaella Zanuttini (cfr. Grammatica dell'italiano antico, edited by Salvi and Renzi 2010) and the Corpus OVI dell'italiano antico (Corpus OVI dell'italiano antico 2020 and cfr. Dardano 2013). 22 See Contini (1941) for examples of negation in different positions in the sentential word order. 23 According to Zanuttini (2010), negation and negative indefinites may occur within the same sentence as well as in different ones. At least two cases should be considered: non belongs to the matrix clause and the negative indefinites belong either to the subordinate clause or to the relative one. See the original work for the linguistic data. 24 As an anonymous reviewer pointed out, these data are compatible with a transition from a strict to a non-strict negative system, in line with what has been proposed for languages such as Catalan (van der Auwera and Alsenoy 2016). However, according to Garzonio and Poletto (2012), there are some good reasons-both distributional and syntactic-to hypostasize that this is not the case. The alternation in the negative concord system would just be a consequence of the syntactic positions in which the negative indefinitives are moved. See the original work for a full analysis. 25 The distribution of words such as niuno in Old Italian-or nessuno in Modern Italian-has been investigated by many scholars. I will refer to some works, among many others, which propose different, and often alternative, analyses. See (Ladusaw 1992;van der Wouden and Zwarts 1993;Haegeman and Zanuttini 1996;Giannakidou 2000;Herburger 2001;Zeijlstra 2011;Poletto 2020b). According to Martins (2001) and Poletto (2014), for example, there are weak negative polarity items (NPI), which can be licensed by yes/no questions and conditional, as well as by negative markers. According to Giannakidou and Yoon (2010) and to Giannakidou (2011), an NPI can be either strong or weak depending on whether it is only licensed by a negative marker in the scope of an anti-veridical operator, or not. The veridicality of an operator is definite in the following way (cfr. Giannakidou 2006): (i) A propositional operator F is veridical if Fp entails or presupposes that p is true in some individuals' epistemic model M E (x); otherwise F is nonveridical; (ii). A nonveridical operator F is antiveridical if Fp entails that p is not true in some individuals' epistemic model: Fp→¬p in some M E (x). Put in different words, a veridical/non-veridical/anti-veridical operator measures the speaker's epistemic attitude toward the truth of an expression: respectively, s/he can be sure of the truth of it (ex. factive structure), uncertain (ex. questions or conditionals) or sure of the falsity of it (as with negation). From this point of view, strong NPIs can only occur in negative sentences, because they require an anti-veridical context. Coherently, they cannot occur in questions or in a protasis of a conditional clause because they are non-veridical operators, whereas weak NPIs can. 26 As an anonymous reviewer pointed out, the etymological derivation of mica and its semantic contribution diverge. In accordance with the literature, I will assume that mica has been generated as a minimizer already in Latin (Orlandini and Poccetti 2012), but lost its nominal properties (Manzini and Savoia 2002), involving the presupposition that the negated event was expected to happen in Modern Italian (Cinque 1976;Squartini 2017). According to Zanuttini (2010), the original strengthened value of mica is still visible in Old Italian, as the following sentence shows (see Garzonio and Poletto 2010 for a theory on the derivation of minimizers in Old Italian): (ii) "Chi sete voi" disse messer. T. "che di rimanere con voi tanto ci pregate?" who are you said sir T. that of stay with you very much Cl.us to beg "Certo, sire" disse elli "io non ve lo celerò mica" (in Zanuttini 2010, p. 572) of course sire said he I neg cl.to-you cl.it hide.Fut.3rd.Sg neg '"Who are you?" sir T. said "that beg us to stay with you very much" "Of course, sire", he said, "I will not hide it from you at all"' According to Garzonio and Poletto (2012) mica can only appear in a postverbal position-the only exception is when it is raised to a preverbal position in a cluster with negation (ex. with né or non)-and always displays negative concord. See the original works for a detailed discussion. 27 There are some exceptions to this pattern: non instantiates a case of constituent negation ('Ti ho detto di chiamare Luca, non Maria!' Eng. 'I told you to call Luke, not Mary") and it appears in structures with verbal elision ('Mi raccomando, non (dire) una parola', Eng. 'I recommend, don't (say) a word'). However, these uses do not weaken the idea that non has a proclitic function on verbs, as all the case of sentential negation shows (see Manzotti and Rigamonti 1991). 28 According to Zanuttini (1997), Italian dialects display four distinct positions where a negative marker can occur. She determines those positions on the basis of their distribution with regard to inflected verbs, past participles, and lower adverbs. The four typologies of negative markers represent four different NegPs located in as many places in the sentential structure from the highest one-which selects the TP-to the lowest one. The standard Italian negative marker non represents an instance of the highest one, which occurs in a pre-verbal position and denies a sentence by itself. Poletto (2008) shows that there is a parallelism between the syntactic distribution of the four types of negation and their etymological origins. See the originals works for a detailed discussion on this topic. 29 As a further proof, consider the impossibility of non to occur in why questions (iia) where Italian displays an element with a maximal projection status (iib). For a discussion on the negative head status of non, see Greco (2020b)'s work: (iii) a.
Perchè no? why not 'Why not?' 30 The syntactic and semantic status of negative indefinites is greatly debated in literature but it is beyond the aim of this work. I will refer to some works, among many others, which propose different, and often alternative, analyses. See (Ladusaw 1992;van der Wouden and Zwarts 1993;Zanuttini 1996, Giannakidou 2000;Herburger 2001;Zeijlstra 2011;Poletto 2020b). See also footnote 25. 31 It is worth knowing that these sentences are grammatical for some Italian speakers, but just in contexts where focalization and topicalization are involved. From this point of view, the co-occurrence of non and negative indefinites yields an affirmative meaning via a double negation mechanism: (iiv) Speaker A: Forse qualcuno non ha visto il film.
'Perhaps, someone did not watch the movie.' Speaker B: NESSUNO non ha visto il film nobody neg has watched the movie 'Everybody watched the movie' (It is not the case that nobody watched the movie) These cases are rare and restricted to the speech only. 32 See note (25) for the syntactic and semantic status of elements such as nessuno (nobody). 33 See (Cinque 1976;Frana and Rawlins 2015;Squartini 2017) for a discussion on mica. It has been argued that it denies the presupposition of a sentence rather than the proposition implicated by a sentence. See also footnote 26. 34 An anonymous reviewer properly pointed out that the preverbal position of mica does not seem to correspond to a third stage of Jespersen's Cycle. In fact, mica is not able to trigger, among other elements, strong-NPIs, such as affatto ('at all'), which require to occur in a negative sentence (see Greco 2020b): (v) Mica ha mangiate *affatto La pizza! mica has eaten At all the pizza 'S/he has not eaten the pizza (as you thought!)' This is also the reason why mica cannot substitute a standard negation in a sentence, as the final stage of Jespersen's Cycle predicts (see Batllori 2016 for similar discussion on the Cataln mica, that has been proposed to miss an uninterpretable (uNeg) feature). 35 However, EN cannot occur in subordinate sentences depending on some verbal classes, such as fear, doubt, prevent, forbid, and deny, as it happened in Old Italian: For a full discussion on the nature of elements as nessuno (n-body), see Zeijlstra (2004) and the references cited there. According to this work, neg-words should not be treated as NPIs, but as "non-negative elements that are syntactically marked for negation, i.e., they carry an uninterpretable [uNEG] feature that needs to be checked against a semantically negative operator carrying [iNEG]" (Zeijlstra 2004, p. 236). I will not analyze the huge discussion on neg-words here, leaving this goal to the original references (see Laka 1990 as the pioneering work on this issue). 38 Some Italian speakers do not accept this sentence because of "nessuno", but they accept it if is changed with "qualcuno" (someone). Differences in the grammaticality judgments are often associated with EN, particularly in languages showing the same negative marker in both expletive and standard negation contexts (see Tubau et al. 2017 for Catalan and Greco Matteo et al. 2020 for Italian). See Greco (2019) for the grammaticality judgments of this sentence. 39 See Delfitto and Fiorin (2014) for grammaticality judgments. 40 From this point of view, it is extremely convenient to consider expletive and standard negation as the result of different syntactic representations involving the same morphological element: when the negative head non is merged in the TP-domain, it gives the standard negation reading; when it is merged in a higher position, i.e., the CP-field, it gives the expletive negation reading. Therefore, the negative marker is always the expression of the unique functional word associated with negation, it just has different interpretations. 41 Many works focused on exclamatives, among many others, see Zanuttini and Portner (2003); Delfitto and Fiorin (2014), and the references cited there. 42 Such an interaction between negation and syntax seems consistent in other languages as well as in other structures. For example, in Modern English, according to Tubau (2020), having negation first merged either in a TP-internal position or in a TP-external one gives some crucial contrasts, as witnessed by polarity-reversing question tags, neither/so-coordination, either/too adverbs, etc. Moreover, it has also been argued that a distinction between low and high negation is the key to understanding the structure of yes-no questions and their response particles (see Holmberg 2016;Wiltschko 2017).