2.1 Previous accounts

The position of negative markers in Romance has been examined in several works by Zanuttini (1997), Garzonio (2008 and subs.), Manzini & Savoia (2011), Poletto (2020). In particular, Poletto claims that each type of negative marker (such as minimizers, preverbal negation, indefinites etc.) occupies a specific functional projection inside a ‘negative circuit’. These positions depend on the etymological source from which the element originated. In turn, depending on its position in the negative circuit, each negator will show different interpretive and syntactic properties. Lastly, these negative elements can move elsewhere in the clausal spine from their original merging site depending on what other features they carry. Mica, for example, is said to originate in a functional projection inside TP, but can optionally move to CP, where it surfaces as clause-initial. We adopt Poletto’s label “Mica preposing” to describe this phenomenon. Based on prosodic evidence, Poletto shows that fronted mica must indeed land in a C-position: mica cannot be the direct replacement of non which is merged as a clitic in TP; instead, mica is a fully-fledged phrasal element, probably projecting its own phrase and not being able to be hosted in the clitic slot. This is also compatible with the fact that both mica and non can appear simultaneously, yielding a double negation interpretation.⁴

Thaler (2018) presents preposed mica as used to convey a stronger emotive involvement. This effect was also described in more formal terms in Squartini (2017), by resorting to the notion of mirativity (Aikhenvald 2012). However, this effect of mirativity and strong emotive involvement might be the consequence of the semantic value of mica, which we will define later in the discussion. Pescarini & Penello (2008: 51) point out very interesting semantic differences between the preposed mica and the clause-internal one. Based on scalar implicatures, they argue that the two constructions have a different nature.

When we take pragmatic scales like numbers, the denial of a point with internal mica can either imply the denial of all smaller elements (in a downward entailing fashion) or simply be the denial of the exact point of the scale. When mica is preposed, only the last interpretation is available. They propose that the second alternative implicature is not available in (11), because with preposed mica there is an operator in CP which creates a conflict with the denial of the entire scale. This difference is very interesting, but needs further experimental support to confirm it (cf. Panizza et al. 2009). To support the different nature of the two mica, they also provide evidence from a northern Italian dialects which displays two morphological variants depending on the position of the negator.

Although this morphological alternation seems convincing, caution is needed, inasmuch as the alternation can also be attributed to prosodic requirements in the metrical structure (cf. Poletto’s prosodic point).

Cerruti (2018; 2020) conducted a sociolinguistic study on the distribution of Italian mica in the CORIS corpus (Favretti et al. 2002). Among other things, Cerruti notices that the bipartite construction NEG+V+MICA can encode the denial of both discourse-old and discourse-new information (77,8% of all instances), he also included those cases where the preverbal negator could be dropped (18,75%), in sentences like ho mica fame ‘I am MICA hungry’, although he considered these cases as not belonging to spoken and written standard, mainly because they were fixed formulae and constituent negations. What is crucial for the present discussion is that the construction MICA+V (4,17%), i.e. the preposed mica, was found only in the denial of discourse-old information, as opposed to the bipartite construction, which can scope over less explicit predicates. This finding is line with Cinque’s intuition, where preposed mica lexicalizes exclusively the denial of active information. To conclude, previous research has noticed the particular behaviour of preposed mica, suggesting that it must be handled as another construction conveying specific meaning. As noticed by Cinque and Cerruti, preposed mica is licensed with explicitly old propositions. With this in mind, in next section, we will corroborate such hypothesis and show that preposed mica requires indeed stricter pragmatic activation in order to be licensed.

2.2 A note on diatopic variation

Besides semantic and pragmatic distinctions for the alternation between fronted mica and the clause-internal one, recent literature has also emphasised the role of diatopic variation. For example, Ledgeway (2020) shows that northern Italian dialects and northern regional Italian employ mica in a post-verbal position, whereas southern varieties prefer to place mica at the beginning of the sentence. The intuition behind this divide lies behind the type of verb movement and the grammaticalisation status of the negator. Similarly, Schifano (2018) builds a taxonomy of verb movement in Italo-Romance by deploying mica as one diagnostic test (more on this in section 4.1.1). In short, these constructions can be geographically marked. In this paper, we will not focus on these diatopically-marked varieties, and we will show that both preposed mica and internal mica co-exist within Standard Italian (as Cinque does). In other words, despite regional preferences, which are emerging from vernacular varieties in the process of the formation of regional varieties (Cerruti et al. 2017), we will empirically support the hypothesis that both constructions can be found in Standard Italian (à la Cerruti et al. 2017). To do so, we plan to collect acceptability judgements from various regions and control both the area of informants and their dialectal usage. The fact that both constructions can co-exist can be already appreciated in data from italiano giornalistico (‘newspaper Italian’), as shown by the following examples.

Equally, in Cerruti’s (2020) data, both constructions were attested by speakers coming from Milan, Florence, Rome, Naples. We can replicate similar results by using other corpora. In the KIParla corpus, containing data from spoken Italian, we find sentences like (16) and (17) (Mauri et al. 2019). Similar results are detectable from the written corpora LIP and CORIS/CODIS (Favretti et al. 2002), which contains written data from press, novels, and essays.

Special attention on the possible diatopic variation will be dedicated to the acceptability rating test (section 3), where we take as informants those speakers who declared scarce or null dialectal usage. This methodological choice was driven by the need to limit possible preferences brought by the dialect and to directly measure the effect of the pragmatic conditions. However, we also included the geographic origin of the speakers in our model, but it failed to show as large effects as the pragmatic condition (see later discussion).

3.2 Results

We report the raw results from the survey in Figure 1, which is a Likert plot made with the R package sjPlot (Lüdecke 2021). The plot displays the rating frequency of each item, distinguished in each row, each colour represents a different acceptability rate. Besides the first and last three rows, the name of experimental items are shortened for ease of graphic representation. In the plot, M and U stand respectively for an item with Mentioned proposition and Unmentioned one. The second letter describes the position of mica, whereas I stands for clause-Internal position and P for the Preposed one.

Figure 1

Acceptability of target sentences.

As expected, between 92% and 98% found the ceiling sentences as completely acceptable and assigned an acceptance value of 6 or 7. Vice versa, word salads were found completely unacceptable, obtaining 1 or 2.

Concerning the target dialogues, when mica scoped over mentioned propositions, the sentence was rated as acceptable. The situation becomes more complex when looking at the items with the unmentioned propositions. When mica is clause-internal the rates are less clear-cut and more medial: the sentences were not judged acceptable as the previous but do not either reach the floor levels of word salads. This fuzzy zone might be the consequence of the optional accommodation that the speaker might attribute to the denial. If the speaker assumes that the proposition p was activated in previous turns, other conversations, encyclopedic knowledge etc., then the sentence could be accepted more easily. We will cover this effect later in the discussion. The acceptability dropped down to the floor levels when the proposition was unmentioned and mica was fronted, where the majority of speakers (78%–94%) rated the sentences as unacceptable. In order to predict the acceptability of the sentence as a combination of both positions and informational status, we ran a generalised mixed model with the interaction of the two conditions.

An ordinal regression was run in R using the package ordinal (Christensen 2019) after checking the parallel regression assumption with the package brant (Schlegel & Steenbergen 2020). The model building started with the maximal random effects structure (participants and items) but failed to converge. Hence, different models were computed with a step-wise procedure (see Appendix for the procedure). The best model, decided by taking the lowest Akaike Information Criterion (AIC) with an Anova comparison, was the one comprising the interaction of the predicate activation and position of mica as fixed effects, and the participant as random effect (model ← clmm (value ∼ Inform*position+ (1|participant))).

Interestingly, the model including the dialectal area did not yield any significant improvement in explaining the variance. First, there was no significant difference (p = 0.49) between the model specifying the dialectal area (AIC = 2976,2, Log likelihood = –1472.1) and the one without it (AIC = 2969,6, Log likelihood = –1474.8). Secondly, no specific correlation emerged in the model including the dialectal area.⁷

Within the best fit, the difference between the intercept, set at internal mica with mentioned proposition, and the ones with the same position but unmentioned p is statistically significant and negative (β = –2,39, Std.Error = 0,18, z = –13,13, 95% CI [–2,75, –2.03], p < .001), showing that with unmentioned propositions, the acceptability rate drops. The effect of the preposed position worsened even more the acceptability of mica in unmentioned p (β = –2,22, Std.Error = 0,25, z = –8,79, 95% CI [–2,71, –1,72], p < .001). On the contrary, there is a significant positive correlation between preposed mica and mentioned propositions, in comparison to the intercept (β = 0,6, Std.Error = 0,18, z = 3,31, 95% CI [0,24–0,97], p < .001). To sum up, the results reflect the tendencies appreciable in Figure 1, when mica is clause-internal and scopes over mentioned proposition, the sentence is perfectly acceptable (median = 6,5, standard deviation = 1,58). The acceptance rate increases slightly when mica is preposed with active predicates. However, when mica is used as negation of unmentioned predicates, the acceptance rate significantly drops, especially with preposed mica (median = 2, standard deviation = 1,42).

Our data confirm Cinque’s intuition that preposed mica requires informational activation of the predicate, whereas internal mica has looser information-structure requirements that can be met by accommodation. If the same lexical item has distinct pragmatic requirements in different surface positions, then the difference might be attributed to the syntactic mechanism. With this in mind, we will now turn to the crucial part of the paper: a syntactic analysis of preposed mica which could account for this pragmatic alternation.

4.1 Preposed mica as corrective focus

We argue that preposed mica is parallel to another syntactic operation available in Standard Italian: focus fronting (Bentley 2010; Cruschina 2012; Bianchi et al. 2015; Cruschina & Remberger 2017). It is commonly accepted that a fronted focus constituent in Standard Italian bears a corrective interpretation (24) (Rizzi 1997), whereas an informational focus must appear postverbally (25) (Cruschina 2021 and references therein). In other words, non-corrective focus constituent cannot undergo focus fronting in Standard Italian, corrective focus is free to surface both internally and fronted.

The corrective focus in (24) has been defined by Cruschina (2012) (but see also references therein) and Bianchi & Bocci (2012) in pragmatic terms as a correction to a previously asserted proposition. They propose that the focused constituent is not compatible with what has been asserted before and the alternative under focus is the one that should be updated to the Common Ground (instead of the antecedent alternative presented previously in the context, but see also section 5).

Similarly, preposed mica can be considered as an instance of corrective focus, as it corrects an active proposition in the discourse, showing that the latter is incompatible with speaker’s knowledge and should not belong to the Common Ground. This is in line with Frana & Rawlins’s (2016) semantic definition of mica, a perspectival FALSUM operator. Note that the same effect can be potentially reached without fronting (as it may happen, in general, with the corrective focus in situ). We argue that internal mica encodes a broader contrastive focus, which can then be specified in the corrective focus if the appropriate conditions are met, whereas the preposed one is specific for correction. We will return to this dichotomy later in the discussion.

Back to this parallelism, Cruschina (2012) already introduced the idea to extend focus-fronting to other elements, in particular to quantifiers (in his work, QP-fronting), also reviewing some other case studies with fronted constituents. For example, Ambar (1999) posits an ad-hoc functional projection for fronted constituents, an Evaluative Phrase, which expresses the speaker’s evaluation on a given state of affairs. Similarly, Molnár (2002) proposes a KonstrastP for Finnish. Hence, extending the availability of a higher corrective projection, reachable by various categories like adverbs does not seem so hazardous and we can proceed further.

We will now look at the framework proposed by Bianchi et al. (2015) in more detail and then we will try to apply it to the specific case of mica. In particular, for fronted focus, they proposed a cartographic explanation (Rizzi 1997; Samo 2019 for a recent review), employing a dedicated left-peripheral slot, a Focus Associated Implicature Projection (FAIP). A FAI head syntactically encodes a conventionalized implicature (in the sense of Grice 1975) on the type of involved focus. Following the characterization given by Grice, conventionalized implicatures are part of the conventional (lexical) meaning of words (for example, from the only temporal meaning of since to the implicature of cause Geis & Zwicky (1971). Bianchi et al. argued that a corrective implicature can also be conventionalized in a syntactic construction, rather than a word.

FAIP could be seen as another instance of a syntacticization of discourse, where the association between a linguistic form and its pragmatic effect is encoded in specific syntactic operations. Potts (2007) reviewed and tried to integrate conventionalized implicatures in syntax (e.g. appositives) and Bianchi et al. based their projection on his work.

FAIP regulates the conventionalization of focus implicatures and, in the case of Italian focus fronting, it holds interpretive and syntactic instructions via a [corr(ective)] feature. For the case of mirative focus, as for another example, FAIP has a [mir] focus feature, which distributes a mirative meaning and will trigger dedicated operations.⁸ With corrective implicatures, FAIP will activate a lower FocusP, which is a landing site for the focus movement (Rizzi 1997).

In sum, in Standard Italian, when an active proposition (or alternative, see section 5) is not compatible with the speaker’s knowledge, corrective focus implicature has its own linguistic form, encoded in FAIP. In informational focus, such contrast is not present, inasmuch as the alternatives/propositions are not salient in the discourse: FAIP will not involve the same syntactic mechanism and movement is banned (at least for Italian). Once we set out the main tenet of Bianchi et al.’s framework, we can now return to mica and explain the parallelism with focus fronting. First, preposed mica seems to encode the same corrective implicature as focus fronting: when a certain proposition p is introduced in the context, it indicates that this one is not compatible with the proposition in the corrective reply. With this in mind, it is easy to assume that if Italian FAIP conventionalizes this corrective implicature with fronting, then FAIP can involve mica as well, as it carries the same import to information structure. Hence, preposed mica can be seen as an instance of focus fronting.

Focus fronting requires the rejected proposition to be explicitly active in the discourse, and so does preposed mica, as shown by our acceptability experiment and the intuition by Cinque. If such kind of contrast is available, then FAI is able to activate the lower FocusP as potential site for the A’-movement of mica. This is also in line with Poletto’s proposal about the position of preposed mica in CP.

4.1.1 The movement of corrective mica

The syntactic mechanism that has been outlined here might pose some problems to the theory of adverbs proposed by Cinque (1999). In Cinque’s work, mica is one of those adverbs which occupy a rigid position, preverbal mica is to be attributed to the verb movement instead. When discussing verb movement, he uses mica to show that it is the verb that moves, whereas ‘mica does not move at all’ Cinque (1999: 50–51). This is clearly at odds with our claim that mica, triggered by FAIP, can move. Cinque’s main point is that mica cannot precede subjects (29), the verb moves to the head of the AdvP where mica stands in the specifier, establishing this precedence relation; the subject is still on top of the entire AdvP, proving that mica cannot move. Cinque also adds in a footnote (p. 185) that mica can precede the subject only when it works as a narrow focus on the subject, possibly as constituent negation but not as denial of an entire predicate.

1. (29)

1. *Mica
2.   Mica
1. Gianni
2. Gianni
1. gli
2. him
1. telefonerà.
2. call.fut.3sg
1. (Cinque 1999: 50)
2.  
1.    ‘Gianni won’t call him at all.’

1. (30)

1. Gianni mica gli telefonerà.
2. ‘Gianni won’t call him at all.’

Differently from Cinque, Moscati (2022) illustrates an example where mica precedes the subject and still works as denial of the entire clause. We collected new empirical evidence on the possibility of the order mica preceding the subject. First, we searched naturally occurring sentences on the internet with this order. We report some of these examples. Then, in order to ascertain whether these sentences are acceptable or only marginal, we ran a second acceptability experiment on Psytoolkit. The employed methodology was the same as the previous experiment, i.e. with a 7-points Likert scale, and ceiling sentences and floor sentences in order to establish references on the employed scale. Besides these sentences, naturally occurring in spontaneous written data, the target sentences of the second acceptability experiment also contained those ones that will be discussed later in the paper (such as syntactic tests and other Cinque’s diagnostic tests on mica). For this second acceptability survey, there was a total of 16 target sentences and 12 fillers. Among the 150 speakers who took part in the data collection, 65 surveys are usable following the same requirements of the first survey. The final dataset comprised 15 speakers from the Gallo-Italic dialectal area, 12 from the southern one, 12 from extreme southern one, 9 from Veneto, 2 from Rheto-Romance area, 2 from Sardinia (this second dataset is also available in the appendix).

1. (31)

1. ⁹Mica
2.    Mica
1. gli
2. the
1. scienziati
2. scientists
1. potevano
2. could.3pl
1. dire
2. say.inf
1. “la
2. the
1. probabilità
2. probability
1. è
2. is
1. 0”.’
2. 0
1.    ‘The scientists could not say “the probability is 0”.

1. (32)

1. ¹⁰Mica
2.    mica
1. Babbo
2. Santa
1. Natale
2. Claus
1. poteva
2. could.3sg
1. avere
2. have.inf
1. la
2. the
1. carta
2. paper
1. uguale
2. equal
1. alla
2. to.the
1. nostra!
2. ours
1.    ‘Santa Claus could not have the same paper as ours.’

1. (33)

1. (context: the speaker is commenting an advertisement of a perfume. In the advertisement, the main focus is on the packaging rather than the perfume itself)
1. ¹¹Mica
2.    mica
1. una
2. a
1. donna
2. woman
1. si
2. refl
1. spruzza
2. sprays
1. il
2. the
1. packaging.
2. packaging
1.    ‘Women do not spray the packaging.’

As anticipated, we checked the acceptability of these naturally occurring sentences. We illustrate the results in Figure 2, which shows that the sentences are acceptable.

Figure 2

Acceptability of sentences with preposed mica and overt subject.

Note that the acceptability of these sentences was tested with contexts eliciting the denial of entire predicate rather than the constituent denial of the immediately following subject, as exemplified in (34).

One theme that has not yet figured prominently in the discussion is the impact of regional variety onto these latter judgements. The dialectal area from the experiment did not show any difference in the acceptability of preposed mica preceding a subject. An ANOVA comparison between two ordinal mixed models was perfomed: one model did not contain the dialectal area and the one had the dialectal area as fixed effect. Again, the model with the lowest AIC (AIC = 424.84, Log Likelihood = –206,42) was the one without the dialectal area specified as fixed effect (model with dialectal area: AIC = 429,84, Log Likelihood = –203.64).

1. (34)
1. a.

1. Persino
2. even
1. gli
2. the
1. scienziati
2. scientists
1. non
2. neg
1. sono
2. are
1. sicuri
2. sure
1. di
2. of
1. questo
2. this
1. nuovo
2. new
1. vaccino…
2. vaccine
1. Hai
2. have.2sg.pres
1. sentito
2. heard
1. cosa
2. what
1. hanno
2. have.3pl
1. detto?
2. said?
1. ‘Even the scientists are not sure about this new vaccine… Have you heard what they said?’

1.  
1. b.

1. Sì,
2. yes
1. ma
2. but
1. mica
2. mica
1. gli
2. the
1. scienziati
2. scientists
1. potevano
2. could.3pl
1. dire
2. say.inf
1. “la
2. the
1. probabilità
2. probability
1. è
2. is
1. 0”.
2. 0
1. ‘Yes, but the scientists could not say “the probability is 0”.’

Based on the natural data and their acceptability, we have positive evidence that mica can be preposed before the subject. Furthermore, mica need not to have a narrow scope on the subject, but can deny the entire predicate. This evidence makes the movement of mica to CP still tenable.¹² With the data just showed, we do not want to claim that verb movement does not occur. Schifano (2018), for example, uses the impossibility of the order *mica > verb in the Teolese variety (near Padua) to show verb movement behaviour, we agree with Schifano on this point. However, we want to remark that the status of negator is also important when considering such variation. Let us assume that two closely-related Northern Italian varieties display the same type of high verb movement, like Venice and Gazzolo (near Verona). Venetian, differently from Gazzolese, can optionally have preposed miga.

1. (35)

1. Miga
2. mica
1. go
2. have.1sg
1. visto
2. seen
1. łe
2. the
1. gondołe.
2. gondolas
1. (Venetian)
2.  
1. ‘I haven’t seen the gondolas at all.’

1. (36)

1. *Mia go visto le gondole. (Gazzolese)
2.   ‘I haven’t seen the gondolas.’

If we have the same pattern of verb movement, then the ungrammaticality of (36) must be ascribed to the nature of the negative marker. Indeed, this prediction is borne out: Gazzolese, differently from Venetian, has a more grammaticalized mica, which can be used as standard negation instead of the presuppositional one (Magistro et al. 2022).

To sum up, we believe that the verb does move higher or lower in the clause, but mica can also be fronted to CP when it is able to convey a specific meaning (see also Bobaljik’s 1999 criticism on the position of mica and verbs).

The idea that adverbs can move to CP was already proposed by Garzonio & Poletto (2014): they put forth the hypothesis that negative adverbs and some of the aspectual adverbs have a quantificational feature that is matched at the focus position in CP. If we assume a quantificational nature of mica (Tubau 2020; Mayer 2021), then the movement is motivated along the lines of Garzonio and Poletto’s proposal. The movement of mica to a higher position was also proposed in a footnote in Coniglio (2009), but, as landing site the the proposed slot is the one usually occupied by the preverbal negation non. Given the data we have just showed, and the points raised by Poletto (2020), we argue that the landing site of preposed mica is in the FocusP projection, activated by FAIP. We will also illustrate other syntactic tests in section 4.3

We will now turn to the case where mica is clause-internal and does not move.

4.2 Contrastive mica

Before we turn to the syntactic mechanisms involved in the internal mica, we need to say a few words about its acceptability. Mica does involve a contrast with a salient proposition in any case, regardless of its position. However, as we have seen from the first acceptability experiment, when the proposition is not explicitly activated, internal mica is more acceptable compared to the preposed one, as already suggested by Cinque.

This result poses two challenges: first, the higher acceptability scores of internal mica compared to preposed mica call for an explanation. Secondly, we might want to find a reason why internal mica in out-of-the-blue did not score extremely high rates, as it happens when it occurs with a proposition explicitly mentioned in the previous context. The problem is crucial in order to go further in the discussion, since sentences judged by the speakers as not completely acceptable might seem to not offer solid grounds for building syntactic hypotheses.

For the case of internal mica, the proposition can be retrieved by accommodating it by resorting to world knowledge, or presupposing that the proposition belonged to speaker’s expectations, or other repair strategies. We use the term ‘accommodation’ in the classical sense of Lewis (1979), i.e. as a repair strategy for a missing presupposition. The hearer will accommodate missing information in order to meet the communication goals (see Beaver & Zeevat 2007 for a formal review). In our case, the hearer can force p to spring into existence using adequate strategies, and rescue the denial meaning of internal mica. However, as Beaver and Zeevat remarked for accommodation, “speakers do not accommodate everything destructively” (sic), but only with specific discourse/pragmatic requirements. Among these, they listed Gazdar’s principle (after Gazdar 1979): “when in conflict, implicatures always prevent global accommodation (i.e. projection) of presuppositions”. This principle can be applied to corrective focus. The corrective implicature of focus fronting encodes rejection specifically against an aforementioned proposition, thus the speaker cannot accommodate the missing proposition with repair strategies: an antecedent p must be explicitly available by definition. The corrective implicature and the implicature engendering the missing propositions are then in contrast, and accommodation is prevented, leading to a pragmatic infelicity. The same effect does not hold with the simple contrastive implicature: in this case, the predicate p can indeed be engendered.

Another functional principle proposed by Beaver and Zeevat, which is relevant for the case at stake, is the Discourse Record Implicature. According to this, presuppositions about what is in the discourse record may not be accommodated. Accordingly, the accommodation may or may not happen, depending on the kind of the context reconstructed by the speaker. Such differences in the accommodation of the discourse record might explain why the sentences with internal mica and unactivated p scored more medial values instead of being judged as completely grammatical.

When the proposition p is not explicitly mentioned, but it is accommodated (as, for example, in forms of speakers’ expectation), we clearly cannot have a corrective implicature. Syntactically, this translates to FAIP not carrying a [corr] feature and hence not activating the dedicated focus projection for movement. As a consequence, mica stays in situ. How to describe the feature bundle of FAIP in this case, then? We extend Bianchi et al.’s (2015) and Cruschina’s (2021) ideas for this case of focus-associated implicature. For FAIP, they define only a mirative implicature and a contrastive one, respectively activated on different features within FAI°. We argue that non-corrective mica is also characterized by its own feature which triggers distinct syntactic mechanisms. In other words, a focus associated implicature different from the corrective one is conventionalized. We propose that for non-corrective mica we have the conventionalization of a contrastive implicature.¹³ We adopt this dichotomy from Bianchi & Bocci (2012), where they make such distinction in order to account for focus fronting in Italian. They argue that only corrective focus can undergo fronting, whereas the contrastive one must stay in situ. Their example of contrastive focus is reported here in (37).

In sentence (37), focus-marking applies to Armani, deleting the other alternatives (in their example the other alternative is overt: a cheap dress from H&M, but its presence is optional and can be easily accommodated). In this case, contrast is conveyed, but the alternative does not come from a previous turn or information. Similarly, we argue that this is the case for non-corrective mica, the alternative is evoked within the same communicative turn and mica expresses contrast with p instead of correction. We will return to the semantic details of this in section 5.

Concerning its exact positioning, two hypotheses can be made about contrastive mica. First, we can simply assume that it stays in situ in the core part of the sentence, as adjunct to the verb. If this is the case, it could occupy a position in its specific projection in the negative circuit, MinimizerP (in Poletto’s system). Following the description of TP as two fields, High Adverb Space (HAS) and Low Adverb Space (LAS) (Ledgeway & Lombardi 2005; Schifano 2018), built upon Cinque (1999), mica is base-generated at the left boundary of LAS before the T (anterior) phrase occupied by già ‘already’. Similarly, Poletto (2020) reviews the possible merging positions of negative elements and puts the presuppositional negation in a T-anterior position.

For a discussion on the merging position of internal mica, based on its precedence relationship with adverbs, see also the extensive work by Manzini & Savoia (2005) and Poletto (2017; 2020). Alternatively, if we want to justify the non-corrective focus meaning associated with mica, we can resort to a low focus projection in the low periphery (à la Belletti 2004). According to Belletti, the vP Left Periphery possesses a specialized projection for new information focus, an interpretation which is not available in the periphery of CP (see Poletto & Bocci 2016; Badan & Crocco 2021 for a review on low focus). For the current discussion, we can dispense of a low focus movement for contrastive mica, since its interpretive effects are already present in the numeration of FAIP and do not need movement. Furthermore, the low focus position in the vP Left Periphery would be too low, considering that internal mica precedes già in TP. Hence, we adopt Poletto’s final schema, which comprises a dedicated projection for clause-internal mica. Back to our discussion, we claim that mica encodes by default the conventionalization of contrastive focus, in particular it would create a contrast with p by proposing the alternative ¬p. Such conventionalization is syntacticized by a FAIP head, which would give interpretive instructions on the type of focus via a dedicated feature. However, if mica is used to correct an antecedent proposition, then we add the conventionalization of a corrective focus in FAIP, a [corr] feature is attributed to enrich the structure, which would activate a focus projection for fronting.

To sum up, according to the activation of alternatives and thus the type of focus (encoded in different values in the FAI head), we can explain the different positioning and pragmatic requirement of mica. The lower position is accessible both by corrective (against active propositions) or contrastive (presenting a proposition), whereas the higher slot in C is available for the corrective reading only. We will address the issue of the two possible merging sites of corrective mica in section 6.

4.3 Syntactic tests

Adopting a dedicated FAIP operator, responsible for the licensing of different types of mica might seem superfluous: when mica is preposed, it would be enough to posit an activated FocusP in the high left periphery, triggered by the active propositions. More generally, in this paper, the architecture proposed by Bianchi et al. has been extended based on the similarity of corrective implicature, but we have not provided any specific test. In this section, we want to point some reasons why we cannot dispense the structure from a left-peripheral operator FAI.

As Cinque noticed, mica cannot appear in some kinds of subordinates, for example attributive relative clauses (38) and conditional protases (39) among others. We report Cinque’s examples here (Cinque 1999: 102).

The marginality of these sentences was also replicated in the second acceptability test, as shown in Figure 3: the majority of the speakers rejected the sentences even though the judgements of (38) are less clear-cut.

Figure 3

Acceptability of mica with a bridge verb (ex. 40), if-clause (ex. 38), relative clause (ex.39) (Cinque’s examples), with internal and preposed mica in wh-questions (ex.41 and 42).

Cinque’s point to account for these cases depends quintessentially on meaning: these types of sentence are typical of irrealis mood which is incompatible with the type of commitment lexicalized in mica. These motivations are straightforward and convincing, but the syntactic model sketched here allows us to explain these restrictions.

If we take, for example, restrictive relative clauses and conditionals, the literature proposed that they cannot host focus fronting because of intervention effects with other operators licensing the sentence type (Haegeman 2010). However, this is not sufficient to explain the limitations on mica: if such restrictions are given because of intervention effects at the high Left Periphery (potentially blocking the movement to FocusP), then the IP should be intact, allowing mica to appear at least postverbally. Notably, this is not possible, as shown in (38) and (39). Instead, the reason for the syntactic crash might be attributed to higher mechanisms, in a CP projection which licenses mica regardless of the position of the latter, FAIP. By adopting Haegeman’s ideas, these clause types are realized by the extraction of an operator to the left periphery; at this point there is reduced access for other operators because of locality effects. The licensing of some clause types such as restricted relative clauses and conditional protases will then inhibit a FAIP projection because of intervention effects. In turn, no type of mica is licensed. As control, when we look at other embedded contexts with bridge verbs, which allow the subordinate to possess a fully-fledged CP with all respective projections (Benincà & Poletto 2004), mica can be merged, as shown by example (40) reported directly from Cinque (1976) and judged more acceptable in our dataset (the plot in Figure 3).

Regardless of the type of focus involved, mica is necessarily activated by the FAI projection: in the absence of the latter, the derivation crashes. We take this piece of evidence to show that a left peripheral operator is necessary for both types of positions. Moreover, in a similar way as what happens with focus constituents, sentences with mica do not allow wh-movement altogether, regardless of its position (Erteschik-Shir 1986; Bocci et al. 2021 for a discussion), as reported in Figure 3.

In section 6 we further explore the hypothesis that mica associates with focus, based on independent studies on the diachronic mechanisms of Jespersen’s cycle. We will provide additional arguments supporting the adoption of a complex model with a FAIP and focus projections for the syntax and interpretation of mica. We will now try to define a semantic model for contrastive and corrective mica to justify their focal properties.

6.1 On the optionality of corrective mica

As we said, while contrastive mica can occur only internally, corrective mica can appear also preposed. Similarly, an XP consistuent under contrastive focus can appear only internally, whereas a corrective-focused XP may occur either internally or fronted. The optionality of movement for corrective focus is problematic within a generative approach, since there is no clear structural/semantic difference within the category of corrective foci.

By maintaining the parallelism between preposed mica and focus fronting, it might be useful to recall the solution put forth by Bianchi et al. (2015) and Bianchi (2019) on the optional movement of corrective focus. Bianchi (2019) sets out two equally important principles which produce constant tension between the “in situ” and “ex situ” constructions. The fronted form follows a scope principle, which requires the operator to be spelled out in its scope position; the clause-internal construction satisfies a syntax-prosody interface rule, the rightmost principle, which requires the nuclear pitch accent to be assigned to the rightmost position. A closely-related explanation was successively put forth by Bianchi & Bocci (2012), where the double positioning is based on the tension between the pragmatic markedness of focus and prosodic requirement, respectively rooting for fronted focus and in situ focus. They based their idea on Bobalijk’s (2002) proposal that in a movement chain, either the lower copy or the higher one can alternatively be spelled out, according to the type of movement (overt or covert). Bianchi & Bocci’s proposal disposed of covert movement, but rather suggests that the lower copy can be spelled out to meet the interface needs with prosody. This approach has a crucial impact on the syntax-prosody interface, inasmuch as it would imply a “lookahead model”, where prosody does not map directly what came after syntax in a modular way, but can already intervene in the earliest stages of derivation.

More specifically, the default focus marking on the rightmost element would lead to overgeneration in the case of mica, where it may appear at the very end of the sentence. However, clause-final positioning is not grammatical in Italian, as found in the data from the second acceptability experiment, visually reported in Figure 4.

Figure 4

Acceptability of sentence (52) with mica appearing after the object.

Titov (2020) explained the alternative positioning of corrective focus in terms of competition between two principles, as well. According to Titov’s formulation, the lower form follows the economy principle which favours Merge over Move (Chomsky 1995), whereas the structure with movement serves the function of mapping a more marked contrastive meaning (the corrective reply). In particular, she argues that Russian low focus constituents can be ambiguous in their informational import, i.e. they can be either informational or contrastive with different scope size; the movement to CP is then justified as a disambiguation strategy to remark the contrastive meaning. In conclusion, she argues that such movement is licensed by the quantificational property of the constituent over a set of discourse alternatives, a specialized meaning disambiguated across all the other possible interpretations which were available prior to the movement. The idea that preposed mica emerged as disambiguating strategy might find some support in historical data. Visconti (2009) noticed that preposed mica only starts being attested in the XIX century. Independently, she reports that during the same century, mica moved from a set of monological functions (contrast, textual linker, repetition) to the specialized contrastive function in dialogues. In Table 1, we report these results from the inspection of historical corpora.

It is not unreasonable to suspect that the preposed mica emerged as disambiguating strategy for corrective denial during the same period in which this particle started to specialize in dialogical contexts.

In short, we believe that mica in situ would bear the contrastive implicature (and if it is the case, the corrective one) and the fronted one would specialize exclusively in the corrective meaning, where optional movement would be a ‘disambiguating tool’, à la Titov.

6.2 Some diachronic speculations

The idea that mica is associated with focus is not entirely new, in particular in diachronic studies. For instance, Garzonio (2019), based both on historical and contemporary data, sketches an explanation in which the Old Italian né mica is situated in an Existential Phrase (ExistP) below a FocP. In his analysis, mica would initially occupy ExistP because it is a quantifier operating on events, FocP, on the other hand, would be the place where the negative coordinator né was merged. Mica was then reanalysed upward in the functional spine, occupying the specifier of FocP while né was lost in Standard Italian. Hence, Garzonio’s analysis of mica already pointed at sensitivity with focus.

Interestingly, there is some research that may lead us to think that mica had already focus sensitivity at the beginning of Jespersen’s cycle (Jespersen 1917; Breitbarth et al. 2020 for a recent review). To begin with, mica started its life as minimizer, denoting the lowest point on a pragmatic scale (cf. Horn 1989), in particular, in Latin, it used to mean ‘crumb’. Now, similar minimizers have been claimed to be associated with the focus particle even (Heim 1984; Horn 1989; Giannakidou 2007; Kuno 2008; Chierchia 2013). Under negation, minimizers would exhaustify all the other alternatives lexically triggered in their pragmatic scale. To take a clearer example, if one has not eaten a crumb of bread, then no bigger alternatives (a slice, a loaf and so on) are eaten. This exhaustifying effect would be conventionalized under a focus feature, in line with the operational definition we provided for focal operations. For reference, we provide an example in (54), showing a tentatively internal structure for the minimizer.

Batllori (2016); Batllori & Hernanz (2013) started to bridge the passage between a negative minimizer and an “Emphatic Polarity Particle”, arguing that the latter are licensed by a complex machinery with a [iNeg] feature and a Focus Operator that encodes the meaning [same]/[reverse], or a Force Operator that encodes [objection].

Gianollo (2020), analyzing the diachrony of Italian nessuno emphasizes the role of focus features. This NPI was originally emphatic (It. nessuno < Lat. Nec ipsum unum, lit. ‘not even itself one’), internalizing a DP focus projection, connected to a higher focus operator. More interestingly, Gianollo (2021) applies similar considerations to the Latin negator non, in comparison with the older negator ne, which was replaced in an earlier instance of Jespersen’s cycle (Schwegler 1988). It is argued that non is merged in a high CP-focus position, where it is endowed with [iNeg] and [uFoc] features. Of course, a rigorous model which could account for the maintenance of a focus feature from the minimizer to the polarity focus is beyond the scope of this paper, as it would imply core discussions about feature acquisition and change. What we want to stress here is that the association of mica with focus, as a result of FAIP, fits in with independent or similar previous studies.

To conclude this section, we want to show some other cases where the diachronic passage between an emphatic polarity particle to simple negation is attributed to the loss of a focus feature. It has been claimed that in some cases of Jespersen’s cycle the emergent negator is deprived of [focus] and turns it into the standard negator. For example, Simpson & Wu (2002) argue that French pas started life as a focus element, triggered by semantic reasons; eventually it turned into an element inside a focus concord shell and, at the end of the cycle, it became a negative concord shell, being a functional element with no specific pragmatic import.

Similarly, Breitbarth (2017) and Breitbarth et al. (2020) argue that Jespersen’s cycle is fed by two economy principles, Minimize Structure (introduced by Cardinaletti & Starke 1994) and Feature Economy (Van Gelderen 2009), which impoverish the internal structure of the emerging negators under certain conditions. Not only do we attest a change in the interpretability of the formal [Neg] feature, but the negation markers do not necessarily convey a focal interpretation anymore.

Preliminary experimental evidence in Magistro et al. (2022) show that Gazzolese mia (originating from Latin mīca as well) does not necessarily interpret a focus feature at PF, inasmuch as less prominent pitch accents are found in this variety, in comparison to less grammaticalized varieties, such as Venetian and Paduan. All these works suggest that we might be on the right track in attributing focus sensitivity to Italian mica, and hence by resorting to a mechanism with a FAI projection and a lower FocusP. Future research would then benefit from the application of this theoretical model to diachronic data.