An Evaluative Head in Romance The Palermitan Verbal Affix -vu

In this work, we analyse the behaviour of a verbal suffix in Palermitan – the Romance language spoken in Palermo, Sicily – expressing an evaluative meaning with sentential scope. The suffix is (V)vu, where V is the thematic vowel of the verb. The Palermitan dialect we study here is the one spoken in the urban area, approximately by 250,000 speakers. This suffix is interesting for several reasons. In the first place, an evaluative morpheme with clausal scope is quite unusual in Romance and to our knowledge it has not been reported outside Sicily. In the second place, as we will illustrate later, its use is spreading among the youngest generation, showing that the dialect is productive and the form under scrutiny is adopted by larger segments of the population, even if it does not exist in Italian. Summary


Introduction
It is well known that a sentence conveys a literal meaning together with information concerning the relationship between the speaker and the context of utterance. This is the case for instance of temporal anchoring, i.e. the location of the events mentioned in the sentence with respect to the temporal location of the speaker. 1 Moreover, besides temporal anchoring, which is an obligatory component, as argued by many scholars, a sentence can convey the speaker's evaluation of the sentential content, its epistemic assessment and several other properties concerning the relation between the sentence and the contextual situation. 2 In recent years, generative grammar has devoted a lot of attention to the interface between sentence grammar and the context. In particular, scholars working in the cartographic framework have proposed, beginning with the '90s, a syntactic implementation of discourse functions -such as topic and focus (see Rizzi 1997) -and of several linguistic items expressing the speaker's attitude toward the propositional content. This is for instance the case of evaluative, evidential and epistemic adverbials (see Cinque 1999). 3 Our work follows the cartographic approach. We analyse the behaviour of a verbal suffix in Palermitan -the Romance language spoken in Palermo, Sicily -expressing an evaluative meaning with sentential scope. The suffix is -vu, preceded by a thematic vowel. The Palermitan dialect we study here is the one spoken in the urban area, approximately by 150,000 speakers.
This suffix is interesting for several reasons. In the first place, an evaluative morpheme with clausal scope is quite unusual in Romance and to our knowledge it has not been reported outside Sicily. In the second place, as we will illustrate later, its use is spreading among the youngest generation, being adopted by larger segments of the population, showing that the dialect, in spite of the influence of Standard Italian, is very much employed.

Evaluative Affixes
An evaluative affix conveys the speaker's evaluation of a certain content. In Italian we find nominal affixes, expressing an affective or derogative evaluation of objects, people or situations, such as for instance -etta in casetta (house-EVAL small), and verbal affixes, expressing an affective or derogative value concerning events, such as for instance -ucchiare in mangiucchiare (eat-EVAL in small bites). 4 The morpheme we analyse here belongs to a different class, in that it qualifies the whole sentence, and not only a part of it, as conveying a positwo authors. As far as legal requirements are concerned, Alessandra Giorgi takes official responsibility for § § 1, 3, 5 and 6. Fabrizio Sorrisi for § § 2 and 4.
2 For a theoretical proposal concerning the syntactic representation of the speaker's spatial and temporal coordinates, see Giorgi 2010. 3 It is impossible to provide an exhaustive discussion of the whole issue and of the related bibliography in this work. We will discuss only the works relevant to our specific topic.
4 For an analysis of these morphemes in Italian, see Gambino 2010 andGrandi, Montermini 2010. Annali di Ca' Foscari. Serie occidentale, 52, 2018, 65-86 e-ISSN 2499-1562ISSN 2499-2232 tive or a negative value. In this respect, therefore, -vu has the same functional role as adverbial forms, which, in many languages, Italian included, express a positive or negative evaluation, such as (s)fortunatamente ([un] luckily), inaspettatamente (unexpectedly), or a locution such as (per) fortuna che ([for] luck that). In all these cases the evaluation by the speaker concerns the whole clause. 5 These adverbs, according to the discussion in Cinque (1999), appear in the left periphery of the clause, above IP, in the same layer where evidentials and epistemics -e.g. allegedly and probably -may appear: 6 (1) Cinque (1999), left periphery:...[ evaluative [ evidential [ epistemic... Evaluative verbal affixes with scope on the entire proposition -sometimes included in the class of items expressing mirativity -are found in many languages, even if, as said above, not in Italian. Examples, according to Cinque (1999), can be found in Menomini, an Amerindian language spoken in Wisconsin, in Korean, in Ngiyambaa, an Austronesian language and in Akha, a Tibeto-Burman one. In Palermitan, beside the morpheme -vu, adverbial constructions expressing the meaning of fortunatamente (luckily), such as furtuna ca (lit.: luck that) and pi furtuna (lit.: for luck) can be found, whereas the adverbs ending in -mente are generally missing. 7 The basic contrast we study is illustrated by the following examples: 8 (2) Accattavu u pisci spada I bought-EVAL swordfish (3) Accattai u pisci spada I bought swordfish 5 In certain cases these adverbs can refer to constituents and not to the whole clause, as in the following case: (i) Gianni ha mangiato probabilmente la torta Gianni ate probably the cake In (i) the epistemic adverb probably refers to the cake. We are not going to consider these cases in this work. For a discussion see Giorgi 2016. It is also important to keep in mind that the notion of evaluation is often taken in a much broader sense than the one considered here.
6 For a semantic analysis of evaluative adverbs, see De Vries 2012. 7 We will briefly discuss below the complementary distribution of the evaluative morpheme -vu with adverbial locutions such as fortuna ca (lit.: luck that). Cf. also Cruschina 2010. 8 In order to assess the extension and the productivity of the morpheme -vu, we interviewed three generations of Palermitan native speakers and asked them to provide a grammaticality judgment about some relevant sentences. We interviewed a group of people aged 18-30, another group aged 40-55 and a third one aged 65-80. The total amounts to 12 speakers. By means of the sentence in (2), the speaker expresses an evaluation on the event, which just took place, whereas in (3), where the verbal form buy+PAST (accattai) appears, the speaker only asserts the existence of an event taking place in the past, without providing any evaluation. Note that in example (2), the one with the evaluative morpheme, the event is located in the past as well.
In order to better understand the value expressed by means of sentence (2), let us consider the following minimal pair: (4) Aeri ci cuntai/cuntavi tutti cuose Yesterday I told-PAST/told-IMPF him everything (5) Aeri ci cuntavu tutti cuose Yesterday I told-EVAL him everything The sentence in (4) only asserts the existence of a terminated event located in the past, whereas in (5), besides mentioning the event, the speaker expresses an evaluation on it, such as for instance the following one: "I was able to tell him everything and that was a good thing!". The 'flavour' of the evaluation is determined by the context. As far as example (5) is concerned, we are suggesting a scenario where the evaluation is positive, but we can easily provide examples where the evaluation is a negative one, as in the following case: The example in (6) is most naturally evaluated negatively, as a default, on the basis of the consideration that falling down the steps has negative consequences. However, given a suitable context, even this sentence could be associated with a positive value. Consider for instance the following case: (7) Carivu r'i scali e quindi u cicchinu mi mancò I fell-EVAL down the steps and hence the sniper missed me In this case, the fact that the sniper missed me is obviously associated with a positive evaluation of my fall. We will discuss more extensively on these issues about the meaning in § 4 below.

On the Morphological Properties of the Morpheme -vu
In this section we will analyse the morpho-syntactic properties of this morpheme, in particular the realisation of the features expressing tense and aspect and of those expressing agreement with the subject. 9 3.1 The Person Feature (8) Accattavi > accattai (Rohlfs 1968) buy-PAST We hypothesise, therefore, that the form -avi temporally precedes the form -ai. Both of them express the first person singular of the simple past. In the following example, we give the complete paradigm of this form: 10 (9) Accattavi/accattai (I bought) Accattasti (you bought) Accattò (he/she/it bought) Accattammu (we bought) Accattastivu (you bought) Accattaru (they bought) In Palermitan we find also another form, looking very similar to these ones, i.e. the imperfect of the indicative. In (10) we give the imperfect paradigm (-a is the thematic vowel): The examples from the first conjugation are however the vast majority. The ending -avi can therefore be analysed as constituted by a morpheme -v-marking tense and aspect and a morpheme -i marking person. In the form -ai, tense and aspect is realised by a 0-morpheme. Analogously, the imperfect ending -ava can be analysed as -v-followed -a. In the evaluative ending -avu we find the morpheme -v-, which we are going to analyse in the next section, combined with a person ending -u. Such a person ending is quite isolated in the Palermitan past paradigms, in that only the evaluative form adopts it to mark the first person singular.
We propose that -u is the reduced form of the first person pronoun eu, realised as an enclitic on the verb. This proposal is justified by the fact that, as discussed in the literature, this is a widespread process in Sicilian. 12 Moreover, precisely this peculiar ending permits to distinguish the evaluative forms from the past ones, making it possible to specialize this morpheme to express this specific semantic value.

Tense and Aspect
In this section we discuss the temporal and aspectual values associated with -vu. In the previous section we sketched a hypothesis concerning the nature of -u; here we investigate what kind of past is realised by means of the morpheme -v-. Is this the morpheme -v-appearing in the simple past -as in accattavi (I bought) -or the one of the imperfect -as in accattava (I bought-IMPF)? From the translations we gave above, it emerges that -vu is usually interpreted as a past form, but we can provide more formal and precise arguments to illustrate this point.
Note, as pointed out above, that this form was in a first moment -i.e. in the grandparents' generation -available only for the verbs having the regular first conjugation perfect, i.e. -(a)vi, derived from the Latin perfect, as proposed by Rohlfs (1968). Consider the following cases: This is an argument for claiming that -v-in -vu is actually the morpheme of the perfect and not the one of the imperfect.
The youngest generation however generalises the use of -vu also to verbs that are not present in the language of the previous generations In what follows we show that -vu is a perfective past and does not exhibit any of the properties characterizing the imperfect. 14 As a first consideration, observe that the -vu form is in general not anaphoric. By anaphoric we mean a verbal form that must refer to a previously given moment in time. The imperfect is anaphoric in that it requires a temporal antecedent and cannot be used otherwise, i.e. out-of-the-blue. The event expressed by means of the imperfect must have a precise temporal location in the linguistic or extra-linguistic context. Consider for instance the following example: If the temporal locution ieri alle 5 (yesterday at 5) is not realised, the sentence is out. Generalizing: if a precise temporal location for the event is neither present in the linguistic context, nor retrievable in the discourse, the sentence is ungrammatical. If the speaker uses a past form, the presence of a temporal locution is optional: 15 (15) (Ieri alle 5) Gianni ha mangiato un panino Lit.: (Yesterday at 5) Gianni has eaten a sandwich '(Yesterday at 5) Gianni ate a sandwich' 13 Modal verbs constitute an interesting case. The form putivu (I can-EVAL) is not accepted by all speakers. Interestingly, it is usually accepted by very young speakers, i.e. teenagers. It seems therefore a further innovation. The -vu forms of the other modals tend to be rejected. For instance, ruvievu (I must-EVAL), or vulievu (I want-EVAL) are usually considered ungrammatical, even if judgments are not crystal clear. The reason for the decreased acceptability could be due to the simultaneous presence of two different modalities: the evaluative one and the one carried by the verb, which might render the interpretation difficult or odd. The issue deserves further study.
14 Many scholars have studied the properties and distribution of the imperfect in different languages. To mention just some of the works concerning Italian, see Bertinetto 1991;Delfitto, Bertinetto 1995;Giorgi, Pianesi 2001, 2004 15 As we pointed out above, in order to express a non-perfective past in the Italian examples we use the present perfect form. Such a form is the most widespread one in Central and Northern Italian varieties and the one usually adopted in so-called Standard Italian. In this case the locution yesterday at 5 is not obligatory and the sentence can be used out of the blue. Let us consider now Palermitan. When the temporal topic is not realised, the imperfect is not available, as in Italian, whereas both the past and the evaluative form are possible: The -vu form patterns with the past one, and not with the imperfect. The second property concerns the continuous interpretation. Simplifying somewhat, when an event is interpreted as continuous, it is conceptualized as an unbounded sequence of sub-events of the same kind. Typically, the imperfect expresses a continuous aspectual value, whereas past forms, such as the simple past or present perfect, do not. For this reason, an event associated with the imperfect morphology can provide the background for another event, whereas, on the contrary, an event with past morphology cannot. In this case the two events can only be located in a sequence with respect to each other. Consider the following examples: (20) Mentre compravo/*ho comprato il pane mia madre mi aspettava fuori dal negozio While I buy-IMPF/buy-PAST/bread my mother was waiting for me outside the shop (21) Mentre accattava/*accattai/*accattavu u pani me matri m'aspittava fuora ru negozio While I buy-IMPF buy-PAST/buy-EVAL bread my mother was waiting for me outside the shop 'While I was buying bread, my mother was waiting for me outside the shop' The presence of mentre (while) forces the interpretation of the adverbial clause as the background, hence the imperfect is available, whereas the past is not. In example (18) the contrast between the imperfect and the simple past is in fact quite sharp. In Palermitan, consistently with what we saw above, the evaluative morpheme patterns with the past and not with imperfect. Third, the imperfect can be used to express habituality, whereas the past cannot. Consider the following examples: 16 (22) Quando ero piccolo giocavo/*ho giocato sempre a pallone nel cortile When I was a child I always play-IMPF/play-PAST soccer in the courtyard (23) Quannu era nicu iucava/*iucai/*iucavu siempre a palluni nu cortile When I was a child I always play-IMPF/play-PAST/play-EVAL soccer in the courtyard 'When I was a child I always used to play soccer in the courtyard' Example (20) illustrates the distribution of the imperfect and the past in Italian in habitual contexts. In Palermitan we find the same one and again the evaluative form patterns with the past and not with the imperfect. Finally, in Italian the imperfect appears in fictional contexts, whereas the past forms are marginal, or marked: 17 (24) Ho sognato che mangiavo/*ho mangiato un gelato I dreamt that I eat-IMPF/ eat-PAST an ice cream (25) Sugnavu ca mi manciava/*manciai/*manciavu un gelato I dreamt that I was eating an ice cream 'I dreamt I ate an ice cream' Again, Italian and Palermitan pattern in the same way, and the -vu form behaves as a past.
In general therefore, with respect to the temporal interpretation, we can conclude that the difference between -avu, and -ai/-avi, is that the -vu form expresses an evaluative affective (i.e. speaker-related) value, whereas the simple past only expresses the temporal (aspectual) value. There is no difference between the two with respect to the temporal interpretation. 18 We can conclude therefore that the -v-appearing in -vu is the one appearing in the simple past, even if -avi in the competence of the intermediate and youngest generation has been progressively replaced by -ai.
With respect to the aspectual properties, the -vu form is perfective, in that it is compatible with achievement predicates (for instance to reach the top of the mountain), which usually are incompatible with imperfective forms. Consider the following examples: 16 The sentences with the past forms are not actually ungrammatical, but odd, in that the past forces an eventive reading. Since the non-habitual reading is here strongly disfavoured, the sentence is perceived as anomalous.
17 For an extensive discussion of these contexts and for the analysis of the possible readings with indicative forms other than the imperfect, see Giorgi, Pianesi 2001. 18 See however § 6 for a brief discussion of the imminential reading of -vu. Achievements predicates are punctual telic predicates, intrinsically perfective, and cannot be combined with imperfective morphology, with the exception of very marked contexts. 19 As is possible to see from the examples, in Italian -cf. example (27) -the present perfect is permitted, whereas the imperfect is excluded. The same applies to Palermitan, where we see once more that the -vu form patterns with the past and with the imperfect. Hence, aspectually, the evaluative morpheme is perfective.
Finally, note that in Italian a stative predicate can be made perfective, given a suitable context. Consider the following Italian example: (30) Ho amato Maria per tre anni I loved Maria for three years A perfective past can combine with the stative predicate love Maria, because of the presence of the temporal locution per tre anni (for three years) identifying a temporal span in which the state holds. In Palermitan we find the same distribution: (31) Amai/amavu a Maria pi tri anni I love-PAST/love-EVAL for three years 'I loved Maria for three years' The evaluative morpheme can be combined with statives under the same conditions affecting the past morpheme. Consider also the following example: 19 For an analysis of this issue in Italian, see Bertinetto 1991. It is in principle possible to combine the imperfect morphology with the achievements and in some cases it might be (almost) acceptable, but in general the results are quite odd. We refer the reader to the quoted reference for further discussion. (32) Na partita aeri fui/fuvu veru bravu In the game yesterday I be-PAST/be-EVAL really good 'In yesterday's game I was really good' Analogously to the temporal locution we saw above, the phrase nella partita di ieri (in yesterday's game) determines a perfective reading. Let us add a few word on the form fuvu. This is the evaluative form of verb be and seems to convey a peculiar meaning: besides expressing the evaluation by the speaker, it also implies that the speaker participated directly to the events mentioned in the clause or at least was present in the background. For instance, in a context in which a goal is scored in a soccer game, the speaker can utter the following sentence: (33) Fuvu contentu quanno succiriu I be-EVAL happy when it happened 'I was happy when it happened' By means of this sentence the speaker not only expresses a (positive) evaluation on the event, but also implies that he was an important participant in the situation described, for instance he himself scored the goal. The grammaticalization of these nuances in the Italian varieties with rich(er) morphology is an important topic that has not been much studied yet and that certainly deserves further attention.

On Interpretive Issues
The -vu form has an 'affective' value. As we anticipated in the introduction, the evaluation by the speaker of the event can be either positive or negative and is contextually defined. The -vi form, which differs minimally from the evaluative one, on the contrary, simply means that an action took place in the past. Consider the following example: (34) Incuntravu a Gianni I meet-EVAL Gianni 'I met Gianni' This sentence can have a positive evaluation in the context in which Gianni is a friend of the speaker, whom the speaker met after a long time. The same sentence can have a negative affective value if Gianni is for instance a person to whom the speaker owns money and therefore the speaker would rather not meet. The interaction with negation constitutes a very interesting point. Consider the negation of the following sentence: Example (33) means that the speaker did not meet Gianni and that, in the situation in which the meeting would have a positive affective value -when for instance Gianni is a friend of the speaker -the non-meeting has a negative one. In the opposite situation -owing money to Gianni -on the contrary, the evaluation would be positive. In other words: the non-negated sentence means "it is a good/bad thing that I met Gianni", whereas the negated one means "it is a good/bad thing that I did not met Gianni". Hence, the evaluative marker has scope on negation. Such a reverse scope interpretation is quite anomalous and calls for an explanation. 20 According to Cinque's (1999) proposal, however, evaluative, epistemic and evidential projections are higher then negation, being over IP. So, what needs to be explained is the fact that a certain semantic value, which is usually realised in a high evaluative projection can be expressed by means of a low verbal morpheme. In what follows we propose a solution to this question.
Note that adverbial evaluatives exhibit the same ordering both in Palermitan and in Italian: with the adverbial locution. In other words, even if it could be deemed reasonable to have the -vu morpheme and the adverbial locution together in the same sentence to emphasise the speaker's evaluation, yet it is impossible, as shown by the following example: (40) *Furtuna ca incuntravu a Gianni Luck that I meet-EVAL Gianni 'Luckily I met Gianni' This observation constitutes an argument in favour of the idea that in the morpheme -vu is in complementary distribution with the adverbial, and that they lexicalize the same features. 22 In particular we propose, following Cinque (1999), that fortunatamente in Italian occupies the specifier position of an EVAL head. Furthermore, furtuna ca (luck that) in Palermitan, and presumably the equivalent fortuna che in Italian, occupies the same projection. Ca, or che, occupies the head position and furtuna, or fortuna, the specifier. We propose that by means of covert movement the morpheme -vu ends up in the head position of the same projection. The only difference is that fortunatamente (luckily), has an intrinsically positive value, which does not need to be contextually identified, as opposed to -vu. 23 Finally, -vu is incompatible with exclamative forms, such as mizzica/ mizzichina/mii, as shown in the following example: 24 (41) Mizzichina quantu manciai/*manciavu! EXCL how much I eat-PAST/eat-EVAL 'Good heavens, how much I ate!' 22 As pointed out by a reviewer, In Italian we can have sentences such as Gianni deve necessariamente partire (Gianni must necessarily leave), where deve (must) and necessariamente (necessarily) coexist. The possibility of such a co-occurrence has still to be explained. It seems to us, however, that the two items -the modal verb and the adverb -do not express exactly the same semantic value, the first being more on the bouletic side, and the second on the deontic one. A possible explanation for this incompatibility can be that exclamatives and evaluatives share the same features -or, more reasonably, they share at least a subset of the same features, exclamatives having an evaluative component -hence they cannot be both lexicalized. This point however deserves further studies in Italian varieties and in other languages. Summarising what we proposed so far, we have shown that the -vu form is perfective, as opposed to imperfective. In this, it is analogous therefore to the past form -avi/-ai, with the only difference that it expresses an evaluation by the speaker. Diachronically, it develops from the Latin first conjugation verbs exhibiting a regular perfect in -avi. As a person ending, it exhibits -u, which we suggest is the reduced form of the first person pronoun eu. Thanks to the presence of the enclitic reduced pronoun -u, this form gets specialised as an evaluative one. We can therefore propose the following diachronic derivation (where a is the thematic vowel of the first conjugation): The morpheme -vu lexicalizes an evaluative meaning, which in Italian is expressed by adverbs such as (s)fortunatamente ([un]luckily), with the only difference that -vu does not specify a positive or negative value, which is contextually determined. Finally, we proposed that even if it is realised as a verbal morpheme, -vu is interpreted in a high evaluative projection.
In the next section, we are going to consider the distribution of this form in embedded contexts.

Embedding and First Person
The evaluative form -vu cannot be embedded, with a single exception we are going to consider in a while. Consider the following sentence: (43) Gianni ci cuntò a Marco ca aeri accattai/*accattavu i fichi r'India Gianni to him-told Marco that yesterday I buy-PAST/buy-EVAL the prickly pears 'Gianni told Mario that yesterday I bought prickly pears' In this case the main sentence has a third person subject and the embedded one a first person. The evaluative form is not possible. However, when the subject of the main clause is a first person as well, the embedded evaluative becomes possible: (44) Ci cuntai a Marco ca aeri accattai/accattavu i fichi r'India. To him-(I)told Marco that yesterday I buy-PAST/buy-EVAL the prickly pears 'I told Mario that yesterday I bought prickly pears' Note that the repetition of -vu both in the main and in the embedded clause gives rise to very marginal sentences: (45) ?*Ci cuntavu a Marco ca aeri accattavu i fichi r'India. I tell-EVAL Marco that yesterday I bought-EVAL the prickly pears 'I told Mario that yesterday I bought prickly pears' The presence of a normal past form in the embedded clause restores grammaticality: (46) Ci cuntavu a Marco ca aeri accattai i fichi r'India. I tell-EVAL Marco that yesterday I buy-PAST the prickly pears 'I told Mario that yesterday I bought prickly pears' This happens even in more complex structures with adjunct clauses: (47) Assira ci cuntavu na storia a Mario e s'addivirtiu, poi quannu c'a cuntai/?*cuntavu a Luca, s'annoiò Yesterday night I tell-EVAL Mario a story and he had fun, then when I told it to Luca, he was bored Concluding, the evaluative morpheme occurs in the main clause and can appear in the subordinate one only when the main subject is also a first person. Note that it is not possible to draw the conclusion that this is a root phenomenon, because it is possible to have an embedded -vu even when the superordinate first person verb is embedded, as in the following case: (48) Tutti sannu ca ci cuntai a Marco ca aeri accattai/accattavu u pisci spada.
Everybody knows that I told Marco that yesterday I buy-PAST/buy-EVAL the swordfish 'Everybody knows that I told Marco that yesterday I bought swordfish' Example (48) contrasts with the following one: (49) Tutti sannu ca Gianni ci cuntò a Marco ca aeri accattai/*accattavu u pisci spada. Everybody knows that Gianni told Marco that yesterday I buy-PAST/buy-EVAL the swordfish 'Everybody knows that Gianni told Marco that yesterday I bought swordfish' In example (47), as opposed to (46), the main clause subject is a third person one and the intermediate subject is a first person. This is enough to license the embedded evaluative morpheme. Consider furthermore that the relation between the clause with the evaluative morpheme and the clause with a first person subject must be local, as shown by the following example: (50) M'arricuordu ca Gianni ci cuntò a Marco ca aeri accatta/*accattavu u pisci spada I remember that Gianni told Marco that yesterday I buy-PAST/buy-EVAL the swordfish 'I remember that Gianni told Marco that yesterday I bought the swordfish' In this example, the main clause subject is a first person, whereas the intermediate one is a third person. This configuration does not license the evaluative marker in the embedded clause. We can conclude that the evaluative morpheme only appears in the domain of the speaker, namely when the immediately superordinate subject refers to the speaker. When -vu is in a main clause, it is still in the domain of the speaker, given the anchor for the main verb is constituted by the utterance event itself, defined by the temporal and spatial coordinates of the speaker.
We propose that this is due to the presence of the reduced first person pronoun -u, which must have a local antecedent. In other words, we argue that the reduced form of the first person pronoun is anaphoric, and must have an antecedent in the local domain. Such an antecedent can either be the superordinate subject, or the speaker, whose hic et nunc defines the utterance event. 25 This distribution is also found in the epistemic use of the Italian credo (I believe) -cf. Giorgi 2010 -which is licensed by a superordinate first person subject exactly like -vu. In the next section we are going to discuss the similarities between those particular usages of credo (I believe) and-vu, taking into account the distribution of the other kinds of high sentential adverbials as well.
25 This observation fits very well in the more complex picture concerning temporal anchoring, as discussed in Giorgi, Pianesi 1997 andGiorgi 2010. According to these authors, in fact, the spatio-temporal coordinates of the speaker must be locally represented in the C-layer both in main and subordinate clauses, in particular with dependences exhibiting an indicative mood. Moreover, in subordinate clauses the spatio-temporal coordinates of the subject of the superordinate clause, i.e. the attitude bearer, are represented in T -cf. Higginbotham 1995. In order to obtain a grammatical sentence with the evaluative morpheme -vu the relevant local projection -either T or the leftmost projection in the C-layer -must contain the speaker's coordinates. We are not going to discuss this issue any further here and refer the reader to the mentioned references. Let us consider now the distribution of the adverb francamente (frankly). This kind of adverbs can only take the speaker as an antecedent, hence they cannot refer to the subject of the superordinate (cf. Jackendoff 1972): (58) Francamente Mario è un cretino Frankly Mario is a stupid(speaker) (59) *Gianni crede che francamente Mario sia un cretino Gianni believes that frankly Mario is a stupid (*speaker) The explanation provided for credo (I believe) also holds in this case, the only difference being that in the case of francamente (frankly) the feature +speaker is not due to first person morphology, as is the case with credo and -vu, but to its intrinsic meaning. Even in this case, wen embedded under a first person its acceptability strongly increases: (60) (?)Credo che francamente Mario sia un cretino I believe that frankly Mario is a stupid (speaker) Concluding, we can say that credo (I believe), analogously to -vu, is morphologically marked as a first person, hence interpretively it refers to the speaker. Francamente (frankly) in Italian is marked as +speaker as well, but as a lexical property. Hence, they cannot be embedded, unless in the scope of +speaker.

Imminential Usages
Given the proposal developed above on the nature of -vu, we illustrate here a last piece of evidence, even if we only have a tentative explanation for the phenomena in question. We think that it is important to remark that single morphemes can be syncretic with respect to their value, combining some of their properties together. The morpheme -vu can in certain cases express an imminential meaning, combined with the evaluative interpretation. For instance, given a context in which the speaker is hungry and the meal is ready, the speaker says: Or, let us imagine a context in which it is late at night and people go on talking forever. The speaker is eager to go home and says: (62) M'innivu! I go-EVAL 'Now, I go!' In both cases, the speaker is evaluating the situation as negative for herself and expresses her will of doing something in the immediate future to overcome the negative effects. As a first consideration, we can say that the imminential value is made possible by the fact that the -vu form is a perfective form, and not precisely a past temporal one, as discussed above in § 3.2.
As argued in Giorgi and Pianesi (1997), perfectivity is usually temporally interpreted as a past, but such a past interpretation -simplifying somehow -is a side-effect of perfectivity. In other words, in principle perfective forms are not intrinsically past ones and for this reason they are also amenable to a future temporal interpretation, which is precisely what happens in this case. 26 This kind of temporal interpretation seems to be specialised in Palermitan for contexts about which the speaker has a negative evaluation and therefore is possible only with -vu. The form manciai (I eat-PAST) cannot appear in these contexts, because it necessarily expresses a past temporal value.
Furthermore, we argued above that the morpheme -vu expresses an attitude of the speaker towards the following content. In these cases, even if the interpretation is not purely evaluative, such a component is still there. Note also that these sentences are realised with an exclamative intonation, which is reminiscent of the Italian exhortative forms, as the following one: Descriptively, we can say that the Italian present tense is often used profuturo, and therefore in this case it can very well be used to express an imminential meaning, and the exclamative intonation provides the exhortative part. The Palermitan examples with -vu express the Italian meaning we see in (63), plus the evaluative component.