Contrastive focus and right dislocation
Contrastive focus and right dislocation
Abstract and Keywords
This chapter examines the interaction of focalization and right dislocation. It first shows that right dislocation can displace higher constituents to the right of lower foci, giving the false impression that these foci have moved upwards. It then examines cases where right dislocation targets constituents containing a focus, showing how such focus is obligatorily evacuated from these phrases. As a result, evacuated foci—including Rizzi’s left-peripheral foci—are preceded by hanging and CLLD topics located at the left-periphery of the clause and only followed by right-dislocated phrases. Furthermore, and in sharp contrast with Rizzi’s analysis, evacuated foci do not c-command the constituents following them, as expected if these are right-dislocated phrases moved to a TP-external position as claimed here. Finally, the chapter also examines the co‑occurrence of wh-phrases and left-peripheral foci, showing how the right-dislocated nature of the constituents following left-peripheral foci correctly predicts when co-occurrence is grammatical or ungrammatical.
As we have seen in the previous chapters, Italian discourse-given phrases can remain in situ (Chapter 2) or be right-dislocated to a position above TP (Chapter 4). We have also seen that contrastive focalization remains in situ whenever right dislocation is absent (Chapter 3).
This chapter pulls together these results to provide a comprehensive analysis of the entire distribution of contrastive focalization in Italian, showing that it is directly affected by right dislocation. The distribution will be shown to be partitioned into two distinct cases depending on what constituents are targeted by right dislocation, with different consequences for focalization.
1. RD is absent or it targets phrases that do not contain a focused constituent—In all these cases, focalization occurs in situ (see also Vallduvì 1992).1 We already saw in Chapter 3 that this is the case when right dislocation is absent. This chapter shows that focalization occurs in situ even when a focus is followed by right-dislocated constituents, provided the focused constituent was not generated within them. When these constituents are generated higher than the focused one, their occurrence to the right of focus gives the impression that focus has moved leftwards, but this is never the case.
2. RD targets a larger phrase that contains a focused constituent—In all these cases, the focused constituent evacuates the targeted phrase by moving leftwards just as high as necessary to enable the right dislocation of the targeted constituent. The final position of the evacuated focus thus depends on what phrase is targeted (p.164) by right dislocation. For example, a focused constituent evacuating a right-dislocating VP will occur in a lower position than a focused constituent evacuating a right-dislocating TP. The final position of the evacuated focus immediately precedes the right-dislocated phrase. In linear terms, the focus appears left-peripheral relative to the dislocated phrase. But it is not so structurally, since it will be shown to never c‑command the dislocated phrase to its right.
As we will see, several results follow from this analysis. First, the described interaction correctly accounts for the different properties displayed by unfocused constituents preceding and following focus, such as the observation that CLLD and hanging topics must precede left-peripheral foci whereas unfocused constituents following evacuated and non-evacuated foci are always marginalized or right-dislocated discourse-given constituents.
Second, the data described as ‘left peripheral focalization’ in the literature will be shown to be a strict subset of the linguistic expressions determined by focus evacuation. They are produced whenever right dislocation targets an entire TP and that TP contains a focused constituent. The term ‘left-peripheral’ only holds in linear terms: these foci do not c-command the right-dislocated TP to their right.
Third, the articulated interaction with right dislocation described in this chapter shows that focalization cannot be fully understood when studied in isolation. It is not possible to identify and explain the range of positions displayed by contrastively focused constituents without taking into account the effects of right dislocation, despite the independent nature of these two operations. More generally, it is not possible to study the distribution of focalization without taking into consideration the distribution of discourse-given phrases.
Fourth and last, positing a fixed projection for contrastive focalization anywhere in the clause is both incorrect and unnecessary. It is incorrect because it cannot account for all instances of focalization where focused constituents occur lower or higher than the location of the posited projection. It is also unnecessary, since as this chapter and this book show the entire distribution of contrastive focalization can be accounted for by focalization in situ and focus evacuation triggered by right dislocation, with the positive theoretical consequence of keeping the grammar maximally simple.
The chapter starts with Section 5.2 discussing the evidence for in-situ focalization for the first set of cases, where the constituents targeted by right dislocation and focalization do not overlap. Section 5.3 examines the second set of cases, introducing focus evacuation, the related predictions, and comparing them against the predictions of current analyses based on Rizzi’s (1997) and Belletti’s (2004) focus projections. Section 5.4 examines the interaction with wh-extraction, showing that contrastive foci and wh-operators can co-occur in the same clause, contra Rizzi (p.165) (1997, 2004), provided the syntax and intonation of right-dislocated constituents in sentences involving evacuated foci is taken into account.
5.2 The interaction between focalization in situ and right dislocation
When focus and right dislocation target independent constituents, i.e. non overlapping ones, any sentence where focus is followed by a discourse-given constituent falls into one of the following two classes: (i) sentences where the discourse-given phrases following focus are generated lower than the focused constituent, and (ii) sentences where at least one post-focus constituent is generated higher than the focused constituent. Two examples are shown in (1) and (2) respectively. In (1) a focused subject is followed by a lower-generated discourse-given object. In (2) a focused object is followed by a higher-generated discourse-given subject.
The post-focal constituent in the first class of sentences is ambiguous between a marginalized or right-dislocated status (hence the parentheses surrounding the comma, which represents the optional pause preceding right-dislocated phrases, which is absent with marginalized ones). Focalization occurs in situ in either case, since in this case the higher focus necessarily precedes the lower constituent whether the latter is marginalized in situ or right-dislocated.
The second class of sentences is more interesting because it may at first appear to provide evidence for leftward focus movement. I will claim that focalization occurs in situ in this case as well, which means that the post-focal constituent must have been right dislocated, as a higher constituent would necessarily precede the lower focus if marginalized in situ. For example, the subject of (2) must have been right dislocated, yielding the structure in (3) with the object focused in situ.
(p.166) This last observation provides us with a test for in-situ focalization. Since a marginalization analysis is excluded, any higher-generated discourse-given constituent following focus is predicted to show the typical properties of right dislocation while those of marginalization should be absent. The rest of this section shows that this prediction is borne out: postfocal constituents of this kind are indeed always right-dislocated and never marginalized.
5.2.1 The status of higher-generated phrases following postverbal focus
In-situ focalization predicts that both post-focus constituents are right-dislocated. As a first test of this prediction, we can replace the post-focus constituents in (4) and (5) with the corresponding negative phrases, as in (6)(a) and (7)(a). As explained in Chapter 3, right dislocation places negative phrases outside the licensing domain of the preceding neg‑marker non, whereas the same phrases are grammatical when marginalized in situ. Since (6)(a) and (7)(a) are ungrammatical, the right-dislocated status of the post-focal constituents is confirmed, which in turn supports the in-situ location of the preceding focus.
The validity of this test is strengthened by the grammatical sentences (6)(b) and (7)(b) where the negative phrases precede the lower focus and thus allow for marginalization in situ, which in turn produces grammatical sentences as expected. The grammaticality of (6)(b) and (7)(b) also confirms that right dislocation of the post-focus negative phrases is the only factor responsible for the ungrammaticality of the (a) sentences. The structures for (6)(a) and (6)(b) are provided in (8).
The above data also provide evidence against positing a fixed intermediate focus projection between TP and VP. If focus raised leftwards to such a projection, the post-focus negative phrases should remain able to marginalize in situ and be successfully licensed, incorrectly predicting (6)(a) and (7)(a) to be grammatical. (The evidence against positing a fixed projection above TP is discussed later in Section 5.3; see also Section 3.2 and the introduction chapter.)
5.2.2 Scope asymmetries induced by right-dislocated indefinites
Further evidence for in-situ focalization comes from the obligatory wide scope displayed by higher-generated post-focus indefinites in negative sentences.
When preceding focus, these indefinite objects may be interpreted within the scope of a preceding c-commanding neg-marker, thus displaying narrow scope. For example, when the entire clause is presentationally focused, the indefinite object in (9) can be interpreted in the scope of the preceding negation, yielding the (p.168) interpretation they did not give us a single biscuit, i.e. they gave us no biscuits. This reading is even more evident when the adverb nemmeno ‘not-even’ is also present, although its presence is not obligatory.
Right-dislocated indefinites, on the other hand, cannot be interpreted in the scope of a preceding neg-marker and must instead take scope over it. This is expected, since they are right dislocated TP-externally and therefore outside the c-commanding domain of the neg-marker. For example, the right-dislocated object in (10) is only grammatical under the odd interpretation where a specific biscuit has not been given to Mary. Note that the indefinite object is not clitic-doubled, but since it follows a clitic-doubled right-dislocated indirect object, its right-dislocated status is not in question.
Let us further examine the scope properties of indefinites following focus. First of all, consider indefinites generated lower than the focused constituent. Since they can be marginalized in situ they are expected to allow for narrow scope. As (11) shows this is indeed the case; the sentence allows for the interpretation where no worker will be fired.
Now consider cases where the post-focus indefinite is generated higher than the focalized constituent. Since focalization occurs in situ, the indefinite can only follow focus if right-dislocated, which in turn predicts an obligatory wide-scope interpretation due to the TP‑external position of right-dislocated phrases.
The prediction is borne out. Consider for example the sentences in (12) involving an indefinite subject and a focused object. When the subject precedes the object, as in (12)(a), narrow scope is possible as predicted by the corresponding structure in (13)(a) where the subject occurs in situ. When the subject follows the object, as in (12)(b), narrow scope is no longer available, as expected if the object is right-dislocated (p.169) outside TP as shown in (13)(b). Being outside TP, the indefinite subject takes scope over negation, forcing the reading where a single specific worker will not leave the factories. This, in turn, makes (12)(b) an infelicitous reply to the provided context sentence, whose most natural interpretation requires narrow scope. The bracketed comma shows that the observations just described hold even when the post-focal subject is not preceded by a short pause, as could be the case under marginalization, confirming that the interpretation associated with marginalization in situ is absent.
A similar argument can be built using idiomatic forms based on indefinites. For example, the indefinite un’anima ‘a soul’ is often used in negative statements to mean ‘anybody’, but this idiomatic meaning is lost if the indefinite lies outside the scope of negation. As (14)(a) shows, the idiomatic meaning remains available when an indefinite subject of this kind occurs in situ preceding a lower-generated locative adjunct containing a focalized DP. The idiomatic meaning, however, is no longer (p.170) available when the indefinite subject follows focus in (14)(b), showing once again that the indefinite has been obligatorily right-dislocated (rather than marginalized in situ) as predicted by the analysis. As the bracketed comma shows, these observations, too, hold even when the post-focal idiomatic expression is not preceded by a short pause.
Once again, note how these data cannot be accounted for by analyses positing an intermediate fixed focus projection inducing leftward focus movement. Under these analyses, the focus constituent would be able to move to the left of the higher-generated indefinite while keeping the indefinite in situ and hence in the scope of negation, thus incorrectly predicting the availability of narrow scope.
5.2.3 Scope asymmetries caused by right-dislocated adverbs
Similar scope asymmetries are also attested with lower adverbs generated between T and the aspectual projection AspectP hosting verbal past-participles. When a discourse-given adverb precedes a focalized past participle, as in (15), it can be interpreted in the scope of the preceding negation. When the same adverb follows the focused past participle, as in (16), it necessarily takes scope over negation. In (16), this forces the unnatural and only marginally acceptable interpretation that laughing is the one action that John never did; he did everything else.
As before, this alternation in scope is exactly what is expected under the proposed analysis, where the post-focus adverb in (16) is right-dislocated and hence outside the scope of negation as shown in (17). The same alternation, however, is unaccounted (p.171) for if the focused past-participle is assumed to raise to a focus projection above the adverb as in (18), since in this case the adverb could remain in situ and hence in the scope of the preceding neg‑marker.2
Focalization in situ is also corroborated by the distribution of lower adverbs like mica (neg-particle) and più (no more) which require licensing by a c-commanding neg-marker and therefore cannot be right-dislocated. As expected, they can precede a contrastively focused past-participle but not follow it, as right dislocation would place them outside their licensing domain. As before, the bracketed comma shows that these observations hold independently from the presence of a short pause right after focus.
These data, too, are unexpected if focus moved to a higher intermediate focus projection, since the past participle would then raise above mica and più, thus enabling them to remain in situ and licensed.
5.2.4 Order asymmetries caused by right-dislocated adverbs
The mandatory right-dislocated status of post-focus phrases generated higher than focus also explains the asymmetric distribution of higher adverbs relative to focus noticed in Cinque (1999). Cinque examined the higher pragmatic, evaluative, and modal adverbs that precede lower adverbs. He noticed that like lower adverbs they follow a rigid order when the entire clause is presentationally focused. For example, as the following data from Cinque (1999: 12) show, the pragmatic adverb francamente ‘frankly’ precedes the evaluative adverb purtroppo ‘unfortunately’, the evaluative adverb per fortuna ‘luckily’ precedes the modal adverb probabilmente ‘probably’, and probabilmente precedes the adverb forse ‘perhaps’.
This rigid order disappears in post-focus position. Consider for example the following sentences from Cinque (1999: 16) showing free ordering between purtroppo and forse, and between francamente and the lower habitual adverb di solito ‘usually’ (the right-dislocated adverbs require a pause right before them).3 (p.173)
Cinque’s observation follows straightforwardly from the obligatory right-dislocated status of higher-generated post-focus phrases. The rigid order of higher adverbs under clause-wide focus reflects their base-generated order (independently of whether this order reflects a fixed syntactic hierarchy as per Cinque 1999 or follows from the adverbs’ semantics as in Ernst 2002). Since focus occurs in situ, the only way these adverbs may follow the focused object in the example sentences is through right dislocation. But right dislocation allows for free ordering, explaining why the original rigid order is lost.
In-situ focalization also accounts for a subtle asymmetry involving wh-extraction from pre- and post-focus sentential complements. Consider the context question in (p.174) (26) which ensures that the subject and sentential complement of the subordinate interrogative clause are discourse given. The two answers in (26)(a) and (26)(b) repeat the same interrogative clause with a new indirect object, thus contrastively focusing it. Crucially, wh-extraction is more readily available in (26)(a), where the sentential complement precedes the subject, than in the more marginal (26)(b) where their order is switched.
This slight difference in grammaticality is exactly what is predicted if focalization occurs in situ. Since the focused indirect object occurs in situ, in both sentences the subject i tuoi fratelli ‘your brothers’ must be right-dislocated (if it were in situ it would precede the indirect object). In (26)(a) the sentential complement precedes the subject and hence it can be marginalized in situ, and therefore it allows for wh‑extraction; see the corresponding structure in (27)(a). In (26)(b), instead, the sentential complement follows the right-dislocated subject and therefore it too must have been right-dislocated. Since no clitic doubling is present, wh-extraction remains possible but it shows the typical marginal grammaticality characterizing wh-extraction from right-dislocated clauses described in Section 4.4.4. The corresponding structure is in (27)(b).
(p.175) The same asymmetry is predicted absent under current cartographic analyses positing a fixed focus projection, since in this case the focused indirect object would raise to a focus projection above the unfocused subject, thus enabling both the subject and the sentential object to occur marginalized in situ in sentence (26)(b), incorrectly predicting wh‑extraction to be non-marginal in this sentence too.
The previous sections showed that when right dislocation applies to constituents not containing focus, focalization occurs in situ. Focused constituents can be followed by marginalized and right-dislocated discourse-given phrases. Post-focal constituents generated to the right of focus can be either marginalized in situ or right-dislocated; their status is ambiguous and analysts must take this into account before drawing any conclusion (unfortunately this crucial test is currently almost always absent in the literature on Italian focalization). Post-focal constituents generated before focus are instead necessarily right-dislocated, since marginalization in situ is inconsistent with their post-focal position.
These results are analytically valuable, because they enable analysts to identify the right-dislocated status of specific constituents even when clitic doubling is absent. They become even more effective when combined with the following two additional criteria that apply to all phrases. First, since marginalized phrases follow the base-generated order because they occur in situ, any set of post-focal constituent altering such order is necessarily right-dislocated. Second, any constituent following a right-dislocated constituent is also right-dislocated.
As a practical application, consider the six grammatical sentences in (28) showing a focused object followed by a discourse-given subject, indirect object, and temporal adverb in all possible permutations. The above criteria immediately tell us that in sentences (a) and (b) all constituents following focus are right-dislocated because the first of them, the subject Carlo, must be right-dislocated because it is generated above the focus. Similarly, all post-focal constituents are right-dislocated in (c) and (d) as well, since the initial adverb ieri is base-generated before the focused object and therefore its post-focal position must be caused by right dislocation. Only sentences (p.176) (e) and (f) allow for the potential marginalization of the indirect object a Maria, since it can be generated lower than the focused object and be marginalized in situ in its original position. The same indirect object might also have been right-dislocated, its status being ambiguous, while the adverb and subject following it can only have been right-dislocated since they are base-generated before the focused object.
Most literature on Italian focalization does not discuss the potential right-dislocated status of the constituents following focus, implicitly assuming that they are never right-dislocated. As example (28) illustrates, this assumption is almost always incorrect, calling into question any conclusions based on it. Right dislocation is much more pervasive than our intuition may at first suggest. It is marginalization that is more uncommon. Even when present, it usually co-exists with an equally possible right-dislocation analysis, as is the case for the indirect object in (28)(e)–(f). Only an appropriate control of right dislocation can ensure accurate and valid deductions about the internal structure of the Italian clause.
5.3 Focus evacuation: the role of right dislocation in left-peripheral foci
As we saw in the previous section, when focus and right dislocation apply to distinct constituents, focalization occurs in situ. But right dislocation may also target constituents that contain focused phrases. As this section will show, in this case—and only in this case—the focused phrase is raised to a position outside the dislocating constituent, in order for right dislocation to apply to a focus-free constituent.
I call this operation ‘focus evacuation’ to distinguish it from the ‘left peripheral focus’ analyses à la Belletti (2001, 2004) and Rizzi (1997, 2004) where focalized constituents move to the specifier of a fixed focus projection. As we will see, focus evacuation accounts for all instances of Italian left-peripheral focus (though more research is needed for the partial focus fronting data of Fanselow and Lenertová 20114). Most (p.177) interestingly, the distribution of focus evacuation will be shown to extend beyond the familiar left-peripheral focus data à la Rizzi (1997); Rizzi’s data constitute a specific subcase that occurs when the constituent containing focus and targeted by right dislocation happens to be an entire TP (as opposed to, say, VP, PP, AP, DP).
Several properties distinguish focus evacuation from these analyses making it possible to test its validity.
1. Dependence on right dislocation—Focus evacuation only occurs when right dislocation is present and targets a constituent containing focus. When right dislocation is absent or targets constituents not including focus, the conditions for focus evacuation are absent and focalization occurs in situ as discussed in Section 5.2. Since right dislocation is an optional process, the analysis immediately accounts for Brunetti’s (2003) observation that leftward focus movement is optional. This contrasts sharply with any analysis proposing that all focalized constituents obligatorily raise to a posited focus projection.
2. Varying position—The final position of evacuated foci is predicted to vary in accord with the constituent targeted by right dislocation. For example, a focused object will move above TP when TP is targeted but only above VP when the target (p.178) is VP. This, too, contrasts sharply with any analysis where all focalized constituents move to a unique and fixed focused projection.
3. Status of post-focus constituents—The constituent containing focus and targeted by right dislocation eventually follows the evacuated focus and is expected to display the typical properties of right-dislocated phrases.5 The same does not hold in analyses where focus moves to a fixed focus projection, since in this case the constituent following focus need not be affected by any operation (or, to put it differently, can be marginalized in situ).
4. Lack of c-command—Since right‑dislocated phrases are not c-commanded by the constituents preceding them (see Chapter 4), evacuated foci are predicted to be unable to c‑command the right-dislocated constituent originally containing them. Analysis moving focus to a higher focus projection instead predicts that left-peripheral foci can c‑command any constituents to their right, since the latter are not obligatorily right-dislocated.
Of the above four predictions, the first has already been examined in the discussion of focalization in situ in Section 5.2. The rest of this section discusses the remaining three as well as several other pieces of evidence. Section 5.3.1 provides the structural details of focus evacuation. Section 5.3.2 examines the different positions taken by evacuated focus, showing in the process that Rizzi’s (1997) data identify a specific subclass within the wider set of data determined by focus evacuation. Section 5.3.3 considers NPI-licensing relations, showing that evacuated foci cannot c-command the right-dislocated constituents to their right. Section 5.3.4, refining Samek-Lodovici (2009), examines the left-peripheral focus data studied in Rizzi (1997) and presents several pieces of evidence supporting their evacuated status. Section 5.3.5 examines the alternative analyses proposed in Benincà (2001), Benincá and Poletto (2004), (p.179) Frascarelli and Hinterhölzl (2007), and Bianchi and Bocci (2012). Section 5.3.6 briefly considers some additional evidence from parasitic gaps constructions. Finally, Section 5.3.7 considers how the proposed remnant movement analysis fares relative to Müller’s principle of unambiguous domination (Müller 1996, 1998).
5.3.1 Focus evacuation
Schwarzschild (1999: 150) convincingly demonstrated that discourse given phrases may contain focused phrases. Consider (29). When B is uttered after A the items John, apple, and green are trivially discourse-given because they co-refer with identical items in A.
Schwarzschild argues that the entire sentence B is also given in the context of A because A entails the proposition that John ate an apple of some colour. Formally, he defines a constituent as discourse-given whenever its existential F‑closure is entailed by the discourse context, where the existential F-closure is obtained by replacing any focused constituent with a variable and existentially closing the result (modulo existential type shifting, see Schwarzschild 1999 for discussion).6 He then explains how under this definition the proposition John ate a RED apple in B counts as given because its existential F-closure ∃Y[John ate a Y apple], meaning that John ate an apple of some colour and obtained by replacing RED with Y, is entailed by sentence A. Using the same definition, Schwarzschild shows that the VP ate a RED apple and the DP a RED apple also count as given because the corresponding existential F-closures, roughly equivalent to the action of eating an apple of some colour and the existence of an apple of some colour, are also entailed by sentence A.
Similarly, the TP Bill saw Rome in (30)(B) counts as given when uttered as a reply to (30)(A) because its existential F-closure, equivalent to the proposition that Bill saw some city, is entailed by (30)(A) even if the object ROME contained in it is contrastively focused.
(p.180) When the same exchange occurs in Italian, as in (31), the discourse-given TP corresponding to (30)(B) becomes a potential target for right dislocation. Since right dislocation is an optional operation, there are two possible answers to (31)(A). When right dislocation does not apply, the focused object ROMA remains inside the TP as in (31)(B1), which is fully analogous to its English counterpart in (30)(B) and discussed no further.
When right dislocation does apply, focus moves leftward to evacuate the targeted constituent and enable its right dislocation, giving rise to answer (30)(B2).
Two issues arise. The first concerns the cause of focus evacuation: why does focus have to leave the constituent targeted by right dislocation? A possible explanation, discussed and formalized in Chapter 6, is that right-dislocated phrases being discourse-given are subject to prosodic constraints requiring discourse-given material to be non-prominent (cf. Féry and Samek-Lodovici 2006). When focus occurs inside a constituent targeted by right dislocation a conflict arises between the need to stress focus and the need to avoid stress on the right-dislocated constituent. Focus evacuation makes it possible to satisfy both requirements.7
The second issue concerns the position of evacuated foci. The analysis just described predicts that evacuated foci will raise only as far as is necessary to enable right dislocation of the targeted constituent, thus left-adjoining to it. This is the shortest movement leaving the targeted phrase focus-free and ready to right-dislocate.
It follows that the final position of evacuated foci will depend on which constituent is targeted by right dislocation. For example, when right dislocation targets a VP containing a focused indirect object, the indirect object will left‑adjoin to VP. An example is provided in (32), where the evacuated indirect object a LUCA ‘to Luke’, here pied-piping the preposition due to the absence of preposition-stranding in Italian, is followed by the right-dislocated VP to its right (to avoid excessive (p.181) clattering, I am using ‘VP’ also for the projection hosting the verbal past-participle). The corresponding derivation is in (33). First, focus evacuation forces the focused indirect object to VP-adjoin as in (33)(a). Then the lower VP segment is right dislocated as in (33)(b), in accord with the analysis of right dislocation proposed in Chapter 4. Note how the focused indirect object a LUCA precedes the dislocated VP but does not c‑command it.
When right dislocation targets an entire TP, focus left-adjoins to TP, giving rise to the more familiar left-peripheral focus sentences discussed in Rizzi (1997) and many subsequent analyses. For example, sentence (34) arises when the TP containing the focused object ROMA ‘Rome’ is targeted by right dislocation. The corresponding derivation is shown in (35). First, the focused object is evacuated via leftward TP-adjunction. Second, the lower TP-node is right dislocated. As a result, the focused object linearly precedes but does not c-command the right-dislocated TP.
The evidence for the above analysis is discussed in detail in the following sections. When discussing sentences involving right-dislocated TPs, I will where convenient use the term ‘left-peripheral focus’ originally used by Rizzi (1997). This term, however, is potentially misleading, as it suggests that sentences where focus is descriptively left-peripheral—i.e. linearly preceding an entire TP—form a special class, whereas they are just a subset of the structures determined by focus evacuation.
The label ‘left-peripheral focus’ also incorrectly suggests that focus raise to a higher position for intrinsic reasons (e.g. to check its features in a higher fixed focus projection), thus obscuring the key role played by right dislocation in triggering the existence of these structures. Finally, the term may incorrectly be interpreted as implying that the raised focus c‑commands the constituents to its right, which is not the case, since the evacuated focus never c-commands the right-dislocated constituent that originally contained it.8
Focus evacuation predicts that the same focused constituent will move to different positions depending on which containing phrase is targeted by right dislocation.
The following examples show several instances of focus evacuation involving right-dislocated phrases of increasing size, such as DP, AP, PP, VP, and TP. The initial (a) sentence in each set lacks right dislocation and focus occurs in situ. The remaining sentences involve right dislocation of increasingly larger constituents, giving the impression that focus moves further and further leftwards. Under an appropriate intonation, with focus duly stressed, they constitute natural replies to the provided context; the few marginal cases have been marked as ‘?’. Due to the absence of preposition stranding in Italian, the preposition preceding focus is always pied-piped with the evacuated focus.
Since the examples lack clitic doubling, the right-dislocated status of the constituent following the evacuated focus is not immediately evident. Clitic doubling is (p.186) inevitably absent when right dislocation targets an entire TP containing a focus, since the head T hosting the clitic is part of the dislocated phrase. But instances of focus evacuation displaying clitic doubling do occur whenever the right-dislocated constituent originally containing focus is smaller than TP. Examples involving focus evacuation out of clitic-doubled APs and DPs are shown in (41) and (42). Right-dislocated verbal past participles disallow for clitic doubling, but as (43) and (44) show they can follow a clitic-doubled constituent, thus leaving no doubt about their right-dislocated status. (Stress falls on focus and each dislocated phrase is preceded by an intonational break.)
The presence of clitic doubling in the previous examples provides a first robust piece of evidence for the focus evacuation analysis. It does unquestionably show that foci can be generated in a larger constituent targeted by right dislocation and that the same foci are extracted before right dislocation takes place. In other words, the visible presence of clitic doubling leaves no doubt that the focus evacuation process proposed in this chapter does occur in Italian. This is not sufficient to refute competing (p.187) analyses of left peripheral focalization based on dedicated focus projections, since right dislocation could occur even in those structures, but it establishes that focus evacuation does exist. Since this process is sufficient to account for the left peripheral focalization data, the onus to prove that movement to a higher focus projection is also present is on the corresponding analyses. As I will show in the following sections, several pieces of evidence show that this cannot be the case.
The focus evacuation analysis also provides a unified account for all the above sentences and any other similar ones, correctly predicting when focus remains in situ, when it moves, and how far up it moves. As discussed in the next section, the same does not hold for analyses à la Rizzi (1997, 2004) and Belletti (2001, 2004) which must appeal to additional movement operations that eventually generate ungrammatical sentences.
184.108.40.206 Overgeneration in current cartographic analyses of focalization
Following Belletti (2004), most analyses of focalization assume without much discussion that all contrastive foci move to Rizzi’s (1997) fixed focus projection above TP, whatever position they may have in linear terms. Consequently, in these analyses any disruption of the base-generated order must be accounted for by independent processes moving unfocused constituents to the topic projections located above and below FocusP. These movements give rise to problematic consequences.
Consider for example sentence (37)(b) and (37)(c) repeated in (45). In a structure à la Rizzi (1997), the word order for (45)(a) requires movement of the PP ‘[PP via t]’ to a topic projection below FocusP, followed by the remnant TP ‘[TP siamo andati t]’ moving to a topic projection above FocusP, see structure (46)(a). Likewise, sentence (45)(b) requires movement of the VP ‘[VP andati via t]’ to a topic projection below FocusP and movement of the remnant TP ‘[TP siamo t]’ to a topic projection above FocusP, see (46)(b).
(p.188) Movement to these topic projections must be freely available, or else the above movements would constitute ad hoc operations postulated to obtain the observed word orders. If freely available, however, these operations generate ungrammatical structures. For example, if we switch the landing site for the TP and PP in (46)(a), which should be possible as we are simply swapping the topic projections targeted by the TP and PP, we obtain the structure in (47)(a) which corresponds to the ungrammatical sentence (47)(b). Similarly, if we swap the final positions of TP and VP in (46)(b) we obtain the structure in (48)(a) which corresponds to the ungrammatical sentence (48)(b).
(p.189) The ungrammatical assessment of these two sentences is crucially sensitive to the intonation being provided, which must be the one expected under the corresponding structure. The constituents ‘[via]’ and ‘[andati via]’ in (47)(b) and (48)(b) constitute left-dislocated topics similar to the initial topic ‘Marco’ in the grammatical sentence (49). Therefore, they should carry a B-accent (Büring 1997: 60) here represented in small caps and when uttered at a normal speed they should be followed by an intonational break and short pause (Frascarelli 2000: 48). This intonation crucially distinguishes the ungrammatical sentences in (47) and (48) from their counterparts in (50) which are unsurprisingly grammatical because the PP and VP are here pied-piped with the evacuated focus and therefore show neither a B‑accent nor a pause before the focused item.
In conclusion, the ungrammaticality of (47) and (48) show that positing a unique focus projection above TP cannot account for the instances of focus evacuation in (45) without introducing additional operations that would then have to be further constrained through additional conditions. These conditions are unnecessary under focus evacuation, which furthermore provides a straightforward and unified analysis of the sentences in (45): they respectively involve evacuation of a PP, pied-piped by a focused DP, from a dislocating PP and VP. Similarly, since no additional operations need to be posited, the ungrammaticality of sentences (47) and (48) is immediately accounted for as due to the extraction of the focalized constituent from the initial PP and VP, which here constitute CLLD phrases and as such constitute an island to extraction. The focus extraction operation itself is unlicensed in these sentences, because focus evacuation can only be triggered by right dislocation in the way described earlier in this chapter.
Focus evacuation makes fine-grained predictions about the availability of licensing relative to focused and unfocused negative phrases. These predictions are all borne out, providing strong support for the analysis.
220.127.116.11 Focused negative phrases
The first prediction concerns the licensing of focused negative phrases. Italian negative phrases need to be licensed by a c-commanding licenser when lower than T but not when c‑commanding T (see appendix A). We thus expect evacuated negative foci to need licensing depending on their final position, which in turn depends on the size of the constituent targeted by right dislocation. Specifically, evacuated negative foci should require licensing by, for example, a preceding negation when evacuating constituents smaller than TP, but require no licensing when evacuating a right-dislocating TP. The same alternation is instead predicted to be absent under a cartographic analysis where all foci share the same fixed focus projection, since in this case licensing is expected to be uniformly present or uniformly absent depending on the position of the posited focus projection.
The available data support the focus evacuation analysis. For example, a focused negative indirect object requires licensing by the preceding neg-marker non when focalized in situ as in (51)(a) and also when evacuating a right-dislocating VP as in (51)(b). Omission of non makes these sentences ungrammatical. The same negative indirect object needs no licensing when evacuating an entire TP, as in (51)(c), since in this case it TP-adjoins and therefore it occurs higher than T.
The sentences in (52) show the same pattern with a different set of examples, showing how a focused negative PP needs to be licensed when focused in situ or evacuated from a right-dislocating AP, but not when evacuating a right-dislocating TP.
18.104.22.168 Unfocused negative phrases following evacuated foci
A second prediction concerns the licensing of negative phrases within the right-dislocated constituent that triggered focus evacuation. Since right dislocation lifts this constituent outside TP and since licensing cannot occur under reconstruction (see appendix A), any negative phrase in this constituent no longer occurs within the licensing domain of its original licenser in TP and will consequently be predicted to be ungrammatical. This holds independently from the final position of the evacuated focus.
The prediction is borne out. Consider the data in (53), involving the right dislocation of a VP and leaving the evacuated focus in VP-adjoined position. Sentence (53)(a) is ungrammatical because right dislocation places the entire VP, including the negative object in it, outside the c‑commanding domain of the neg-marker non, as shown by the corresponding structure in (54).
Note that licensing remains possible when the VP is not right dislocated, as in (53)(b) where the negative object occurs in situ and precedes the focused indirect object, which is also in situ. Grammaticality is also preserved when the VP is right dislocated but contains a non-negative object, as in (55). These last two sentences show that failure in licensing the negative object is the only possible reason for the ungrammaticality of (53)(a), thus further strengthening the evidence for the right-dislocated status of the VP in (53)(a) predicted by focus evacuation.
The same kind of licensing failure is also expected—and found—when right dislocation affects the entire TP. For example, under sentence-wide presentational focus, preverbal negative subjects in specTP license lower negative phrases such as the negative object in (56). When the negative subject is contrastively focused, however, the object is no longer licensed, see the ungrammatical (57). As shown in the corresponding structure (58), the licensing failure follows from the extra-clausal position taken by the right-dislocated TP, which blocks the necessary c-command relation between licenser and licensee.9
The ungrammaticality of (57) also shows that the subject evacuates the dislocating TP without first moving to specTP, where its trace would be sufficiently high to act as a licenser. This is unsurprising, since other instances of A'‑movement affecting subjects, such as subject wh-movement, have been shown to involve extraction directly from the base-generated position of the subject (Rizzi 1982: 118; Brandi and Cordin 1989). The ungrammaticality of (57) also shows that the subject cannot move to specTP and focus there while the rest of the clause is marginalized. This, too, is expected since—as claimed in this book—focalization of the subject should occur in situ whenever right dislocation is absent, i.e. in the subject’s base-generated specVP position.
22.214.171.124 The distribution of the neg-marker ‘non’
Focus evacuation also accounts for the peculiar distribution of the sentential neg-marker non ‘not’ relative to negative subjects. Compare (59) and (60). As (60) shows, when the entire TP is presentationally focused and stress falls rightmost, the neg-marker may not co-occur with a preverbal negative subject (Zanuttini 1991; Penka 2011).10 However, if the preverbal negative subject is focused and stressed, as in (61), the neg-marker can be present, even though it forces a double negation interpretation (Penka 2011).11
(p.194) Under focus evacuation, the entire pattern follows immediately. The observed differences follow from the presence of a right-dislocated TP in (61), where the focused subject has been evacuated, and its absence in (60), where focus evacuation does not occur.
In (60), the preverbal negative subject in specTP needs no licensing and already negates the entire clause, therefore adding the negative marker is ungrammatical (see for example the account in Haegeman 1994, where economy considerations disallow the overt neg-marker non when it shares the same Agr projection of nessuno).
In (61), the clause-initial focused subject can only be a product of focus evacuation unleashed by the right dislocation of the containing TP, because when right dislocation is absent subjects focus in situ, i.e. post-verbally in specVP. The evacuated subject and the neg-marker within the right-dislocated TP do not share the same clause at surface. The neg-marker thus becomes possible again. But since the negative focused subject NESSUNO does not c‑command the right-dislocated TP containing the neg-marker non, no neg-concord is possible, forcing the attested double negation interpretation.
126.96.36.199 Problems raised by NPI-licensing to analyses positing fixed focus projections
Current cartographic analyses assuming a dedicated focus projection FocusP above TP cannot account for the data examined in the previous sections. For example, focused negative phrases in FocusP, including negative subjects, would c‑command lower negative phrases, incorrectly predicting their successful licensing. For example, (57), repeated in (62), would incorrectly be predicted grammatical in a structure à la Rizzi (1997, 2004) such as (63) because the raised subject ‘NESSUNO’ inevitably c‑commands the lower negative object. (Placing FocusP lower than T would have to explain why preverbal negative foci can be licensed in such a low position, given that their unfocused counterparts cannot, see appendix A.)
Before interpreting the observed licensing failure as evidence against positing a fixed focus projection above TP, we should consider whether focalization might interfere with licensing (Adger, p.c.). Such an interference is demonstrably absent.
To begin with, if focalization blocked licensing, we would expect licensing to be equally disrupted when focus intervenes between licenser and licensee. Yet the exact opposite holds. As (64)–(66) show, negative markers and negative subjects do license marginalized negative phrases across an intervening focalized argument (see also the (p.195) examples in Chapter 2 where a marginalized negative phrase is licensed across a contrastively focused verb).
Similarly, if focus interfered with licensing, we would expect licensing to be disrupted when focalization applies to the licensee. Again, this is not the case. The negative marker in (67) successfully licenses a negative object focalized in situ.
As the following examples show, even in English, where preverbal focused subjects c‑command the rest of the clause, focused negative subjects remain able to license lower NPIs, confirming that focalization does not interfere with NPI‑licensing.
Playing devil’s advocate in order to test the evidence for focus evacuation even further, we may wonder whether the licensing of negative-phrases by negative subjects is limited to negative subjects in specTP but disallows for any higher licenser. (p.196) Under this assumption, preverbal focused negative subjects could simply be too high to license postverbal negative constituents, with no need to invoke focus evacuation and the consequent right dislocation of TP. But this hypothesis, too, runs against the available empirical evidence. For example, in (70) the negative indirect object in the sentential complement is licensed by the matrix verb dubito ‘I doubt’, which is clearly located higher than TP since it precedes the complementizer che ‘that’. Likewise, in (71) the negative object is licensed by the covert interrogative yes/no-operator in the specifier of CP (or the specifier of the relevant projection in the analyses decomposing CP into multiple projections).
In conclusion, the study of negative-phrase licensing strongly supports the focus evacuation analysis. It also provides robust counter-evidence for alternative analyses positing a high fixed focus projection.12
We may further test the validity of focus evacuation by examining the properties of fronted constituents occurring immediately after clause-initial foci, such as the object la mela following the focused A MARIA in (72). Henceforth, I will call these post-focal phrases ‘PF‑phrases’. The corresponding sentences are grammatical provided that the initial focus is stressed and all post focus constituents, PF-phrases included, are unstressed and interpreted as discourse-given.
Under focus evacuation, PF-phrases and the TP following them are both predicted to be right-dislocated. The right-dislocated status of the TP follows from the clause-initial position of the focus, which can only obtain if the focus has evacuated a right-dislocating TP. Without right dislocation, focus would occur in situ inside the TP.
As for the PF‑phrases, they cannot be analysed as marginalized in situ, since in this case they would not be fronted before their TP. They cannot have been fronted after focus evacuation either, since they would then precede focus. They can, however, be right-dislocated independently of the TP. In fact, there is nothing surprising about PF-phrases and their TP being right-dislocated independently from each other and being ordered as they are, since right dislocation is known to apply to multiple constituents and to dislocate them in any possible order. The corresponding (p.198) structure for (72) is provided in (73), showing the PF-object la mela and the following TP abbiamo dato each right-dislocated to the specifier of a distinct RP projection. (A complete derivation is available in the footnote to this sentence.13 See also (p.199) appendix B, which provides the reasons why the representation adopted here is preferable to one produced by a rightward movement analysis of right dislocation.)
The structure in (73) contrasts sharply with the corresponding analysis à la Rizzi (2001, 2004), where the initial focus would occur in a focus projection above TP, while the PF‑object would have to be analysed as a left-peripheral topic as in (74).
The two structures differ in what they tell us about PF-phrases. Under focus evacuation PF-phrases are right-dislocated and are therefore inevitably expected to show the exact same properties as other right-dislocated phrases. The analysis also fits Neeleman and van de Koot’s (2008, 2012) and Neeleman and Vermeulen’s (2012) crosslinguistic generalization that foci never move across topics, since focus moves above the PF-phrase as part of the remnant TP and, furthermore, right-dislocated phrases do not qualify as topics, as will be amply shown later in this section.
The Rizzi-like structure in (74) makes the opposite prediction. PF-phrases are analysed as post-focal topics and consequently their properties are predicted to diverge from those of right-dislocated phrases accordingly. Furthermore, as Neeleman and Vermeulen (2012: 4) point out, structures like (74) constitute an exception to the generalization that foci cannot move above topics, since focus here does move across a topic, namely the PF-phrase.
This section will show that PF-phrases are indeed right-dislocated phrases and not topics, as predicted by the focus evacuation analysis. Some preliminary evidence in this direction is already present in the literature on information structure. For example, Brunetti (2009) shows that PF‑phrases share the same pragmatic function (p.200) of right-dislocated phrases, namely they are discourse-given non-contrastive phrases, or ‘tails’ in Vallduvì’s (1992) terminology. Similar evidence is also available in Frascarelli and Hinterhölzl (2007), who show how naturally occurring PF-phrases display the typical properties of right-dislocated phrases, as discussed later in Section 5.3.5. PF-phrases also show the same intonational contour of right-dislocated phrases, being preceded by the same intonational break and optional pause. This is for example the case for the object in (72), where the intonational break is represented by the preceding comma. Furthermore, the PF-phrase and the TP following it can be freely ordered, much like any other set of right-dislocated items. For example, the object and the TP of (72) can be swapped as in (75). Under the focus evacuation analysis, these findings are all expected because PF-phrases are necessarily right-dislocated phrases.
The next few sections provide a more in-depth testing of the status of PF-phrases, showing that they systematically share the properties of right-dislocated items. To bring this convergence into relief, PF-phrases will be also shown to diverge from the properties of pre-focal left-peripheral topics, such as hanging topics (HT) and clitic left-dislocated topics (CLLD) (Cinque 1990; Benincà 2001; Benincà and Poletto 2004).14 As summarized in the table in (76), and as further discussed in the sections (p.201) to follow, right-dislocated phrases will be shown to differ from HTs with respect to clitic doubling, contrastivity, the distribution of bare NPs, the deletion of prepositions, the licensing of epithets, and the sensitivity to strong islands. The first three properties also distinguish right-dislocated phrases from CLLD topics. PF‑phrases align with right-dislocated constituents with respect to all these properties, as predicted by the focus evacuation analysis.
188.8.131.52 Evidence for the right-dislocated status of post-focus phrases
184.108.40.206.1 Preposition dropping
Unlike HTs, right-dislocated phrases cannot drop any preposition normally associated with the argument they express. For example, in (77), the right-dislocated indirect object a Maria ‘to Mary’ must retain its preposition. (p.202)
We can therefore test whether PF-phrases can drop prepositions. As illustrated by the examples in (78), this is never the case. Presence or absence of clitic doubling in the following TP is uninfluential. As expected, the same sentences become grammatical once the preposition is supplied, see (79).
When the order between focus and PF-phrases is swapped, the HT analysis is expected to become possible again because HTs are located above TP and therefore they always precede evacuated foci adjoined to TP. As the data below show this is indeed the case. The corresponding structure for (80) is provided in (81).
The test thus supports the right-dislocated status of PF-phrases and their difference from HTs. The preceding sentences also show that HTs can co-occur with evacuated foci, thus excluding potential interferences with focus as a possible cause for the ungrammaticality of the sentences in (78). Finally, they confirm that HTs must precede evacuated foci, as expected if they are located in projections higher than TP.
220.127.116.11.2 Epithet licensing
HTs allow for epithet licensing whereas right-dislocated phrases and CLLD topics do not (Benincà 2001). The asymmetry is illustrated in the following examples.
If PF-phrases are right-dislocated they should be unable to license epithets. As the data in the following examples from Samek-Lodovici (2009) show, the prediction is borne out. The two sentences in (85) show a subject and an object HT licensing epithets, while (86) shows that epithet resumption is no longer possible when the same subject and object occur as PF-phrases.
Once again we expect an HT analysis and the associated epithet licensing to become possible again when the same subjects and objects precede focus. As the data in the following example show, this is indeed the case. These sentences also confirm again that focus does not interfere with epithet-licensing, leaving the right-dislocated status of PF-phrases predicted by focus evacuation as the only cause for the ungrammatical sentences in (86).
18.104.22.168.3 Sensitivity to strong islands
Yet another source of evidence comes from the study of strong islands. As (88) and (89) show, HTs are insensitive to strong islands (Cinque 1990, see also Zeller 2006 for Zulu and Vermeulen 2007 for Japanese). Right-dislocated items are instead sensitive to them. This is shown in (90) and (91) where indirect objects cannot be right-dislocated out of a subject and complex NP island respectively. (CLLD-topics are also sensitive, see Cinque 1990: 59.)
As in previous cases, we may exclude any interference with focalization, since as shown in the following examples the same phrases behave as HTs and are insensitive to strong islands when they precede focus.
Right-dislocated constituents are never contrastive, whereas both HTs and CLLD topics can be so (Büring 1997, 2007; Benincà and Poletto 2004; Frascarelli and Hinterhölzl 2007: 101; Brunetti 2009; see also Lambrecht 1981, 1986 for French). For example, in (96) and (97) the initial constituents in the conjoined clauses—here underlined—respectively occur as contrasted HTs and CLLD topics providing a grammatical answer to the question ‘who will speak to whom?’. The same constituents occur as right-dislocated phrases in (98) where each conjunct clause is grammatical on its own, but the entire sentence is ungrammatical showing that right dislocation disallows contrastivity.
(p.206) When the same constituents occur as contrasted PF-phrases, as in (99), they too are ungrammatical, confirming the right-dislocated nature of these constituents predicted by focus evacuation. Once again, each conjunct is grammatical in isolation but the sentence as a whole is ungrammatical independently of whether clitic doubling is present or absent. Even the ellipsis of the second TP—an operation available with right-dislocated phrases—does not improve the overall assessment, see (100). As usual, main stress must fall on the focused subject: it is deceptively easy to inadvertently shift main stress to the indirect object and get the grammatical but irrelevant sentence where the indirect object acts as a focus following a contrastive topic.
As with all previous properties, the examined constituents can be contrastive when they precede focus, since in this case an HT/CLLD analysis becomes possible. For example, (101) and its ellipsis counterpart in (102) are grammatical as an answer to the question ‘I know someone will speak to John and Andrew, but who exactly will speak to them?’. As in all previous cases, this shows that interference with focalization cannot be the cause of the ungrammaticality of (99) and (100).
The absence of contrastive PF-phrases is also confirmed by Brunetti’s independent study of the interpretative properties of the constituents preceding and following left-peripheral focus (Brunetti 2009). Brunetti notes that only constituents preceding focus can be interpreted as left-peripheral contrastive topics, and that the constituents following focus, such as PF-phrases, always show the same interpretation of right-dislocated constituents (see also Lambrecht 1981, 1986 for French). For example, in sentence (103)(a), the phrase a Maria ‘to Mary’ can be interpreted as a contrastive CLLD topic, while the entire sentence is interpreted as a partial answer to the question in (103) which presupposes that different recipients received a phone call at different times. The same partial answer interpretation, however, is unavailable (p.207) when the same phrase follows focus as in (103)(b), making the sentence infelicitous under this context (as symbolized by ‘#’).
The focus evacuation analysis immediately accounts for Brunetti’s observations. Since the post-focus phrase a Maria in (103)(b) is right-dislocated, it cannot have the contrastive topic interpretation necessary for making (103)(b) felicitous in the provided context.
22.214.171.124.5 Absence of clitic doubling
Left-peripheral object topics require clitic doubling independently of their HT or CLLD status. The same is not true for right dislocation, where clitic doubling is optional (see Chapter 4 for discussion). The contrast is illustrated by examples (104) and (105). Note that the object in the second sentence is necessarily right-dislocated since it follows a clitic-doubled right-dislocated indirect object.
We may test the status of object PF-phrases by examining whether they, too, require a clitic. As the following sentence shows, a clitic is unnecessary, consistently with the right-dislocated status of PF-phrases predicted under focus evacuation.
As in the previous tests, the presence of focus does not interfere with the property being tested. When the object precedes focus an HT/CLLD analysis becomes available again forcing the presence of the object clitic. (p.208)
126.96.36.199.6 Availability of bare NPs
Bare NPs can occur as HTs and CLLD topics provided they carry the B-accent normally associated with contrastive topics (Büring 1997), see (108). They instead resist right dislocation, see the examples in (109) respectively with and without the partitive clitic ne ‘of them’.15
If the focus evacuation analysis is correct and PF-phrases are right-dislocated, they should disallow bare NPs too. The prediction is borne out.
As (111) shows, bare NPs remain possible when bare NPs occur as HTs and CLLD topics preceding clause-initial focus, confirming that right dislocation is the relevant factor causing the ungrammaticality of (110).
If PF-phrases are right-dislocated, as predicted by focus evacuation, they might also be expected to allow for clitic doubling like any other right-dislocated phrase. Yet they have been described by many linguists, myself included, as lacking it (Benincá 2001; Benincá and Poletto 2004; Belletti 2004; Samek-Lodovici 2009). Indeed, when considered outside an appropriate context, sentences with PF-phrases appear to be more natural without clitic doubling; compare for example the two sentences in (112), where the clitic is present only in the second highly marginal sentence.
There are native speakers, however, who find clitic-doubled PF-phrases acceptable (Cardinaletti, p.c., also Brunetti 2004 in Section 188.8.131.52). Clitic doubling becomes indeed possible when an appropriate context is supplied. This is for example the case in (113) and (114) where the presence of clitic doubling in the context sentence appears to be sufficient to make clitic-doubled PF-phrases fully acceptable. Native speakers considering (114) should make sure to place main stress on the auxiliary of the context sentence, rather than on the indirect object.
The availability of clitic doubling with PF-phrases becomes particularly evident when we consider NPs extracted from a quantified expression and right dislocated. These quantified NPs are exceptional in that they require obligatory clitic doubling. Compare (115), showing clitic doubling by ne ‘of them’, against the ungrammatical (116) (p.210) where the extracted fragole is not clitic doubled (once right dislocated, the NP requires a preposition, possibly for case reasons). If PF-phrases are not right-dislocated phrases and disallow clitic doubling, quantified NPs should never be able to occur as PF-phrases. If, instead, PF-phrases are right-dislocated phrases and as such allow for clitic doubling, then quantified NPs should be able to occur as PF-phrases while also forcing clitic doubling. As (117) attests, where the presence of the clitic ne is mandatory, this is indeed the case. The absence of a cliticless counterpart enhances the assessment, making these sentences fully natural. It follows that PF-phrases share the properties of right-dislocated phrases even with respect to clitic doubling.
184.108.40.206 Free word order after evacuated focus
As pointed out by an anonymous reviewer, if left-peripheral foci are evacuated foci followed by right-dislocated phrases, we expect the right-dislocated phrases to be freely ordered. As showed in Samek-Lodovici (2006: 861), this is indeed the case; see (118) where under an appropriate intonation every possible order of the three post-focal constituents is grammatical.
The same reviewer wonders whether there might be restrictions on the occurrences of non clitic-doubled objects in these contexts, given the sentence in (119) which the reviewer finds ungrammatical. (p.211)
Sentence (119) shows how easily complex right dislocation data become misleading when assessed outside an appropriate context. Once such context is provided and the exchange is made more natural by using slightly less generic verbs, these sentences become grammatical. See the grammatical (120)(a), structurally identical to (119) and like it involving a non clitic-doubled object.
As this section showed, PF-phrases display all the hallmarks of right-dislocated phrases: they are discourse-given, cannot omit prepositions, cannot express contrastivity, allow for clitic doubling as well as for its absence, are sensitive to strong islands, disallow epithets, cannot be bare NPs, and, finally, they are not in situ, so they cannot be marginalized phrases. All these properties are immediately accounted for under focus evacuation, where PF‑phrases must be right-dislocated phrases.
The observed convergence with right dislocation is instead unaccounted for under a fixed focus projection analysis à la Rizzi (1997). Under such an analysis, two issues arise. First, why do the properties of PF-phrases coincide with that of right-dislocated phrases? If both sets of phrases occur in the same position, then an explanation needs to be provided for why right-dislocated phrases follow TP while PF-phrases precede TP. Under focus evacuation this issue is absent, as the TP following the PF-phrases is right-dislocated too. Second, why, when left-peripheral focalization is absent, an initial non-subject expression, such as A Gianni in (121), can only be a topic (since it does show all the associated properties) but not a right dislocated phrase (since it never shows the properties associated with right dislocation).
Under focus evacuation, this is inevitable, since A Gianni in (121) does not follow an evacuated focus and therefore it cannot be a right-dislocated phrase. In an analysis involving Rizzi’s split-CP, where PF-phrases would be analysed as a specific type of topic that happens to be located between FocusP and TP this question remains (p.212) unanswered. If these topics can precede TP when focus is present, they are incorrectly predicted to do so even when focus is absent.
5.3.5 Existing analyses of post-focal phrases
PF-phrases have also been studied by linguists who assumed a cartographic analysis of left-peripheral foci. This section considers the studies most relevant for the evacuation analysis proposed here. I’ll start with Benincà (2001) and Benincà and Poletto (2004), and argue against their claim that PF-phrases are focused. I then consider Frascarelli and Hinterhölzl (2007) and show that the properties of PF-phrases that they identified are exactly those expected under the focus evacuation analysis. Finally, I consider Bianchi’s (2012) and Bianchi and Bocci’s (2012) claim that left-peripheral focus is available with corrective but not contrastive foci and show that the interesting data that they proposed follow immediately from the evacuation analysis once care is taken to consider the conditions under which right dislocation is pragmatically licensed.
220.127.116.11 PF-phrases are not focused—Benincà (2001) and Benincà and Poletto (2004)
Benincà (2001) and Benincá and Poletto (2004) claim that PF-phrases are foci. If true this would constitute evidence against the analysis proposed here where these phrases are necessarily right-dislocated.
A crucial tenet of their analysis concerns the claim that PF-phrases obligatory lack clitic doubling, a property that—minor exceptions aside16—is typical of foci (Benincá 2001). As we saw in Section 18.104.22.168, however, this claim is empirically incorrect. PF-phrases do allow for clitic doubling once appropriate contexts are supplied.
Analysing PF-phrases as foci also incorrectly predicts that constituents that allow for focalization but resist right dislocation should be able to occur as PF-phrases. (p.213) Negative phrases provide the right class of constituents for testing this prediction. They allow for focalization in both preverbal and postverbal position, see (122), but they cannot be right-dislocated, see (123) where the negative phrase follows a clitic-doubled right-dislocated indirect object (see also Section 4.3.3 and Appendix A). If PF-phrases are foci, negative phrases should be able to occur as PF-phrases. But they cannot, see (124). This is expected under the focus evacuation analysis where PF-phrases constitute right-dislocated phrases and therefore exclude negative phrases.
Furthermore, Frascarelli and Hinterhölzl (2007) showed that corpus occurrences of PF‑phrases are discourse-given. This is indeed a general property of PF-phrases. For example, sentence (125) is only felicitous in contexts that make the PF-phrase la torta ‘the cake’ discourse-given and unfocused. Sentence (125) is grammatically acceptable as a correction of sentence (126)(a), with its subject contrastively focused, or as an answer to (126)(b), with the subject now presentationally focused. But it is never possible under contexts that make the PF-phrase non discourse-given. For example, (125) cannot be an answer to (126)(c), where the subject would be contrastively focused. Yet precisely this last case is the one predicted possible by Benincá and Poletto’s analysis since in this case the PF-phrase la torta in (125) supplies the non-contrastive ‘relevant information’ focus described as possible in their analysis (Benincá and Poletto 2004: 2).
The obligatory discourse-givenness of PF-phrases excludes their use as new-information foci. Theoretically, they could still occur as contrastive foci, since this latter type of focalization can be discourse-anaphoric. But contrastive foci are by definition contrastive, whereas PF-phrases are not, as we saw in Section 22.214.171.124.4. The discourse-givenness character of PF-phrases is thus a serious obstacle for any analysis claiming that these phrases are focused. This problem is instead absent under focus evacuation, where discourse-givenness follows straightforwardly from the right-dislocated status of PF-phrases.
Benincá and Poletto (2004) also observe that PF-phrases are subject to weak crossover effects. For example, they find (127) ungrammatical. If correct, this property would distinguish PF‑phrases from right-dislocated constituents, since the latter do not display it, see (128).
When care is taken to include a context and assess the PF-phrases relative to it, they no longer appear to show any weak crossover effects. For example, sentence (129) shows no such effects when assessed in the provided context. This suggests that the marginal ungrammaticality of (127) is a reflex of right dislocation and the need to ensure that the necessary pragmatic conditions relevant to its licensing are properly satisfied, rather than a sign of weak crossover effects.
126.96.36.199 Word order and prosodic contour—Frascarelli and Hinterhölzl (2007)
Further evidence for the right-dislocated status of phrases following evacuated foci comes from Frascarelli and Hinterhölzl (2007), who examined a large corpus of naturally occurring Italian data and identified several asymmetries distinguishing the constituents immediately preceding and immediately following left-peripheral foci. Crucially, pre-focus topics display the typical properties of left-peripheral topics while post-focus ones, including the PF-phrases discussed in the previous sections, show the properties of right-dislocated constituents, thus supporting the focus evacuation analysis.
More precisely, Frascarelli and Hinterhölzl show that pre-focus phrases express aboutness and contrastive topics, must precede focus when carrying this interpretation, cannot be iterated, follow a fixed order, and are respectively associated with an L*‑H and an H* tonal contour. The post-focus phrases instead must be discourse-given, can be reiterated, can be freely ordered relative to the TP following focus, and are associated with an L* tone; all of which are properties typical of right-dislocated phrases.
Frascarelli and Hinterhölzl call post-focus phrases ‘familiar topics’. They assume a left‑peripheral focus analysis à la Rizzi (1997) which they further refine by positing distinct types of topic projections. Familiar topics are claimed to be situated immediately above TP as shown in (130) (with phrasal reiteration marked through the symbol ‘*’). As such, they linearly precede any material within TP. Right-dislocated phrases are assumed to be merged in the same position, but unlike pre-TP familiar topics they are maintained to involve remnant movement of the TP to their left, causing their right-peripheral position (see also Frascarelli 2000, 2004).
The analysis in (130) thus shares with focus evacuation the claim that PF‑phrases and right-dislocated phrases share the same position, but with several important differences. First, while FP-phrases are analysed as right-dislocated, the TP following them is not: it is marginalized in situ. All problematic predictions stemming from positing a fixed focus projection above TP thus apply to this template as well. For example, it remains unexplained why a focused preverbal negative subject cannot license a postverbal negative phrase or polarity item in the marginalized TP.
Second, it is unclear where the TP is moved to in sentences like (131) where the TP follows a left-peripheral focus but precedes a right-dislocated phrase, here the object. (p.216) The target position would have to occur between the higher FocusP and the FamiliarTopicP projection hosting the object, but the proposed template only shows projections for familiar topics in this region. It follows that TP could either be itself right-dislocated into a FamiliarTopicP projection, with the analysis becoming more and more similar to focus evacuation, or alternatively that the TP moves to a different type of topic projection, in which case the existence of such projection and its occurrence between FocusP and FamiliarTopicP* would have to be stipulated. The issue of why this new position is not available to PF-phrases would also arise.
Third, under this analysis the position of each topic class relative to the other topic classes and to focalization is stipulated, as nothing accounts for why the order is not different. The template does precisely that: it stipulates what occurs before what else, but it does not provide a reason for why such order must occur.
The focus evacuation analysis, on the other hand, provides some straightforward explanations for Frascarelli and Hinterhölzl’s observations. The problematic predictions associated with positing a high fixed focus projection are dispensed with in the way described in previous sections. PF-phrases and right-dislocated phrases are structurally identical, with no need to stipulate that right-dislocated phrases require remnant TP-movement while PF‑phrases do not. Remnant TP-movement occurs in both cases as part of the analysis of right dislocation (see Chapter 4) with the evacuated focus and—when present—any pre-focus topics being contained in the moved TP (or CP when pre-focus topics are present). There is no need to stipulate new projections between focus and PF-phrases. Finally, since the constituents following evacuated foci are always right-dislocated when foci are left-peripheral (i.e. when evacuated from a right-dislocating TP), all aboutness and contrastive topics must necessarily precede focus or else they too would share the properties of right-dislocated items. This explains the observed order between these topics, focus, and familiar topics (i.e. PF-phrases).
188.8.131.52 Contrastive and corrective foci—Bianchi (2012) and Bianchi and Bocci (2012)
Bianchi (2012) and Bianchi and Bocci (2012) note that distinct foci show a subtle pragmatic difference in their felicity conditions according to their evacuated or non-evacuated status. For example, the clause-initial focus in (132) is highly acceptable as a corrective reply to sentence (133) but less so as a contrastive reply to sentence (134). The contrast in acceptability is clearly perceivable and has been furthermore confirmed by Bianchi and Bocci through a larger empirical test involving 18 monolingual speakers of Italian. (p.217)
Bianchi (2012) and Bianchi and Bocci (2012) interpret the above contrast as evidence that left-peripheral focus is possible under corrective focalization but not under contrastive focalization. Bianchi (2012) also observes how the uncovered contrast goes against a strict cartographic perspective where even foci found in different positions in linear terms are claimed to occur in the same fixed focus projection above TP, since the observed pragmatic contrast should then be absent.
Interestingly, the observed contrast is expected under focus evacuation. In (132), the focused object must have been evacuated from the right-dislocated TP si era messa, or otherwise it would occur within TP. Right dislocation, however, can only apply to discourse-given constituents where givenness is intended as in Schwarzschild (1999). When (132) occurs as a reply to (133) this prerequisite is satisfied because the proposition that Mary had put a dress on is directly entailed via existential closure by the information that she had put on a cheap H&M dress. The same prerequisite is instead failed when (132) occurs as a reply to (134) because the observation that Mary was elegant is not sufficient to entail the proposition that she had put on a specific dress. In other words, the difference described in Bianchi (2012) and Bianchi and Bocci (2012) is genuine, but it emerges directly from the givenness conditions that must be satisfied for licensing right dislocation, rather than being formal properties associated with the different syntactic positions taken by contrastive foci.
If this analysis is correct, we expect (132) to be possible in contrastive focus contexts as well, provided that the right-dislocated TP is given. This is indeed the case. For example, (132) becomes felicitous again when the sentence is assessed as a reply to (135), which has a meaning very close to (134) but it also explicitly mentions that Mary had put on a dress, thus ensuring that the post-focus TP in (132) counts as discourse-given.
Bianchi and Bocci’s data thus provide further evidence for the focus evacuation analysis, since the contrast in acceptability that they observed is a direct consequence (p.218) of the right-dislocated status of the TP following focus and the associated givenness prerequisite.
5.3.6 Parasitic gaps
As mentioned in Chapter 3, the different positions of evacuated and non-evacuated foci become apparent when considering parasitic gaps, which are licensed by clause-initial foci but not postverbal ones.17 (Barolo is the name of a great Italian wine.)
In the (a) sentences, the focalized subject precedes the auxiliary and thus constitutes an evacuated focus followed by a right-dislocated TP. As such, the focused object adjoins TP and c-commands both variables after the focus evacuation step, enabling the licensing of the parasitic gap. The main derivational steps for (136)(a) are shown in (138).
(p.219) In the ungrammatical (b) sentences of (136) and (137), the focused object occurs post-verbally and is focalized in situ. Therefore, it is too low to c‑command and license the variable of the parasitic clause. As explained in Section 3.2, this structural distinction is unavailable in strict cartographic analyses that assume the focused object to be identical for both sentences, making it impossible to account for the observed alternation.18
Müller (1996) showed that in German and English remnant movement satisfies a principle of unambiguous domination that prevents a constituent that has undergone a movement of a specific type (scrambling, topicalization, wh-extraction) from being extracted from a constituent undergoing the same type of movement. As Müller points out elsewhere, it is possible to state this principle as in (139), which states that two constituents A and B cannot undergo the same type of movement when one is contained in the other (for further discussion including the relation with the Minimal Link Condition and the Phase Impenetrability Condition see Müller 1998, 2004, 2011).
Under specific circumstances, the structures associated with focus evacuation appear to violate unambiguous domination (thanks to Neeleman and van der Wal for pointing this out). This is for example the case with the post-focal phrases described in Section 5.3.4 of which (140) is an example. As the corresponding structure (141) shows, the object la mela and the TP abbiamo dato undergo the same type of movement—namely right dislocation—even though the object is generated in the TP. (The complete derivation is provided in footnote 13.)
In the rest of this section I will show that (i) the violation of unambiguous domination is a general property of Italian right dislocation that occurs even when focus evacuation is absent; (ii) that such violation occurs independently from the way right dislocation is represented; (iii) that the same violation also occurs under an analysis à la Rizzi (1997, 2004) not involving right dislocation. In other words, Italian right dislocation appears to be a genuine exception to Müller’s principle and the issue (p.221) raised by this observation is fully general, affecting all current analyses of right dislocation and/or left-peripheral foci, not just the analysis of focalization and right dislocation proposed in this book.
Consider sentence (142), which lacks contrastive focalization and focus evacuation. The object and the following VP are both right-dislocated, since they are preceded by the intonation break and optional pause typically preceding right-dislocated phrases. The object is also clitic-doubled by the clitic li, and thus cannot have just scrambled leftwards (Italian disallows for clitic doubling within a clause). The VP follows the right-dislocated object, therefore it, too, must be right-dislocated. The sentence violates unambiguous domination because the object and the VP undergo the same type of movement—namely right dislocation—even though the object is generated within the VP.19
The violation of unambiguous domination occurs whatever representation is assigned to right dislocation. Under the remnant movement analysis adopted here, (p.222) the trace of the right-dislocated object lies within the lower but equally right-dislocated VP, see (143). The same is true under the alternative right TP-adjunction analysis discussed in Appendix B, see (144). Since unambiguous domination holds for both bound and unbound traces (Müller 1996), the different hierarchical order of the right-dislocated constituents across the two representations is irrelevant.
Unambiguous domination is also violated under analyses of left-peripheral foci à la Rizzi (1997, 2004). Consider for example (145). The final object is right-dislocated, as attested by the presence of clitic doubling. Following Frascarelli and Hinterhölzl (2007), assume that this object occupies the specifier of a post focus topic projection above TP. But then the TP, too, must occur in a similar projection, since it precedes the object and follows the initial focus (not to mention the identical prosodic contour introducing the TP and the object and signalling their shared right-dislocated status). The resulting structure, in (146), violates unambiguous domination, since the trace of the object lies in a constituent, TP, which underwent the same type of movement (note that the focused PP must first exit the TP, yielding the additional tk trace in the lowest TP).
In conclusion, Italian right dislocation appears to be immune to unambiguous domination. While more research is needed in this area, this section also showed that the issue raised by this observation is independent from the specific representation of right dislocation and analysis of left-peripheral focalization proposed in this book.
The previous sections provided several independent arguments for analysing left-peripheral foci as focus evacuation based on the range of possible positions taken by evacuated foci and the right-dislocated status of post-focus constituents.
Under focus evacuation, the leftward movement of focused constituents is not an independent operation to be understood on its own terms and independently from any other operation affecting the clause. Rather, it follows from the interaction of two independently available operations, namely focalization in situ and right dislocation. Both can and do co‑occur in the same clause. When right dislocation targets a phrase containing a focalized constituent, a conflict is unleashed, since these operations cannot both take place because focused phrases cannot be right-dislocated. The result is the evacuation of focus from the dislocating phrase.
Focus evacuation differs from a strict cartographic alternative where all foci move to a fixed high focus projection in two fundamental ways. First, it views focus fronting as the exception rather than the rule, since it only occurs when the focalized constituent lies inside a phrase targeted by right dislocation. In all other cases, the vast majority, contrastive foci occur in situ. This enables focus evacuation to account for more sentences more accurately. Second, it views Rizzi’s left-peripheral focus data as a specific instance of focus evacuation that occurs when right dislocation targets TP. As we saw in Section 5.3.2, additional instances of focus evacuation involving smaller right-dislocated phrases are possible too.
Overall, the focus evacuation analysis successfully reduces any instance of leftward focus movement to the interaction of focalization and right dislocation. This provides a unified analysis of Italian contrastive focalization that spans over both (p.224) moved and unmoved foci, correctly predicting the actual position of focus in each case as well as the syntactic status of the constituents following it.
5.4 On the co-occurrence of focus and wh-phrases
The study of wh-extraction lends further support to the focus evacuation analysis. Under focus evacuation, the position of focused constituents is unrelated to the position of wh‑phrases. Therefore, since Italian shows no relativized minimality effect between the two, wh-phrases and foci are predicted to co-occur within the same clause. As this section will show, this prediction is amply borne out in both main clauses and subordinate clauses. This result is particularly significant because it refutes a crucial piece of evidence claimed in support for the existence of a higher focus projection in Rizzi (1997, 2001) and since then almost unanimously maintained as valid in all the literature that followed: namely, the alleged impossibility for wh-phrases and foci to co-occur in main clauses.
Building on Samek-Lodovici (2006), but adding new data concerning main clauses, this section will show that the distribution of wh-phrases relative to contrastive foci is governed by the position of the wh-chain relative to the right-dislocated constituents following a focus. As summarized in table (147), there are only three possible ways in which a wh-chain and a right-dislocated constituent can overlap in a sentence involving focalization: (i) the wh-chain lies entirely outside any right-dislocated phrase; (ii) the wh-chain originates in a right-dislocated phrase but is not entirely included in it (i.e. the wh-phrase is extracted from it); (iii) the entire wh-chain lies within the right-dislocated phrase, which in these cases is usually a clause. Wh-extraction will be shown to co-occur with focalization in all three cases except those involving extraction from a clitic-doubled phrase.
This observed distribution is consistent with the proposed analysis of focalization: nothing blocks wh‑extraction except for clitic-doubled right-dislocated phrases, since they form strong islands to extraction as discussed in Section 4.4.4 (see also Cardinaletti 2002; Samek-Lodovici 2006). The observed co-occurrences of wh-phrases and contrastive foci (p.225) are instead unaccounted for if wh‑phrases and focalized constituents are assumed to share the specifier of a high fixed focus projection whether at surface or at LF, since obviously they could not occur in the same clause. In the following, I consider each case in turn focusing on main clauses and only briefly discuss subordinate clauses—already examined in Samek-Lodovici (2006)—at the end of the section.
5.4.1 Wh-chain outside right-dislocated phrases
The simplest way to examine this case is to consider sentences lacking right dislocation altogether.20 The corresponding sentences necessarily involve focalization in situ, since focus evacuation is triggered—and hence always followed—by right dislocation. We can, however, distinguish the following two subcases depending on whether the wh-operator is generated above the focused constituent or below it.
Wh-extraction is grammatical, and in fact very natural, in either case, showing that contrastive focalization is not subject to relativized minimality effects. The following sentences respectively involve extraction of a wh-subject generated above a focused object, extraction of a wh-object across a higher focused subject, and extraction of a temporal wh-adverbial across a focused subject. For space reasons, context sentences are placed in parentheses.
Under the focus evacuation analysis, the successful co-occurrence of wh-operators and foci in-situ in all the preceding examples is expected. Foci and wh-phrases do not share the same position at any stage of the derivation, therefore wh‑operators may successfully raise to their final position. The opposite holds of analyses maintaining that wh-operators and contrastive foci share the same position, whether at surface or LF, since they incorrectly predict all the above sentences to be ungrammatical.
5.4.2 Wh-chain across a right-dislocated phrase
As discussed in Section 4.4.4, wh-operators cannot be extracted from right-dislocated constituents when clitic doubling is present but they may be marginally extracted when clitic doubling is absent. The extraction of wh-phrases from within a right-dislocated clause and across a focused constituent is predicted to show the same pattern. It should be impossible when clitic doubling is present and marginally available when clitic doubling is absent.
Consider first the case where focalization occurs in situ. The following three examples vary in the type of wh-operator and constituent being focused. In each example, sentence (a) illustrates the case where clitic doubling blocks wh-extraction while sentence (b) concerns the case lacking clitic doubling and allowing for (p.227) wh-extraction. In all (b) sentences, the right-dislocated clause follows a clitic-doubled right-dislocated indirect object; this is necessary to ensure that the non-clitic doubled clause is right-dislocated rather than marginalized in situ.
(p.228) Wh-extraction from a right-dislocated phrase is also possible with evacuated left-peripheral foci, i.e. the cases excluded in Rizzi (1997). In the following examples, wh-extraction occurs from the right-dislocated TP immediately following left-peripheral foci, as schematized in (158). As discussed in Section 4.4.4, wh-extraction from tensed right-dislocated TPs like this one is marginal even when contrastive focalization is absent, therefore the marginal status of the following examples is to be expected and unrelated to the presence of left-peripheral foci. All sentences also inevitably lack clitic doubling because the clitic-hosting head T is part of the dislocated constituent.
(p.229) These assessments do not match those reported for the same patterns in Rizzi (1997: 291) which reports sentence (162) as ungrammatical. Rizzi provides this sentence without a discourse context. The lack of a context makes the interpretation of the post-focus clause as right-dislocated problematic, because discourse-givenness is a key prerequisite for licensing right dislocation. Consequently, speakers are forced to approach (162) as if it lacked right dislocation and under this forced interpretation (162) is indeed ungrammatical. And we know why: under focus evacuation you can’t have a left-peripheral focus without the following TP being right-dislocated; when the TP is not right-dislocated, focus should occur in situ.
When the same sentence is uttered with respect to an appropriate context and with an appropriate intonation, it is grammatical. See the examples (163) and (164), involving a context as well as a more contentful TP, so that it can be more naturally re-proposed to the hearer’s attention in right-dislocated form.
In conclusion, even the second class of wh-extraction structures patterns as predicted by focus evacuation, being marginally possible across in situ and evacuated foci when clitic doubling is absent and blocked when clitic doubling is present. In contrast, if contrastive focus and wh-operators shared the same position, all preceding sentences ought to be ungrammatical.
The third and last possible configuration occurs when the entire wh-chain is contained within a right-dislocated constituent, as shown in (165). In this case, there is no extraction of the wh-operator from the dislocated clause and nothing prevents these structures from being grammatical. The following grammatical examples, involving focalization in situ confirm this prediction.
Right-dislocated interrogatives of this kind may also follow evacuated foci found in clause-initial position. Like any other right-dislocated phrase, the interrogative clause must be discourse-given and therefore these sentences must be assessed relative to a discourse context that satisfies this requirement. The dislocated interrogative must also receive the typical intonation of right-dislocated phrases in declarative clauses, rather than the intonation of questions, since due to their right-dislocated status these interrogatives act more like a reminder of a previous question than a newly posited question. When these conditions are satisfied, the corresponding structures are grammatical, as shown by the following examples.
Like the data in the previous section, the assessment provided for these patterns do not match the one provided in Rizzi (1997: 291). Once again, however, it is important to observe that Rizzi’s original sentence, repeated in its original form in (171), is presented without a discourse context, making it difficult for speakers to approach the post-focal interrogative clauses as right-dislocated. The initial focus in (171) is also not followed by the comma representing the intonational break and brief optional pause typically preceding right dislocated constituents, and this, too, discourages the right-dislocation reading required for grammaticality. Furthermore, (171) ends with a question mark, thus strongly suggesting the raising interrogative intonation of standard questions. But this intonation is incompatible with right-dislocated status, making a right-dislocated reading of the post-focal interrogative completely impossible. Under these circumstances, (171) is inevitably perceived by a native speaker of Italian as a sentence where right dislocation is absent and yet focus has raised above CP nevertheless, as schematically shown in (172). Such a structure is indeed ungrammatical and predicted to be so even under focus evacuation, since focus may move only when right dislocation is present.
Even Rizzi’s original example becomes grammatical when presented under a suitable context, with a comma after the initial focus, and when assigned an exclamative (p.232) intonation; see (173) where the subject has been changed to a plural or else no suitable context is possible. You should imagine two people speaking with A asking the first question—this is the context—and B using Rizzi’s sentence to correct the initial question. The context and the exclamative intonation license an interpretation where the interrogative clause is discourse-given. This, in turn, enables its right dislocation, which triggers focus evacuation and thus makes the sentence grammatical.22
In conclusion, contrastive foci and wh-operators can co-occur in the same main sentence. This holds of in-situ and left-peripheral foci alike, as expected under focus evacuation. Obviously, the prerequisite for focus evacuation and wh‑extraction must be satisfied: the right-dislocated interrogative clause must be licensed by a suitable context and intonation and, where relevant, the interaction of wh-extraction with clitic doubling must be taken into account.
5.4.4 Subordinate interrogative clauses
The observations made so far for main clauses are also valid for subordinate clauses. A few examples follow. Sentence (174) shows co-occurrence of wh-operator with postverbal foci, while (175) shows clause-initial foci preceded and followed by a wh-phrase respectively. (p.233)
Indeed, the co-occurrence of foci and wh-phrases in subordinate clauses is also acknowledged in Rizzi (1997: 330, 2001). For example, Rizzi (1997: 330, footnote 18) mentions (176) as marginally grammatical, and personally I find this sentence fully acceptable provided it is uttered with an exclamative intonation.
Rizzi (2001) accounts for these sentences by proposing that the left-periphery of subordinate clauses differs from that of main clauses in that focus and wh-operators target distinct projections. There are two problems with this analysis, though. First, no reason is provided, nor does one come to mind, for why the distinct focus- and wh- projections available in subordinate clauses should not be available in main clauses. Unless a reason is identified, the stipulation that such projections are present in subordinate clauses but absent in main clauses lacks in explanatory power, since it offers just a description of the observed data.
Second, Rizzi’s (2001) analysis appears to assume that in subordinate clauses left-peripheral foci precede wh‑operators, but this is incorrect. As (176) showed, left-peripheral foci may also follow wh-operators. Two additional examples are provided here.
Under focus evacuation, the attested parallelism between main and subordinate clauses is instead expected. Since right dislocation applies to main and subordinate clauses alike, focus evacuation applies to both types of clauses too, with identical outcomes relative to its interaction with wh-phrases.
5.4.5 An aside on the position of right dislocation
The availability of right-dislocated interrogative clauses shows that the final position of right-dislocated phrases is, to a limited extent, dependent on what constituent is dislocated. As discussed in Chapter 4, right-dislocated phrases occur outside the extended projection of V (or N in nominals, see Samek-Lodovici 2010). In most cases, including the focus evacuation cases discussed in Section 5.3, the extended projection at issue coincides with TP. When the dislocated constituent includes an entire wh-chain or a major part of it, as is the case with the sentences discussed in Sections 5.4.2 and 5.4.3, the dislocated phrase is necessarily larger than TP, and consequently right dislocation, too, occurs higher than TP. For example, consider again sentence (173), repeated in (179) without its licensing context.
The right-dislocated constituent includes the wh-operator che cosa and must therefore be larger than TP; for convenience I will consider it a CP, but a more precise label taking into account the internal structure of CP—except for FocusP—would be equally fine. Since the entire CP containing the main clause is right dislocated, the RP projection hosting the dislocated CP in its specifier must itself occur above CP. The final structure is provided in (180). It is determined by the following derivational steps: (i) the wh-operator moves to specCP; (ii) the focused indirect object left-adjoins to CP as part of focus evacuation; (iii) the CP is right dislocated to the specifier of RP; (iv) the entire CP remnant including the adjoined focus moves to specXP. (p.235)
This analysis tells us that the position of right dislocation is not structurally fixed. As shown in Chapter 4, it must occur outside TP. But whether it occurs right above TP, as in most cases considered in this book, or right above CP as in the case just discussed depends on the size of the dislocated constituent. The correct generalization is that right dislocation occurs above TP but as low as the specific content of the sentence and of the dislocated phrases allows.
More research is needed on this aspect of right dislocation, examining how high in the structure it can be pushed and why TP provides a lower bound. But the above discussion shows that even the distribution of right dislocation cannot be accounted for through a strict cartographic analysis. Rather, its distribution resembles that of evacuated foci in that it occurs as low as structurally possible (once the basic condition that it be located outside TP is satisfied), but it can be pushed higher when that is necessary for the satisfaction of specific constraints, such as the requirement that wh-operators be raised to specCP.
In Italian, the simultaneous presence of wh-extraction and contrastive focalization in the same clause does not give rise to relativized minimality effects. Wh-extraction is always possible except when originating in a clitic-doubled right-dislocated phrase. This distribution is expected under the focus evacuation analysis, but surprising under any analyses requiring foci and wh-operators to share the same position, whether at surface or LF.
Wh-operators may precede or follow focused constituents in both main and subordinate clauses. When they follow focus, the corresponding interrogative is discourse-given and right-dislocated, i.e. simply repeated for the benefit of the hearer. In these cases, the associated intonation is affected accordingly and it is incompatible with the raising intonation of interrogative clauses.
Together, focalization in situ and focus evacuation provide a unified account for the entire distribution of contrastive focus, explaining all possible linear orders displayed (p.236) by focused, marginalized, and right-dislocated constituents in Italian clauses and the distinct syntactic properties associated with each constituent under each order.
The different positions taken by evacuated and in-situ foci also explain the attested asymmetries affecting the syntax of negative phrases, parasitic gaps, and wh-extraction with respect to these two sets of focalization cases, all following from the availability of post-focal marginalized constituents after in-situ foci and their absence in the relevant cases involving evacuated foci.
In so far as the analysis advocated here is correct, it forces us to reconsider the many analyses of the information structure of Italian revolving around the existence of a dedicated focus projection above TP, since the conclusions that have been drawn on the basis of that assumption can no longer be maintained as valid.
(1) Vallduví (1992) provides interesting arguments for the in-situ analysis of contrastive foci. To my knowledge, Vallduví’s work also first pointed out the relevance of right dislocation for an appropriate understanding of the distribution of contrastive focalization. His analysis differs from the one presented here in that it extends in-situ focalization to instances of clause-initial focalization, which I instead claim to occur only when focus is contained in a larger phrase targeted by right dislocation and to require focus movement out of the dislocating constituent, see the rest of this introductory section and Section 5.3.
(2) When lower adverbs are focused, as in (i), they may follow the past participle while retaining narrow scope. This, too, is expected, since the adverb is focused in situ and hence within the scope of negation. The past-participle’s movement above the adverb constitutes an instance of the movement operation examined in Chapter 3 letting lower-generated unfocused phrases optionally raise above a higher focus.
(3) Cinque does not provide minimal pairs illustrating the rigid order between these adverbs, so I added a few examples here. Example (i) shows that the evaluative adverb per fortuna ‘luckily’ must precede the modal adverb probabilmente ‘probably’ under clause-wide focus. Yet the same two adverbs are freely ordered when occurring in post-focus position in (ii). Likewise, evaluative purtroppo ‘unfortunately’ precedes forse ‘perhaps’ under clause-wide focus in (iii), but the two are freely ordered after focus in (iv).
(4) Fanselow and Lenertová (2011) examine cases of partial focus fronting in German and Czech where the fronted item constitutes a subpart of the focused constituent. An example for Italian is provided in (i), where the context question in (i)A focuses the entire VP, yet the reply in (i)B only fronts the direct object.
These cases, too, could conceivably be reconciled with the analysis proposed in this book by assuming that in (i)B focalization is limited to the evacuated object and the following TP is treated as discourse-given and right dislocated. Under this hypothesis what is noteworthy in the dialogue in (i) is the fact that speaker B ignores the focalization and givenness assignment imposed by A’s question and replies as if focalization applied to the object alone. Preliminary support for this hypothesis comes from the observation that partial focus fronting deteriorates when the information treated as discourse-given by B is less obvious, and therefore arguably less easily accommodated as discourse-given by the dialogue participants. For example, as (ii) shows when B’s reply contains more information, partial focus fronting becomes infelicitous (as represented by the symbol ‘#’). The constituent imposed as discourse-given in (ii)B is more informative and less immediately associated with the fronted focus as is the case in (i)B, possibly because mushroom risottos are typically and inevitably eaten, whereas ordering them at the restaurant is a more specific activity contrasting many other conceivable ones (cooking risotto at home, cooking it at a restaurant, eating it at home, eating it at a restaurant, etc.).
A further development of the hypothesis just described would have to consider how to account for the interesting properties associated to partial focus fronting described in Fanselow and Lenertová (2011). Ideally, it would also identify the conditions determining when a speaker can manipulate the focus and givenness assignments associated with the current discourse context in the way partial focus fronting appears to be doing.
(5) While the constituent originally containing the evacuated focus is necessarily right-dislocated, it is not possible to conclude that every constituent following an evacuated focus is right-dislocated in all circumstances, as mentioned in Samek-Lodovici (2006: 1, 2009: 334). This conclusion is not possible because a focused constituent XP could be generated within a phrase YP itself located before another lower constituent ZP, as shown in (i). As discussed later in this section, right dislocation of YP in this case causes the evacuation of the focused XP to a left-adjoined YP position which would still precede the unmoved ZP as schematically shown in (ii) where ‘[YP…tk…]R,i’ is the right dislocated YP constituent originally including the focused XPF.
Note, however, that for this configuration to occur YP must be strictly contained in the TP relative to which right dislocation takes place. When right dislocation targets the entire TP, the evacuated focus necessarily precedes the entire right-dislocated TP, as in (iii). In this case, a ZP immediately following the subject could no longer be in situ, since it is outside its original TP. As will be explained later in this section, in this case ZP is itself right-dislocated independently from TP, as schematically shown in (iv).
(7) A possible alternative, here left unexplored, is that right dislocation marks the domain of focus, in accord with the hypothesis that focus movement marks the focus background proposed in Neeleman et al. (2007) and Neeleman and van de Koot (2008).
(8) As a native speaker, I find left-peripheral focalization possible with contrastive, corrective, and even presentational focus, but on the possible types of focalization available in this position see also Calabrese (1992), Rizzi (1997), Belletti (2001, 2004), Brunetti (2004, 2009), Bianchi and Bocci (2012), and Bianchi (2012).
Brunetti (2004) argues that the unavailability of left-peripheral presentational focus reported by some speakers follows from the specific tests used to distinguish the two focus types. For example, she claims that QA-pairs, like (i) are inadequate as a diagnostics for new-information focus because for these speakers the background proposition ho mangiato ‘I have eaten something’ is too salient and accessible to be repeated again in the answer as a discourse-given constituent as in answer A1. Other speakers, however, are less constrained by the salience of the background proposition and will accept (i)A as grammatical. Most of my informants and myself belong to the second group, since we find (i)A grammatical when the post-focus clause is uttered with the typical intonation of right-dislocated constituents. Calabrese would appear to belong to this group too, as a similar pair involving a left-peripheral focused subject is reported as grammatical in Calabrese (1992: 100). Interestingly, and supporting Brunetti’s point, even the informants of mine who found A1 marginal, found A2 acceptable, where further repetition strengthens the discourse-givenness of the clause ho mangiato ‘I have eaten something’, making its right dislocation acceptable in the last clause.
(9) Sentences (i)–(iii) supply additional examples but with non-subject negative foci. These cases, too, fail to establish a c-command relation between licenser and licensee due to the right-dislocated position of the post-focus TP. Unlike negative subjects, however, these negative constituents do not occur as licensers (i.e. located above T) under clause-wide presentational focus and are thus less suitable for examining the effects of focus evacuation on licensing.
(10) The syntax of the neg-marker non is discussed in several, very different analyses; see among others Zanuttini (1991), Ladusaw (1992), Isac (2004), Zeijlstra (2004), Penka (2011). Many of these analyses do not address the alternation discussed here. No analysis, amongst those surveyed, takes into account the right-dislocated nature of post-focus TPs.
(11) The presence of a double negation interpretation for this type of sentence is widely attested. Nevertheless, the literature on negative concord also reports cases allowing for a simple negative interpretation under an informal/colloquial register and characterized by variable acceptability across different speakers. See Godard and Marandin (2006), Manzotti and Rigamonti (1991), and Benincá (1988).
(12) Interestingly, focus evacuation appears to also extend to absolute participial constructions. Sentences like (i), typically uttered with an exclamatory intonation, are deficient in that they lack an auxiliary. Under sentence-wide presentational focus the initial negative adverb acts as licenser for the following negative object, see the contrast between (i) and (ii). Yet licensing collapses when the same adverb is focused (and hence stressed) as in (iii). Note that when the negative object is replaced with a non-negative counterpart and licensing is no longer required focusing the initial adverb is unproblematic, see (iv). As usual, it is essential to assess these sentences in the context provided.
If these constructions only involved deletion of the top TP projection, we would expect the negative adverb and object in (i) to be ungrammatical, as they would be located lower than T and need a licenser. Nor can we assume that licensing occurred prior to deletion through a neg-marker contained in the deleted TP projection, as in this case the negative object in (ii) could be licensed too in the same fashion.
Rather, the fact that the negative adverb mai ‘never’ is both licensed and acting as a licenser in (i) suggests that the root projection of these constructions—let me call it AbsP for ‘AbsoluteP’—has propositional import similar to the one found in TP and that the adverb has raised from its base-generated position located between T and the lowest available position for past-participles (Cinque 1999) to the specifier of AbsP, where it is able to pass its negative feature to AbsP and to license the lower object; the resulting structure is provided in (v).
The ungrammaticality of (iii) then follows straightforwardly from the syntax of focus evacuation. The adverb cannot focus in situ, since its base-generated position is too low to be licensed. When the entire AbsP is targeted for right dislocation, however, the focused adverb is left-adjoined to AbsP, as shown in (vi), thus reaching a sufficiently high position for its own licensing and accounting for (iv). Nevertheless, it would be unable to c-command the negative object in the dislocated AbsP, hence accounting for (iii).
Obviously, important aspects of the analysis just sketched must be further researched in order to clarify the nature of AbsP and its role in the licensing of negative phrases. The focus evacuation analysis, however, provides a preliminary account of why the focalization and licensing patterns in these constructions mimic so closely that of focused negative subjects in TPs.
(13) Sentence (72), repeated in (i) below, places the right-dislocated object la mela before the right-dislocated TP abbiamo dato. In order to obtain this order, the object must move out of the original TP before the TP’s right dislocation. It is this movement, shown in step 2 of derivation (ii), that is responsible for the TP‑adjoined trace ‘tj’ following the focus in the final structure.
(14) These two topic classes have been given different names across distinct works. CLLD and HT topics are respectively called ‘CLLD’ and ‘LD’ (left dislocation) in Cinque (1990), and ‘LD’ and ‘HT’ in Benincà (2001). I used ‘CLLD’ and ‘HT’ and avoided the ambiguous ‘LD’.
Both HTs and CLLD topics express salient discourse-given referents and occur above TP. CLLD topics can be expressed by any syntactic category, whereas HTs can only be expressed as DPs or NPs, even when relating to an argument normally introduced by a preposition (Cinque 1990; Benincà 2001; Benincà and Poletto 2004). For example, the PP a Maria ‘to Mary’ in (i) is a CLLD topic whereas the DP Maria in (ii) is an HT since it expresses the indirect argument of the verb without the preposition normally introducing it. As a native speaker, I perceive the two sentences as having the same meaning, but further research is needed to determine whether their pragmatic import is indeed identical.
HTs and CLLD topics are also distinguished by clitic doubling, which is always obligatory for HTs and optional—but for objects where it is obligatory—with CLLD topics. This, too, is illustrated by the two sentences in (1) and (2), with the CLLD topic in (1) showing an optional indirect object clitic le ‘to-her’, while the same clitic is obligatory with the HT in (2).
DPs and NPs expressing subjects and objects are ambiguous between a CLLD and HT analysis because their category, the lack of a preposition, and their clitic doubling properties cannot distinguish between the two. Unambiguous testing of HTs requires DPs and NPs expressing arguments normally requiring a preposition (Benincà 2001), while CLLD topics require non-nominal phrases.
Native speakers can easily distinguish evacuated foci, including clause-initial foci, from HTs and CLLD topics because evacuated foci carry main stress, whereas HTs and CLLDs do not and require a distinctive intonation of their own (see Frascarelli and Hinterhölzl 2007).
(15) Bare NPs can be marginalized in situ, see (i). The absence of right dislocation in this sentence is confirmed by the impossibility of clitic doubling, see (ii), which follows from the condition C violation due to the c-command relation between the clitic and the NP. In the examples in the main text, marginalization is controlled for by the addition of a right-dislocated indirect object.
(16) Benincà (2001: 45) notes that left-peripheral focus might allow for optional clitic doubling when the focused constituent is an indirect object, even though the choice of lexical verb appears to play a role too. Her examples follow in (i) and (ii).
While writing this book, I found a second set of cases that involves object DPs where the quantifier is stranded in the base-generated position of the object. Interestingly, the clitic is obligatory here, rather than optional. Compare (iii) and (iv) (past participles agree with the object clitic in gender and number).
(17) The licensing of parasitic gaps by clause-initial focus is reported absent in Catalan (Villalba 2000: 255), suggesting the presence of crosslinguistic differences in the analysis of focalization across these two languages.
(18) Even the weaker hypothesis that focus raises to a high fixed focus projection covertly is problematic. Engdahl (1983) showed that simple parasitic gap constructions cannot be licensed under covert movement, as shown by her example in (i). Nissenbaum (2000: 12) showed that under appropriate conditions requiring multiple interrogatives the licensing of parasitic gaps by covert movement becomes possible in English, see (ii). This observation, however, does not hold with Italian in-situ focalization, see (iii). Overall, covert movement does not appear to license parasitic gaps in Italian and is thus not a solution for the alternation discussed in the main text.
Which senator (you) have persuaded to take the jeep, without before to-call an opponent to put a bomb
(19) Two additional examples are provided here. The first, in (i), involves right dislocation of the intermediate CP and the final object; note how both are clitic-doubled. This sentence violates unambiguous domination for the same reasons considered in the main text provided that it is not possible for the object to right dislocate within the right-dislocated CP, as schematically shown in (ii). If such embedded dislocation is permitted, the dislocated object is part of the CP, and since the CP would also count as the object’s chain domain, unambiguous domination would be satisfied within CP.
The second example is provided in (ii). Contrary to appearances, it consists of a single clause where the modal verb vorrei acts as the auxiliary (Cinque 2004). The object gli amici and the final VP vederli più spesso are both right-dislocated. They are preceded by the intonation break and optional pause preceding right-dislocated phrases and the object is clitic-doubled by the clitic li on the verb.
(20) The first case also includes the complex but here uninformative case where right dislocation applies to an item generated within the wh-phrase itself, as in (i). The resulting sentence is grammatical because right dislocation can occur before wh-extraction thus ensuring that both movements do not occur from an unselected specifier. An example is provided in (ii).
(21) This is the government department for the collection of taxes. I used the name used in the USA. The British name is HMRC (Her Majesty’s Revenue and Customs).
(22) A similar point applies to sentence (i), which is listed as ungrammatical in Abels (2012) based on the judgement of a native speaker informant. This sentence, too, becomes grammatical when an appropriate intonation and context are provided, as is the case in (ii), where the added comma after the focus and the final exclamation mark suggest a more suitable intonation, and (iii), where an adequate context ensures the discourse-givenness of the right-dislocated clause.