Contrastive focus and marginalization
Contrastive focus and marginalization
Abstract and Keywords
This chapter examines the distribution of contrastive focalization when any present post-focal constituent is marginalized in situ, i.e. when no constituent is right dislocated. In these cases contrastive foci occur in situ and any attempt to analyse them as having moved to a higher left-peripheral focus projection above TP or intermediate focus projection located between TP and VP quickly runs into serious problems. This chapter also establishes the existence of an independent operation moving lower discourse-given constituents following a focus to a position immediately above it, potentially making the affected focus clause-rightmost. This operation will be eventually reconsidered in the final chapter, where it will be shown to follow from the constraints requiring Italian stress to occur clause-rightmost. Finally, the chapter examines the binding relations between postverbal foci and postverbal unfocused constituents.
This chapter investigates the position of Italian postverbal contrastive foci showing that it cannot be captured by templatic analyses positing a unique focus projection (e.g. Rizzi 1997, 2004; Belletti 2004: 29; Cecchetto 1999). Once right dislocation is controlled for, the movement of focused constituents predicted by these analyses is demonstrably absent. The available evidence instead supports focalization in situ (see also the evidence in Chapter 5 emerging from the study of the interaction of focus and right dislocation). This chapter also examines movement of postverbal constituents across different information status assignments, showing that postverbal constituents may raise above a preceding phrase only when they are discourse given and the preceding constituent is focused. Both results will be accounted for in Chapter 6, where I will argue that focalization occurs in situ in order to keep stress as close to the right edge of the clause as possible, consistently with the constraints governing the allocation of prosodic prominence in Italian. Similarly, postverbal constituents may raise only above a higher focus and only if they are discourse-given because under only these circumstances does movement improve the alignment of the stress associated with focus with the clause right edge, again in accord with prosodic requirements.
In Italian, contrastive focus in postverbal position occurs naturally. A native speaker accidentally hearing sentence (1) naturally interprets the postverbal subject as contrastively focused, even if unaware of the preceding discourse context and even if the final stress is of the non-emphatic kind found in simple declaratives. The likelihood of a contrastive reading can be increased by overtly listing the possible alternatives in the sentence itself, as in (2), or in its immediate discourse context, as in (3), but this is not a necessary requirement. (p.43)
Most of the evidence for in-situ focalization discussed in this chapter is obtained by closely examining the distribution of postverbal focused phrases relative to marginalized constituents. Given a discourse-given constituent A and a contrastively focused constituent B generated lower than A as in (4)(a), raising focus to a higher projection would place B above A as in (4)(b) whenever A is marginalized in situ. In contrast, in-situ focalization would leave A before B as in (4)(a) and predict the order in (4)(b) to be ungrammatical.
This test is only possible if right dislocation is controlled for, as the order <B A> may also arise when focus remains in situ and the higher constituent A is right-dislocated to the right of B. Disentangling marginalization from right dislocation requires some care. Right dislocation is possible also without clitic doubling (see Chapter 4) and can easily be confused with marginalization. As the following sections will show, once these crucial controls are in place, postverbal focused phrases turn out to never precede higher-generated discourse-given phrases, which, in turn, supports focalization in situ.
Controlling for right dislocation will also be essential to establish the second empirical result of this chapter, namely that postverbal constituents may raise above a higher one only when they are discourse-given and the higher constituent is focused, as in (5)(a). This will require establishing that the same movement is ungrammatical whenever the higher constituent is not focused as illustrated by the patterns in (5)(b) and (5)(c), which in turn require controlling that A is marginalized rather than right-dislocated, since grammatical instances of order (5)(b) do exist but are due to A’s right dislocation rather than B’s movement. (p.44)
I will start in Section 3.2 by showing that positing focus movement to a left-peripheral focus projection is sharply inconsistent with the properties displayed by postverbal foci relative to the licensing of n-words and NPIs, parasitic-gaps, and wh-operators. Section 3.3 examines movement to the TP-internal focus projection proposed in Cecchetto (1999) and Belletti (2001, 2004), showing that this movement, too, must be excluded. Section 3.4 examines the movement of discourse-given constituents above a higher focus and its absence when the higher constituent is not focused. Section 3.5 provides two further tests for the established results by checking whether they generalize to Cinque’s lower adverb hierarchy and then by testing the binding relations they predict.
3.2 In-situ vs. left-peripheral focalization of postverbal foci
Belletti (2004) assumes that all contrastive foci must move to the high focus projection of Rizzi (1997). She accounts for postverbal contrastive foci by assuming that they move to the focus projection just mentioned, followed by the movement of the entire TP remnant to an even higher Topic projection, so that the raised foci still occur postverbally in linear terms. A sentence like (6) would thus have the analysis in (7).
There are several important pieces of evidence against this analysis and in favour of in-situ focalization. Let me begin with n‑words, which, as described in appendix A, must be licensed by an appropriate c‑commanding licenser (p.45) when located postverbally and lower than T, but need no licensing when preverbal and higher than T. Focused n-words follow the same generalization: they require licensing if postverbal—see (8) where the initial licensing neg-marker non cannot be omitted—but they do not require licensing if preverbal, see (9).
If the postverbal n-words in (8) are focused in situ, their need for licensing is expected, since all n-words c-commanded by T need licensing. The same property cannot be accounted for under a left-peripheral analysis à la Belletti (2004) where the postverbal foci in (8) share the same position of the preverbal foci in (9), since this incorrectly predicts that postverbal foci do not need licensing.
Besides predicting that licensing is unnecessary, a left-peripheral analysis also incorrectly predicts licensing to be impossible. As observed by Cardinaletti (2002), Samek-Lodovici (2006, 2009), Brunetti (2004), and, for Zulu, Cheng and Downing (2009), the associated representation does not allow for the necessary c‑commanding relation between the initial neg-marker non and the n-word. For example (8)(a) would be expected to have the representation in (10), where non does not c-command NESSUNO and therefore cannot license it (p.46) (remember that this licensing relation is not restored under reconstruction, see appendix A).1,2 (p.47)
(p.48) The same problem applies to focused postverbal NPIs such as alcunché in (11). While licensing is expected and possible if the focused NPI occurs in situ, it is no longer possible if the NPI is analysed as hosted in a focus projection above TP where it would no longer be c‑commanded by its licenser.
Another argument supporting focalization in situ for postverbal foci comes from the distribution of wh‑phrases relative to contrastive focus. As Rizzi (1997) observed, there are sentences where wh-phrases and left-peripheral contrastive focus appear to occur in complementary distribution when they are both preverbal as in (12). Rizzi derived this effect by maintaining that focus and wh-operators compete for the specifier of the same fixed focus projection (but see Section 5.4 in Chapter 5 for a different analysis). If postverbal foci, too, are raised to the same specifier, then they too should show a similar complementary distribution relative to wh-phrases. Yet this is not the case. See for example question (13), asking whom John—rather than someone else—was introduced to. When main stress is placed on the focused object, (13) is perfectly natural and grammatical, despite the co-occurrence of wh-phrases and contrastive focus. This is unexpected if the focused object and the wh-operator are competing for the same position in this sentence. It is instead accounted for if the postverbal focused object occurs in situ as in structure (14).
(p.49) A final argument comes from an interesting asymmetry in the distribution of parasitic gaps. As (15) and (16) show, fronted focused objects license parasitic gaps whereas postverbal focused objects do not. If postverbal foci focalize in situ, the data in (16) are expected: as (17) shows, in-situ focalized objects are too low to c-command the parasitic gap and therefore they cannot license it either.
In contrast, when analysed as left-peripheral foci, the postverbal focused objects in (16) are maintained to share the same position of the fronted foci in (15) and are consequently incorrectly predicted to behave similarly and license the parasitic gap to their right. The correspondent structure is provided in (18).3 (p.50)
The evidence discussed so far provides some strong reasons in support of focalization in situ while also highlighting some serious problems for a left-peripheral analysis of postverbal contrastive foci. In the next section, I will show that focus movement to a lower intermediate focus projection just above VP is also excluded.
3.3 In-situ focalization vs. raising to an intermediate focus projection
Brunetti (2004: 124, section 5.5) argues at length for the in-situ position of postverbal foci, including a detailed discussion and refutation of the arguments provided in Belletti (2001) in support of the presence of an intermediate focus projection.4 This section provides additional evidence for analysing postverbal focused phrases as positioned in situ.
As mentioned in the introduction, given two phrases A and B generated within VP with A higher than B, if focused phrases raise to an intermediate focus projection above VP the order <BFAM> should be attested, since, on its way to the focus projection, BF will necessarily raise above AM, which is marginalized in situ and therefore inside VP. If, instead, focalization occurs in situ, we expect the order <BF AM> to be impossible, since B cannot move above A. In this latter case we also expect the order <A BF > to be possible because BF would occur below A whether (p.51) A remains in situ or moves to a local higher position (note that A in this case is not necessarily marginalized, since marginalized constituents are post-focal; for this reason it is not marked with the subscript ‘M’).
As we will see, the order <BF AM> is always ungrammatical and the order <A BF> always possible. Contrastive focus in postverbal position thus does not raise to an intermediate focus projection; rather, it occurs in situ. I start by testing postverbal subjects and objects, then consider experiencer objects and sentential complements, then subjects and sentential complements, and finally Cinque’s lower adverbs.
3.3.1 Postverbal subjects and objects
The first test concerns the ordered pairs <S OF> and <OF SM> involving discourse-given subjects and contrastively focused objects. If focalization required raising to an intermediate projection above VP à la Cecchetto (1999) and Belletti (2004), the focused object should be able to precede the marginalized subject, which occurs in situ. This prediction is refuted by the data in (19) and (20): the focused objects always follow the subject as expected if focalization occurs in situ.
The test uses a negative subject to ensure that it is truly marginalized in situ and not right-dislocated when it follows the focused object in sentence (b); as we saw in Chapter 2 negative phrases resist right dislocation. The same caution is unnecessary with the focused object because foci are never right-dislocated.
All sentences are designed so as to avoid any ambiguity in the identification of the subject and the object, since this easily interferes with grammaticality assessments. The subject is always singular and the object plural, so that auxiliary agreement univocally identifies the subject. The size of the subject and the object is also systematically varied to control for potential effects related to prosodic size.5 All sentences must be assessed with respect to the provided discourse context (p.52) or else marginalization is not triggered, leaving the corresponding sentences ungrammatical for an irrelevant reason. Finally, be aware that as explained in Section 2.3.1 of Chapter 2, this type of test is unsuitable for those native speakers of Italian that always disallow for non-final VP-internal negative subjects. Even these informants, however, converge with other speakers when VP-internal subjects are not involved; see for example the examples in the next subsection, which follow the same logic of the examples below but lack VP-internal subjects.
(p.53) Note that the above data are not sensitive to the particular order of subject and object in the context sentences. The same assessments emerge when the context is changed into an assertion or a question involving a preverbal subject rather than a postverbal one. For example, the assessments in (21)(a)–(b) also hold when elicited under the two alternative contexts in (22), where the subject is mentioned before the object. The same holds for the other data presented above.
The observation that focused objects cannot raise above marginalized subjects is also confirmed by the following data from Anconetan Italian (Cardinaletti 2001: 131), where the marginalized status of the subject is guaranteed by the lack of agreement (see Section 2.3.3). If focused objects could precede marginalized subjects, we would expect the order <OF SM> to be possible even under default agreement, but this is not the case.
Overall, the distribution of postverbal focused objects relative to postverbal subjects is inconsistent with the presence of an intermediate focus projection, whereas it is accounted for if they focus in situ.
3.3.2 Experiencer objects and infinitival complements
The focus pattern just examined in the previous section is also found with ditransitive verbs selecting for an experiencer object and an infinitival complement, showing that the absence of focus movement generalizes to different types of constituents and grammatical functions.
As discussed in Chapter 2, non-finite clausal complements are generated below experiencer objects. The same order is found when the clausal complements are focused, showing that they focus in situ rather than raising to an intermediate focus projection. This is shown in the data below, involving a negative object to control for right dislocation. Note that the heaviness of the marginalized object is not a factor: Example (25) is assessed like the other two even though it involves a heavier object. (p.54)
3.3.3 Postverbal subjects and infinitival complements
In-situ focalization is also supported by sentences involving postverbal subjects and infinitival sentential complements. As (27) shows, focused infinitival complements cannot raise above a discourse-given postverbal subject, as expected if focused in situ and unexpectedly under the intermediate focus projection hypothesis.
(p.55) At first, restructuring verbs like volere ‘to wish’ appear exceptional in that their infinitival complement must precede rather than follow a marginalized subject even when focused, see (28).
As explained in Cinque (2004), however, restructuring verbs share the properties of auxiliaries. As such, they are generated in T and form a single clause with the following lexical verb.6 Sentences involving restructuring verbs are thus expected to show the same focalization patterns of mono-clausal sentences involving an auxiliary and a past participle. This is indeed the case: compare (28) with (29). In both cases, the lexical verb moves to a higher aspectual projection stranding the marginalized subject behind. Cinque’s analysis thus explains—and in fact finds further support in—the divergent focalization patterns of restructuring and non-restructuring verbs in (27) and (28).7
(p.56) Cinque’s analysis of restructuring verbs is also essential to the understanding of the pattern displayed by focused subjects in restructuring constructions involving a discourse-given verb and object. Focalization in situ leaves the subject after the raised lexical verb and before the marginalized object as in (30)(a), while moving the subject above the raised verb as in (30)(b) is ungrammatical.
Summing up, the distribution of focus and marginalization in clauses involving infinitival complements confirms the absence of focus movement. With non-restructuring verbs, focused sentential complements cannot raise above discourse-given subjects. With restructuring verbs, focused subjects may not raise above the aspectual projection targeted by the lexical predicate. Both observations follow if focalization occurs in situ.
3.3.4 Floating quantifiers
The previous sections showed that once we control for right dislocation, focused constituents do not raise above higher-generated constituents. What these tests could not ascertain is the presence or absence of movement that is sufficiently local to not alter the base-generated word order of the constituents involved (thanks to Chris Collins for raising this issue). We can test for this kind of local movement by examining the distribution of floating quantifiers. This test, too, confirms that movement is absent and, therefore, that focalization occurs in situ.
Floating quantifiers can be stranded in situ, thus making the original position of a raising DP visible. For example, under a sentence-wide new-information focus the rightmost possible position for the subject-related floating quantifier tutti ‘all’ in a ditransitive clause necessarily precedes the object and the indirect object, consistent with its base-generated specVP position; compare the grammatical (31)(a) where the floating quantifier precedes the object and indirect objects, with (31)(b) and (31)(c) which are unacceptable under the provided context. (Sentence (31)(c) is acceptable when the quantifier is focused on its own, i.e. under a different context. (p.57) This is expected, as explained later in Section 3.4, but irrelevant here, where the goal is to establish the initial position of the subject when focus is not a factor.)
We may now test for the presence of local movement by examining whether a focused subject may move immediately to the left of its quantifier under an appropriate context focusing the quantified DP while leaving the quantifier discourse-given. Such context is provided in (32) and allows for the grammatical answer in (32)(a). As (32)(b) shows, however, local movement of the quantified DP over the quantifier is unavailable, confirming that focalization occurs in situ. The structure of (32)(b) is provided in (32)(c). The two sentences in (33) apply the same test to focused objects with identical results. In all examples the final constituent is negative to control for right dislocation.
Note that there is no reason to suspect that some unknown factor specific to these simple sentences blocks focus movement or disallows quantifier stranding, because focus (p.58) movement and quantifier stranding is possible when the focused phrase occurs left-peripherally as in example (34). (This construction also involves right dislocation of the clause following focus, hence the ‘R’ subscript in the example; see Chapter 5 for discussion.)
In conclusion, the distribution of contrastive focus in postverbal position is inconsistent with the presence of an intermediate focus projection above VP. Under this hypothesis focus should precede higher-generated phrases marginalized in situ as well as discourse-given quantifiers stranded in situ, but the correspondent sentences are ungrammatical. The observed distribution instead follows straightforwardly if focalization occurs in situ, since this predicts that focused phrases will inevitably follow both higher generated discourse-given phrases as well as their own quantifiers.
3.4 Rightmost focus
Besides preceding lower marginalized constituents, postverbal focused phrases may also optionally follow them. For example, a postverbal focused subject may precede a marginalized object as in (35)(a), but also follow it when the object raises above the subject as in (35)(b). When no other constituent follows the focused constituent, this movement leaves focus aligned with the right edge of TP producing the rightmost focus pattern examined in many studies of Italian and similar Romance languages (e.g. Antinucci and Cinque 1977; Calabrese 1982, 1986, 1992; Bonet 1990; Vallduví 1992; Saccon 1993; Belletti and Shlonsky 1995; Zubizarreta 1994b, 1998; Samek-Lodovici 1996, 2005; Belletti 2001; Szendröi 2001, 2002).
In this section, I will show that even rightmost focus involves focalization in situ. Furthermore, I will show that the raising of lower-generated phrases (p.59) responsible for rightmost focus is induced by focalization, since it is unavailable when focus is absent or when higher and lower constituents are both part of a larger focus. This result will be eventually used in Chapter 6 to argue for a prosody-induced analysis of rightmost focus along the lines pursued by Zubizarreta (1994b, 1998), Szendröi (2000, 2001), and Samek-Lodovici (2005), where the position of focus is determined by the constraints governing the position of the associated stress. The results in this section will also be shown to go against the hypothesis of an intermediate focus projection à la Cecchetto (1999) and Belletti (2004).
3.4.1 Discourse-given phrases raising above higher foci
Let me start with the empirical evidence for the existence of rightmost focus. The data below examine the possible orders of two postverbal items A and B, with A focused and generated above B. The order <AF BM> is always possible, consistent with the earlier sections showing that constituents might focalize in situ followed by lower marginalized constituents. More interestingly, the order <B AF> is also possible. Since the same order is impossible when A and B are both marginalized, the optional raising of B must be dependent on A’s focalization.
As in the previous section, I start by considering the position of subjects and objects, then consider experiencer objects and sentential complements, then subjects and sentential complements, and finally Cinque’s lower adverbs.
Postverbal subjects and objects—Postverbal focused subjects may precede lower objects, as in all (a) sentences below, but they also naturally occur in clause rightmost position, as shown by the (b) sentences. This latter order emerges when the object raises above the focused subject.8 In all examples, the object is negative to ensure it (p.60) cannot be right-dislocated. The examples use subjects and objects of different size to control for potential heaviness effects.
Experiencer objects and infinitival complements—Experiencer objects can be focused in situ followed by a lower discourse-given CP complement, as in (39)(a), but it is also possible for the complement to raise above the focused objects and strand them in rightmost position, as in (39)(b).
(p.61) Since in these constructions n-word licensing from the matrix clause into the infinitival CP-complement is not possible, there is no way to control for its right-dislocated status when it occurs in postfocal position. The infinitival complement in (39)(a) could thus be either marginalized or right-dislocated, hence the ‘M/R’ subscript. What really matters here, though, is the availability of rightmost focus in (39)(b), where right dislocation is not a factor.
Postverbal subjects and infinitival complements—The higher generated focused subjects may precede but also follow the lower infinitival complement, providing further evidence for the occurrence of rightmost focus.
3.4.2 The role of focalization
In all above examples, the lower discourse-given phrase raises above a higher constituent focalized in situ. When focalization is absent, the same movement operation can no longer occur, showing that it is induced by focalization.
A first piece of evidence in this respect comes from the properties of marginalized constituents. As we saw in Chapter 2, their order is fixed and reflects their base-generated position. Raising of the kind under discussion is thus absent when the higher and lower constituents are both unfocused.
Even when the higher constituent is focused, movement of the lower phrase is only licensed if it targets a position preceding focus, thus pushing the focused phrase closer to the right edge of the clause. Compare (41) and (42). When the subject is focused, the lower object can follow it, as in (41)(a), or raise above it, as in (41)(b), whose structure is given in (41)(c). When focus is shifted to the verbal past-participle and the subject is no longer focused, however, the object must follow the subject as in (p.62) (42)(a), and movement above the subject in (42)(b) is no longer possible. The structure of (42)(b) is provided in (42)(c).
Similarly, when focus encompasses all constituents, as is the case under sentence-wide focus, the movement here under study is absent. Since full subjects do not remain in situ when the entire clause is focused, we have to examine the movement of objects relative to quantifiers stranded in specVP by the raising subjects. As the contrast between (43)(a) and (43)(b) shows, objects cannot raise above the stranded quantifiers in focused clauses. Raising, however, becomes possible again if only the quantifier is focused and the object is discourse-given, see (44)(a) and (44)(b).
(p.63) An appropriate analysis of Italian focus must explain the peculiar properties of the movement operation just examined. Why is it restricted to discourse-given phrases and why must it involve raising above a higher focused constituent?
In Chapter 6, I will argue that the observed movement improves the alignment of main stress relative to the right edge of the clause, as required by the constraints governing the Italian stress system. When the constituents involved are either both discourse-given or both focused the related stress configuration no longer obtains making movement ungrammatical.
3.4.3 Problems affecting the intermediate focus projection analysis
Belletti’s intermediate focus projection hypothesis can model the raising of lower discourse-given phrases above focus but it cannot capture its dependency on the presence of focus. The raising discourse-given constituents would have to move to a topic projection immediately above the intermediate focus projection, arguably to check a topic feature. This dissociates the observed movement from focalization. But the same operation is then predicted to be possible even when focalization is absent, which is incorrect. This is further illustrated by the examples (45) and (46). In (45) the entire lower clause constitutes a discourse-given sentential complement and the position of subject and object within this complement cannot be altered, yet this should be possible if the object could raise to a higher TopicP projection. Likewise, the object ought to be able to raise above the subject in (46), where focus occurs left-peripherally, yet this is not the case.
(p.64) Finally, note that the incorrect prediction just examined is independent of the specific assumptions governing movement to the higher TopicP projection. Whether discourse-given phrases move to specTopicP on their own or as a larger remnant constituent as proposed in Belletti (2004) for presentational focus,9 the operation is incorrectly predicted to be possible even when neither constituent is focused.10
This last section considers two further pieces of evidence for the claims made so far. Section 3.5.1 examines lower adverbs, showing that they too disallow focus raising and allow for rightmost focus like all constructions considered so far. Section 3.5.2 examines the binding properties of focused and discourse-given phrases showing that focused constituents bind into lower-generated phrases even when the latter have raised above focus, thus revealing the A'‑nature of the involved movement.
3.5.1 Evidence from lower adverbs
The adverbs located between raised active past participles and VP examined in Cinque (1999) provide an independent testing ground for the presence of in-situ focalization and the absence of focus raising to intermediate positions. Cinque argues that the order of lower adverbs visible in sentences like (47) is base-generated. Indeed, when the sentence is interpreted under presentational focus the order cannot be altered. As we saw in Chapter 2, the same fixed order is found when the adverbs are marginalized in situ.
We want to consider what pattern arises when an adverb gets contrastively focused. If focalization occurs in situ, we expect that for any two adverbs Adv1 and Adv2, with Adv1 higher than Adv2, a focused Adv2 may not raise above a discourse-given Adv1, as schematized in (48). Positing an intermediate focus projection above lower adverbs—an even lower position would leave some adverbs unable to focus—makes the opposite prediction since the lower adverb must raise above the higher marginalized one in order to focus.
The available data support in situ focalization. The context sentence in (49) shows the adverbs in their base-generated order and implies that John’s understanding of his pupils’ speech is getting worse. The replies in (49)(a)–(d) vary only in the position of the lowest adverb male (badly), which is focused in contrast with the adverb bene (well/properly) in the context sentence. The replies imply that John is beginning to understand his pupils better. Crucially the focused adverb male may occur in situ, as in (49)(a), but it cannot raise above any higher adverb, as shown in (49)(b)–(d).11 (p.66)
Low adverbs also provide independent evidence for the optional availability of movement of lower discourse-given constituents above higher foci, a fact also noticed in Cinque (1999: 14) who observes that focus may alter the adverbial hierarchy by optionally placing focused adverbs in clause rightmost position. For example, the discourse context in (49), also allows for reply (50), where the discourse-given object nessun allievo (any pupil) moves above the adverb male (badly).
To wrap up, the interaction of focus and marginalization with respect to Cinque’s lower adverbs shows the same properties found with other constituents: impossibility of focus raising, in-situ focalization, and optional raising of lower discourse-given phrases above focus.
The binding relations between focused and discourse-given phrases in postverbal position confirm the results found so far. Postverbal subjects, whether focused or not, always bind lower marginalized objects, as expected if both occur in situ. Discourse-given objects raised above focused subjects, instead, fail to bind them and are themselves bound by the subjects under reconstruction, showing that they raise to an A'-position.
There are three variables at play in the following binding data: the direction of binding between subject and object, their linear order, and which of them is focused. This yields a total of eight cases which are systematically considered below. Whenever the final item is neither focused nor negative, we also have to consider whether that item admits a right-dislocation analysis which might affect the assessment.
<S OF> cases—Let us start with the cases testing binding between a discourse-given subject and a following focused object. In this case the subject binds the object but not vice versa, as shown in schematic form in (51) where the arrow signals the direction of binding.
A first piece of evidence is provided in (52) where the subject binds the following focused object but not vice-versa. To ensure an accurate judgement, subject and object differ in number, so that agreement always identifies the intended subject.
The reported judgements are only valid relative to the intended bound interpretation, and the same applies to any other example in this section. When the pronoun is interpreted as bound by an independent referent (52)(b) becomes grammatical, but this reading is irrelevant here. The reported judgements are also insensitive to phrasal size, remaining invariant even when nessuno ‘nobody, anybody’ is replaced by nessun ragazzo ‘no boy, any boy’. To enhance readability, the suffixes ‘M’, ‘F’, and ‘R’ appear on the last item of the entire subject or object, with no explicit marking of the boundaries of the affected constituents.
(p.68) Anaphoric binding yields the same pattern, see (53) (see also Cardinaletti 2001). The subject of (53)(a) is indefinite, since definite subjects are independently known to be ungrammatical under the <VSO> order when neither focused nor marginalized (Benincà 1988: 124).
<SF OM> cases—If we keep the <S O> order but focus the subject rather than the object, the binding relations remain unaltered, with focused subjects binding the marginalized objects but not vice versa, exactly as expected if both occur in situ. The corresponding data for quantifier binding are given in (55).
Anaphoric binding yields the same pattern. Note that since objects may right dislocate without clitic doubling (see Chapter 4), the object in (56)(a) is ambiguous between a marginalized and a right dislocation analysis. As we will see in Chapter 4, binding into right-dislocated phrases is possible but it involves reconstruction of the right-dislocated item. Sentence (56)(b), too, is ambiguous between a right-dislocated and a marginalized analysis of the object but its ungrammatical status confirms that an object cannot bind a higher focused subject whether marginalized or right-dislocated.
(p.69) <O SF> cases—As (57) shows, raised discourse-given objects cannot bind a following focused subject, whereas focused subjects may bind raised objects, confirming that the latter raise to an A'-position and may reconstruct in their original position.
The corresponding data for quantifier binding are provided in (58). Similar judgements emerge when replacing ‘nessuno’ with the heavier ‘nessun ragazzo’.
<OF S> cases—The final case, involving binding relations between raised focused objects and a following subject as shown in (60), is untestable because as we saw in Section 3.3 focused objects may not raise above discourse-given subjects. (p.70) The corresponding data for quantifier binding in (61) confirm the impossibility of focus raising.13
The corresponding data for anaphoric binding allow for the grammatical sentence in (62)(b) because the final subject allows for a right-dislocation analysis on a par with (63), where a right-dislocated subject follows an equally right-dislocated clitic-doubled indirect object. As discussed at length in Chapter 4, the subject dislocates to a clause external position, thus allowing the focused object to remain in situ. The sentence is grammatical because the reconstructed subject binds the object. For the same reasons, in (62)(a) the reconstructed copy of the subject is not c-commanded by the in-situ focused object, therefore binding fails.
In summary, the attested binding relations between postverbal subjects and objects follow from the analysis developed in the precedent sections. Constituents focused and marginalized in postverbal position occur in situ. Therefore, focused subjects bind marginalized objects but not vice versa. Furthermore, discourse-given objects (p.71) may A'-raise above focused subjects but they cannot bind them. Rather, they are bound by them under reconstruction.
126.96.36.199 Divergent binding relations with the universal quantifier ‘ogni’
The above discussion also shows the importance of teasing apart marginalization and right dislocation lest data that are ungrammatical under one construction are incorrectly interpreted to be grammatical because they are grammatical under the opposite construction, determining erroneous structural conclusions.
The need to distinguish between the two constructions is well illustrated by the binding data in Cardinaletti (2001), where discourse-given objects are reported to bind a focused subject that precedes or follows it, against the results discussed in the previous section. The relevant data from Cardinaletti (2001: 122, 129) are listed in (64) and (65), showing Cardinaletti’s original assessment. The absence of agreement disambiguation makes it difficult to easily determine which constituent is the subject and which the object, potentially affecting their grammaticality assessment. However, Cardinaletti’s judgements are also supported by the corresponding unambiguous sentences in (66) and (67) and should therefore be considered valid.
The divergence between Cardinaletti’s data and the binding relations examined in the previous section follows once we understand two important properties that set the universal quantifier ogni apart from the negative quantifiers examined in the previous section.
First, as noted in Szabolcsi (2001), Beghelli (1993), and Beghelli and Stowell (1997), universal quantifiers may attain higher scope than negative quantifiers. This property underlies the distinct distribution of universal and negative quantifiers in Italian, with universal quantifiers allowed to precede and take scope over preverbal subjects, see (68), while negative quantifiers are unable to do so, see (69). (p.72)
Second, contrary to Cardinaletti’s assumptions, DPs quantified by ogni can right-dislocate and may do so with or without clitic doubling. This is shown by (70) and (71). In sentence (70) the right-dislocated status of the quantified object is confirmed by the presence of clitic doubling. In sentence (71) the quantified subject follows a clitic-doubled object and hence it too is necessarily right-dislocated. It follows, that (64) and (66) allow for a right-dislocation analysis of the quantified object. This result contrasts sharply with the properties of negative quantifier nessuno, which always resists right dislocation, as shown by the corresponding ungrammatical sentences in (72) and (73). These sentences are ungrammatical because right-dislocated phrases occur clause-externally, thus placing the negative phrase outside the licensing domain of the initial neg-marker non ‘not’ (see Chapter 4 for discussion). The same problem does not affect the phrases involving ogni in (70) and (71), since they are not subject to NPI-licensing.
Together, these two properties explain Cardinaletti’s data. In (65) and (67), the object has moved to the higher position exceptionally available to universally quantified phrases and binds the subject from that position. The same position is not available to negatively quantified phrases, explaining why the negative objects examined in the previous section fail to bind a following subject. As for (64) and (66), the universally quantified object is right-dislocated clause-externally and as such allowed (p.73) to reconstruct into the pre-subject position just discussed, from which binding is possible. The same is not possible with negative quantifiers because they cannot raise to the same pre-subject position and cannot be right-dislocated without failing NPI-licensing.
Once we control for the absence of right dislocation and examine the interaction between contrastive focus and marginalized phrases, two important structural results emerge. First, postverbal phrases focalize in situ. Second, lower discourse-given phrases may raise above a higher-generated phrase, but only when the latter is focused. Both results are stated in (74) and (75), followed by a table listing the main grammatical and ungrammatical sequences of focused and marginalized postverbal constituents discussed in this chapter and supporting these results.
These results show that focalization in postverbal position may only trigger movement operations that remove discourse-given constituents located to the right of focus, thus reducing the number of constituents separating focus from the clause right edge. Any movement that increases or leaves unaltered the number of intervening constituents is instead ungrammatical, witness the impossibility of raising focused phrases or reshuffling postfocal discourse-given phrases amongst themselves.
This behaviour is exactly what is expected under an analysis where the distribution of focus is affected by prosodic constraints, as proposed in Zubizarreta (1998), (p.74) Szendröi (2001), Samek-Lodovici (2005). Focus requires prosodic prominence, but the constraints governing prosodic prominence in Italian favour a clause-rightmost position. Any movement operation that improves the alignment between focus and the clause right edge is thus welcome, whereas any operation that worsens it is ungrammatical. The corresponding formal analysis is provided in Chapter 6.
The distribution of focus just examined also shows that there are no dedicated positions for focalized constituents in the clause and hence no corresponding projections either. Postverbal constituents focus in situ, hence in different positions. As discussed in this chapter, any attempt to analyse these focused constituents as sharing the same position runs into severe problems. In particular, locating them in a fixed higher left-peripheral position à la Rizzi is inconsistent with the properties of postverbal focus relative to the licensing of n-words and NPIs, as well as the distribution of parasitic gaps and interrogative operators relative to contrastive focus. Positing an intermediate focus projection à la Cecchetto (1999) and Belletti (2004) is in turn inconsistent with the observed absence of leftward movement of postverbal focused phrases. In conclusion, the study of the interaction of focus and marginalization shows that a templatic analysis of focus based on unique fixed projections is not viable because postverbal constituents are focused in situ.
We may still wonder whether a fixed focus projection is necessary for left-peripheral focus, where focused phrases of different origin might at first appear to share a unique position above TP. Left-peripheral focus will be addressed in Chapter 5, where we will see that the properties of fronted contrastive foci, too, are inconsistent with the presence of a fixed left-peripheral focus projection. Rather, their distribution is determined by right dislocation, which affects the constituents containing them. To see this, I will first have to discuss the properties of right dislocation in detail, which I do in Chapter 4.
(1) Since licensing under reconstruction is not available, the only conceivable structure potentially able to license postverbal negative focused phrases under a left-peripheral analysis requires the licensing neg-marker to occur higher than left-peripheral focus, with overt subjects placed even higher, as shown in (ii) which provides the structure for (i).
Even this hypothetical solution, however, is untenable. Consider sentence (iii)a where the neg-marker and the focused postverbal object occur in a non-finite subordinate clause introduced by the complementizer di. Rizzi (1997: 288) shows that this complementizer lies in FinP, which is located lower than left-peripheral focus. Consequently, a left-peripheral analysis of (iii)a would have to maintain that di is part of the remnant constituent raised above FocusP, as shown in structure (iii)b, or else it would follow the focused object. Since the neg-marker non necessarily follows di—see the ungrammatical inverted order in (iii)c—it too must be part of the remnant phrase (here a FinP rather than a TP). But this contradicts the above hypothesis, where the neg-marker c‑commands Rizzi’s CP-level focus projection.
(2) Zubizarreta (2010) and Belletti (2004) examined this issue, too, reaching different but ultimately non-tenable conclusions. Zubizarreta (2010: 156, fn12) proposes that the licensing neg-marker might be located higher than the fronted negative phrase, hence outside the remnant TP and c-commanding the relevant constituents. The empirical evidence against this hypothesis is discussed in the previous footnote. Furthermore, this hypothesis still does not explain why licensing is needed at all, since licensing is not necessary with preverbal focused phrases.
Belletti (2004: 37) accepts that a preceding neg-marker cannot license a postverbal NPI focused in a left-peripheral position but considers this a correct prediction. Belletti’s original contrastive focus data and judgements follow (see example (51) in Belletti 2004: 37).
The context sentence for (i) is highly unnatural, which in turn makes the answer in (i)a equally unnatural, making a reliable grammaticality judgement difficult. Sentences similar to (i)a, however, are acceptable once an appropriate context is provided, see (iii) and (iv) which have been checked with several native speakers.
The ungrammaticality of (ii)a is only perceived by some native speakers of Italian and it is almost certainly due to a constraint shared by those speakers against the presence of direct objects before a rightmost focus. Speakers insensitive to such constraint, like myself, find (ii)a grammatical. More importantly, sentences not involving direct objects, like (v), are grammatical across both sets of speakers and require the focused subject NESSUNO to be licensed by the negative marker non, against Belletti’s analysis. This sentence, too, has been checked with several speakers.
This leaves only (ii)b, which involves an infrequent and structurally complex NPI of the ‘che NP’ form. These NPIs deserve further research, but the fact that simpler and more frequently used negative phrases and polarity items are grammatical when contrastively focused in postverbal position shows that licensing by a preceding neg-marker is possible, against what is predicted under a left-peripheral analysis.
(3) Under a left-peripheral analysis of postverbal focalization, reconstruction of the initial remnant TP would have to be considered possible, since a subject within the remnant TP can still bind a focused anaphoric object as in (i), and a null subject cannot bind a referential object as in (ii). The structure for (i) is given in (iii). The availability of reconstruction for the remnant TP and the fronted focus provides a potential explanation for why under this analysis the fronted foci in (15) can license parasitic gaps even though they do not c-command their own trace. It follows that the same lack of c-command between the postverbal foci in (16) and their traces cannot be considered the cause of their ungrammaticality.
(4) Brunetti (2004) is mostly concerned with establishing the equivalence between contrastive and presentational foci, showing that they share a similar syntax. The analysis proposed in this book shares with her analysis the claim that postverbal foci occur in situ but it differs in the analyses of left-peripheral foci, which raise to a higher position c-commanding the rest of the clause in Brunetti (2004) whereas in my analysis they are followed by a right-dislocated clause that they do not c-command, as explained in Section 5.3.
(5) Prosodic size does affect the distribution of focused pronominal objects. When subject and object are not heavy, pronominal objects show the same distribution of other constituents and cannot precede marginalized subjects when focused, see (i). They might however marginally precede subjects when the subject is heavier, as in (ii). The higher position of focused pronominal objects in (ii) might follow from the independently attested structural and categorical properties of pronominal items, which are known to differ from those of non-pronominal DPs; see Cardinaletti and Starke (1999) and Déchaine and Wiltschko (2002) for discussion.
(6) As Cinque points out, this analysis explains the several properties that distinguish restructuring verbs from non-restructuring ones. For example, Cinque’s analysis accounts for why clitic climbing is allowed with restructuring verbs, see (i), but not with non-restructuring verbs, see (ii) (object clitic in bold). As in any other simple clause, the object clitic in (i) precedes the item in T, which here happens to be the restructuring verb voleva ‘wished’. Sentence (ii), instead, involves two distinct clauses and therefore clitic-climbing is blocked.
(7) The past participle of (29)(b) is focused in the aspectual projection above VP and hence not in situ. Crucially, however, it occurs in this projection independently of its focused status and movement into this position is not related to focalization.
(8) Is rightmost focus necessarily caused by movement of the lower phrase (Grimshaw p.c.)? Analyses of syntactic structures based on the flexible base-generation hypothesis proposed in Abels and Neeleman (2006), Ackema and Neeleman (2002), and Neeleman and Weerman (2001) allow for the generation of rightward higher constituents that could capture rightmost focalization with no resort to movement. Higher focused constituent could focus rightmost when generated to the right of a lower item. For example, the pattern <O SF> could be base-generated as in (i), where specVP is generated rightward.
This analysis, however, makes incorrect predictions with respect to in situ marginalization. For example, if focused subjects were generated rightward, we would expect unfocused subjects to be able to do so too and appear marginalized to the right of a focused object, as in structure (ii). But this produces the ungrammatical <OF SM> order.
This problem could in principle be fixed through an optimality-theoretic analysis where the attested precedence relations emerge from the interaction of grammar constraints. In Section 6.4.2, I’ll briefly consider the general structure of such an analysis.
(9) The argument against the existence of a lower focus projection extends to new information focus as well. This is shown in (i) and (ii), where the constituent corresponding to the wh-phrase in the context question is presentationally focused and yet we find the exact same patterns found with contrastive focalization. This evidence supports the arguments in Brunetti (2004) who claimed that contrastive and new information focus are non-distinct as far as Italian syntax is concerned. It also calls into question the existence of a lower projection even for new information focus. (In the examples new information focus is marked by the subscript ‘NewF’.)
(10) More specifically, Belletti (2004: 37) proposes that a sentence like (i) is represented as in (ii) (where ‘IP’ has been replaced with ‘TP’). When the same analysis is applied to a sentence without auxiliary and with left-peripheral focalization as the example in the main text, repeated in (iii), the resulting structure is (iv). This structure incorrectly predicts the sentence to be grammatical, since the subject may move to the lower topic projection and the VP-remnant containing the object to the higher topic projection.
(11) Anna Szabolcsi (p.c.) notes that the original analysis of adverbial movement in Cinque (1999) involves movement of the entire phrase containing the adverb and its complement, rather than movement of individual adverbs, a point I will return to in Section 6.6. Crucially, raising of focalized adverbs is ungrammatical even under Cinque’s analysis of movement. For example, in (i)a the focused adverb male ‘wrongly’ raises while pied-piping the lower object as required, but the sentence remains ungrammatical.
(12) Cardinaletti (2001: 129) claims that discourse-given objects may bind a following focused subject on the basis of example (1) below, where Gianni is the object. I personally find (i) ungrammatical (hence the star in parentheses). The difference in judgement could be due to the absence of agreement disambiguation between subject and object, which makes (i) easily susceptible to the irrelevant grammatical reading where Gianni is the subject of the sentence. Under this reading, (i) is grammatical because postverbal subjects may bind following focused objects.
In a footnote, Cardinaletti also notices that the following sentence—which provides a more reliable diagnostics because agreement properly disambiguates between subject and object—is ungrammatical, in accord with the results reported here (Cardinaletti 2001: 129, footnote 13).
(13) Interestingly, (61)a is worse than (61)b, even though both are clearly ungrammatical. As mentioned, both sentences are ungrammatical because they raise a focused constituent. In addition, however, binding in (61)a is impossible even under reconstruction, since the reconstructed object does not c‑command the subject. In (61)b, on the other hand, the reconstructed object is c‑commanded by the subject, thus potentially allowing for binding and possibly determining the difference in assessment.