Abstract and Keywords
This chapter examines Italian marginalization. The name ‘marginalization’ was originally used by Antinucci and Cinque (1977) as a general label for any discourse given constituent found to the right of focus, including right-dislocated ones. After Cardinaletti’s research in 2000 and 2001, this label has been restricted to post-focus discourse-given constituents that occur unstressed in situ, making marginalization similar to the in-situ destressing operation found with English discourse-given phrases. The first part of the chapter establishes the basic word order of Italian, followed by a discussion of marginalization based on Cardinaletti’s work as well as new evidence ranging from the distribution of Italian negative phrases to the analysis of Cinque’s low adverbs. The criteria that distinguish marginalized phrases from right-dislocated ones are particularly important for an effective analysis of contrastive focalization; confusing one with the other is very easy and frequently leads to incorrect conclusions.
The term ‘marginalization’ distinguishes discourse-given phrases that are destressed in situ, examined in this chapter, from discourse-given phrases that are right-dislocated above TP, examined in Chapter 4.1 It is hard to overemphasize the importance of this distinction for a proper understanding of the syntax of Italian and especially for the analysis of Italian focus. The syntactic properties of focus, including its position, must often be inferred from the properties of the constituents surrounding it. Since Italian focused constituents can be followed by both marginalized and right-dislocated phrases, as explained in Chapters 3 and 4, an accurate identification of these phrases as marginalized or right-dislocated is necessary to avoid invalid conclusions. Yet, few studies examine this aspect, most works usually incorrectly assume that the constituents following focus occur in situ. Similarly, some studies of right dislocation, too, fail to distinguish genuinely right-dislocated phrases from marginalized phrases, assigning the properties of one type of phrase to the other.
(p.26) As I will show later in this book, once we control for the marginalized vs. right-dislocated status of post-focus constituents, a series of important results follows, including the absence of overt focus movement when right dislocation is absent (Chapter 3), the existence of right dislocation without clitic doubling (Chapter 4), the observation that Italian left-peripheral focus involves a wider distribution than the one originally studied in Rizzi (1997), and the realization that left-peripheral foci of all types are triggered by right dislocation when it targets a constituent containing focus (Chapter 5).
Marginalized phrases are characterized by the four properties listed in (1) (Cardinaletti 2001, 2002; Calabrese 1982, 1992; Antinucci and Cinque 1977). Properties (c) and (d), follow from property (b): since marginalized phrases occur in situ, as per property (b), they are inevitably ordered along their base-generation order and disallow clitic doubling, as stated in (c) and (d), consistently with the observation that clitic doubling in Italian only occurs with constituents located outside TP.
This chapter presents and further extends the empirical evidence supporting the above properties, examining the distribution of marginalized postverbal arguments and lower adverbs, as well as the related agreement and binding phenomena. Most of the data presented in this chapter also provide the discourse context in which they should be assessed. The context determines which constituents are focused and which marginalized because discourse-given. Native speakers should always read the context before assessing the data, as the latter are ungrammatical whenever the context is not taken into account (and so they should be, since ignoring the context corresponds to making an assessment under a clause-wide presentational focus where neither marginalization nor contrastive focus are licensed).2
2.2 Italian basic word order
To examine how marginalization, right dislocation, and contrastive focus affect clause structure we first need to establish how the clause is structured when they are absent. To do so, we examine clauses that are entirely presentationally (p.27) focused—i.e. constituting new information—because elicited as answers to general questions like Any news? or Why are you angry/happy? or out-of-the-blue declaratives. In all these cases, the discourse context presupposes nothing but general world knowledge. This ensures that marginalization, right dislocation, and contrastive focus are absent because no constituent is discourse-given, as required by marginalization and right dislocation, or interpretable as part of a presupposed or context-induced set of alternatives as required for contrastive focus. The word order found in these presentational clauses instantiates what is usually referred to as the basic, canonical, unmarked word order of the language. I will refer to this order when examining the distribution of marginalization, right dislocation, and focus.
For Italian declarative clauses, this basic word order is <S V O IO> with main stress falling rightmost in the clause (Antinucci and Cinque 1977; Renzi 1988: 120–1; Belletti and Shlonsky 1995).3 Note that subjects are located in preverbal position, in contrast with languages like Spanish, Portuguese, and Greek where postverbal subjects in specVP are also possible (Ordoñez 1997; Zubizarreta 1998; Costa 1988; Alexiadou and Agnanostopoulou 1999). For example, verbs like dormire ‘to sleep’, sorridere ‘to smile’, camminare ‘to walk’, lavorare ‘to work’, vedere ‘to see’, and sentire ‘to hear’ disallow a postverbal subject under presentational focus, see example (2) where Marco is the only child of the couple engaged in the conversation.4 Yet they should allow for it if Italian subjects could remain in specVP, because the verb moves to a higher head hosting the past-participle suffix. Answer A2 can only be made grammatical under a context focusing on the subject, for example as an answer to the question Who slept?
The basic word order for internal arguments is harder to pin down. For example, in (3) answers A1 and A2 show opposite orders for the object and indirect object but they are both grammatical answers to the context question. (p.28)
More refined tests, however, show that direct objects precede other internal arguments. This underlying word order is for example revealed under wh‑extraction in ditransitive clauses involving sentential complements. As shown in (4), wh‑extraction from within the sentential complement is only possible when the complement follows the direct object (Calabrese 1982; Cardinaletti 2002). The contrast follows if the sentential complement lies in situ in (4)(a) but not in (4)(b), where its raising to a specifier position above the object adversely affects wh-extraction.
Further evidence for the proposed order follows from the possible interpretations associated with the adverb solo ‘only’ (Renzi 1988). Assume that John fancies Mary and consider the context question in (5). From a purely structural perspective, sentence A1 with its canonical <V O IO> order allows for two interpretations: a first one where the adverb modifies the entire VP, as in (6)(a), and meaning that the only action that Mark did was giving flowers to Mary, plus a second one which is infelicitous under the given context where the adverb focuses the object as in (6)(b) and meaning that Mark gave Mary only flowers.
Now consider answer A2, where the order of object and indirect object is switched. It only allows for the infelicitous interpretation where the adverb modifies the indirect object alone, as in (7)(a). Crucially, it does not allow for the felicitous interpretation where the adverb modifies the entire VP. Yet, if indirect objects could be base-generated before objects, this interpretation should be available, as shown in (7)(b). The ungrammaticality of A2 shows that indirect objects are base-generated lower than objects. The interpretation focusing the entire VP is unavailable because the structure of A2 requires movement of the indirect object within the focused VP as shown in (7)(c) and since this movement serves no purpose and movement is costly, the structure is ungrammatical (the formal analysis of this intuition is provided in Section 6.3.3). (p.29)
Finally, and importantly for the analysis of marginalization and right dislocation to follow, note that presentationally focused constituents cannot be clitic-doubled (Calabrese 1988: 557; Kuchenbrandt, Kupisch, and Rinke 2005; Gerlach 1998: 27).5 For example, the following sentences with clitic-doubled postverbal internal arguments are strongly ungrammatical under clause-wide presentational focus (here ensured by the presence of clause-final stress, which is resisted by marginalized and right-dislocated phrases).
We may conclude that Italian has the <S V O IO> basic word order. With this knowledge in hand, we may proceed and examine the distribution and properties of marginalized phrases.
2.3 In situ marginalization
Teasing apart marginalized phrases from right-dislocated ones requires some care because their most eye-catching property—namely their occurrence in right-peripheral position past the item carrying main stress—is common to both. For example, the (p.30) postverbal unstressed subject in (9) could be either marginalized or right-dislocated, as symbolized by the subscripts ‘M’ and ‘R’.
We can nevertheless investigate the properties of marginalization by examining items that allow for marginalization but resist right dislocation, such as negative phrases like nessuno ‘nobody/anybody’ and negative polarity items (NPIs) like alcunché ‘anything’. As explained in detail in appendix A, when negative phrases occur in postverbal position and are c-commanded by T, they must be obligatorily licensed by a c‑commanding licenser located in or above T such as, for example, the neg-marker non in (10) (Zanuttini 1991; Longobardi 1991; Acquaviva 1999; Penka 2011). This requirement prevents negative phrases from being right-dislocated because right dislocation would place them above TP and therefore outside their licensing domain. Compare the grammatical sentences in (11), showing a non-negative right-dislocated object and indirect object, with the ungrammatical sentences in (12) where the right-dislocated phrases are negative. Similar examples can also be built with non-negative NPIs such as alcunché ‘anything’ and alcun ‘any’.
In the above examples the right-dislocated constituents are clitic-doubled, but negative phrases resist right dislocation even when clitic doubling is absent. As (13) shows, non clitic-doubled right-dislocated negative phrases are also ungrammatical. The negative phrases here follow a clitic-doubled right-dislocated argument to ensure (p.31) that they are themselves right-dislocated. Note how these sentences become grammatical again in (14) when the negative phrase is replaced by non-negative expressions and licensing is no longer an issue.
Having established that negative phrases cannot be right-dislocated, we may use them to examine the properties of marginalization. Like right-dislocated phrases, marginalized phrases always follow the item carrying main stress and never carry stress themselves. Unlike right-dislocated phrases, however, marginalized phrases occur in situ and are not clitic-doubled (Cardinaletti 2001, 2002). It follows that negative phrases should be possible when marginalized, since they remain in the licensing domain of their licenser, even if they cannot be right-dislocated. This is indeed true; see for example the alternation in (15). Sentence (a) is grammatical because the subject nessuno is marginalized in situ, thus following the raised verb but crucially preceding the right-dislocated object. In contrast (b) is ungrammatical because the same subject now follows the right-dislocated indirect object, thus being right-dislocated itself as in the previous cases. Native speakers assessing these data must consider the provided context, possibly reading it aloud. This is necessary to ensure that the negative phrase does indeed acquire the discourse given status necessary to license marginalization.
(p.32) Another example, this time involving a negative object, is provided in (16). Once again the negative object is grammatical in (16)(a) where it is marginalized in situ and ungrammatical in (16)(b) where it follows the clitic-doubled right-dislocated object alla festa.
Negative phrases thus provide a powerful tool to separate marginalization from right-dislocation. The next sections exploit this and other diagnostic tools to investigate the position of marginalized constituents.
2.3.1 Evidence from the ordering of negative phrases and NPIs
A first piece of evidence for the position of marginalized phrases comes from the just mentioned observation that postverbal negative phrases can be marginalized. Since they are grammatical, they must lie within the domain of a licenser. A neg-marker in T provides the lowest possible licenser. Therefore, marginalized negative phrases must occur lower than T, consistently with Cardinaletti’s claim that marginalization occurs in situ (Calabrese 1982; Cardinaletti 2001, 2002).
Marginalized phrases also disallow for clitic doubling. This holds even in contexts that can trigger right dislocation and the associated clitic doubling. For example, sentence (16)(a) becomes ungrammatical as soon as an object clitic is inserted, see (17). As Cardinaletti (2002) points out, Italian disallows for clitic doubling within the clause. If marginalized phrases occur in situ, hence TP-internally, the absence of clitic doubling follows straightforwardly.
A third piece of evidence comes from the observation that whenever multiple marginalized constituents are present, they necessarily follow the canonical base-generated order, as expected if marginalization occurs in situ (Cardinaletti 2001, 2002). The sentences in the following examples provide further evidence for this claim. In (18), the subject and the object are both marginalized. The subject is a negative quantified (p.33) phrase, and the object is a an NPI. They are both licensed by the c‑commanding neg-marker and hence excluding a right dislocation analysis. The sentence is fine when the subject precedes the object but ungrammatical when the order is reversed.
It is worth being aware that some native speakers of Italian systematically disallow for non‑final VP-internal subjects, negative subjects included. This aspect of their grammar makes the following test unsuitable for them, since they inevitably find all sentences involved ungrammatical. Even these speakers, however, converge with the other informants on tests that follow the same logic but do not involve VP-internal subjects (e.g. (20)). Furthermore, the reported difference in grammaticality is clearly perceived by the other informants.
The marginalized subject and locative arguments of an unaccusative verb also obligatorily follow their base-generated order.
The base-generated order is also obligatory when marginalization applies to verbs taking experiencer objects followed by a clausal complement.
(p.34) Finally, the base-generated order is also obligatorily adhered to in sentences involving marginalized subjects and sentential objects (the object includes the NPI alcunché ‘anything’ licensed by the initial neg-marker, thus ensuring its non right-dislocated status). Compare (21)(a), where the base-generated order is respected, with the ungrammatical (21)(b), where the order of the two phrases is switched. Sentence (21)(a) provides a pragmatically acceptable reply to the context sentence in a situation where some people who have already eaten do not wish to publicly admit it (possibly because invited to an important dinner). The person uttering (21)(a) is correcting the person uttering the context sentence.
The rigid order observed in all the above examples provides strong support for the claim that marginalization occurs in situ, since different orders should be possible if marginalized phrases were able to move. In fact, even ultra-local movement of marginalized phrases, i.e. movement so local that it cannot alter the base-generated order, is excluded. If such movement were present, it should be possible to raise DPs ultra-locally while stranding a quantifier in their base-generated position, thus altering the canonical order <Quantifier DP> of Italian nominals. But this is not the case; see (22) where the marginalized object NPI following the marginalized quantified subject ensures that right dislocation is absent. As (22)(b) shows, stranding the subject quantifier via ultra-local movement of the DP i ragazzi (the boys) is ungrammatical.
2.3.2 Evidence from anaphoric and quantifier binding
The binding relations between marginalized phrases are also consistent with their in situ position (see also Cardinaletti 2001). Since in situ subjects c-command in situ (p.35) objects, we expect marginalized subjects to bind marginalized object anaphors but not vice versa, as in the schema in (23) where the arrow shows the direction of binding between the postverbal marginalized subject and object.
As (24) and (25) show, the prediction is borne out.6 The possessive anaphor propri ‘own’ is successfully bound when occurring as a marginalized object in (24) but not as a marginalized subject in (25) (on the anaphoric status of ‘propri’ see Renzi 1988: 614).
The context sentence of (25) is a passive because this is the only construction that can express the desired meaning. To check whether this choice is a factor in the ungrammaticality of (25), rather than the position of marginalized phrases, consider (26). It tests whether negative objects like the one in (25) are possible under a passive context sentence when binding is not present. Sentence (26) is marginal but clearly better than (25), confirming that it is a binding failure that causes the ungrammaticality of (25). (p.36)
The same asymmetric relation between marginalized subjects and objects are also found under quantifier binding. Consider first the sentences in (27): the quantified subject nessun ragazzo ‘no boy’ in specTP binds the possessive pronoun suo ‘his’ in object position in (27)(a), but the same possessive pronoun in subject position cannot be bound by the quantified object in (27)(b). (On the pronominal nature of ‘suo’ see Renzi 1988: 614.)
The exact same pattern is found when subject and object are marginalized, showing marginalized subjects binding—and hence c-commanding—marginalized objects, but not vice versa, as predicted if marginalized phrases occur in situ. Note that the shift from active to passive in the context sentence is not responsible for the ungrammaticality of sentence (29), as the same sentence is fine when the possessive is not interpreted as bound.
2.3.3 Evidence from agreement loss in regional Italian
Cardinaletti (2001: 131, 2002) proposes evidence for in situ marginalization based on the regional variety of Italian spoken in the area of Ancona. In this variety, preverbal and right-dislocated subjects require agreement in number and person with an auxiliary in T, as in (30)(a) and (30)(b) (note that the subject of (30)(b) is necessarily right-dislocated because it follows a clitic-doubled object). As Cardinaletti (2002) points out, agreement in this case follows from the presence of a preverbal pronominal pro doubling the dislocated subject, as shown in (30)(c), and causing obligatory agreement with T as with any other preverbal subject. (p.37)
In contrast, focused and marginalized postverbal subjects, respectively shown in (31)(a) and (31)(b), allow for a default third person singular agreement even when plural.7 This is expected if these subjects are located lower than T and hence unable to enter into an agreement relation with the auxiliary in T, as expected if marginalized phrases occur in situ (data from Cardinaletti 2001 and Cardinaletti p.c.).
Further support comes from the fixed order of marginalized constituents in clauses showing agreement loss. As (32) shows, marginalized subjects must precede marginalized objects, reflecting their base-generated order. Compare this with the right dislocation sentence in (30), where the order between subject and object remains free.
2.3.4 Evidence from past participle preposing
Further evidence comes from past participle preposing (Benincà 1988; Cardinaletti 2002). Italian active past participles obligatorily raise to an aspectual phrase above VP (Cinque 1999). This is shown in (33)(a)–(b), where the past participle, in bold, necessarily precedes the adverb bene (well), the lowest adverb in Cinque’s adverbial (p.38) hierarchy. Sentence (33)(c) shows that the same past participle may also optionally raise even higher and precede the negative particle mica and the adverbs completamente ‘completely’ and già ‘already’, which are located relatively high in Cinque’s hierarchy.
As expected, past participles may also precede subject-related floating quantifiers stranded in specVP position, see (34)(a). Consequently, preposing of the aspectual phrase containing the past participle necessarily carries with it any subject and object stranded VP-internally. This is shown in (b) and the corresponding structure in (c). (A suitable discourse context for sentence (b) is provided by sentence (a).)
If marginalization occurs in situ, then past participle preposing should be blocked whenever marginalized subjects and objects are marginalized in post-auxiliary position, since this makes it impossible to prepose the aspectual phrase headed by the past participle as an unbroken phrasal constituent (Cardinaletti 2002). The prediction is borne out. Whether the marginalized subject and object both follow the auxiliary, as in (35)(a), or just one of them does, as in (35)(b) and (35)(c), past participle preposing is no longer possible.8 (p.39)
2.3.5 Evidence from the ordering of lower adverbs
Finally, marginalization in situ is supported by the study of lower adverbs. Cinque (1999) shows that the basic word-order of lower adverbs—i.e. the adverbs that occur between the highest position available to an active past participle and its complements—is fixed. For example, when the entire sentence is presentationally focused, the adverbs in example (36) cannot occur in any other order (Cinque 1999: 6).
If marginalized phrases remain in situ, then lower adverbs should preserve their fixed order even when marginalized. The prediction is borne out. Consider example (37), where the context sentence states that the expected worsening of John’s sight is indeed happening, while the reply in (37)(a) explains by contrastively focusing the verb that what is getting worse is actually John’s hearing. The marginalized adverbs in sentence (37)(a) follow the same fixed canonical order of the adverbs in the context sentence. Any change in this order is ungrammatical, see (37)(b)–(e). Note that the negative object must be licensed by the preverbal neg-marker, hence both the adverbs and the final object cannot have right-dislocated status.
The fixed sequence of marginalized lower adverbs can also be exploited to examine the position of marginalized objects, which have already been shown to obligatorily (p.40) follow marginalized subjects. As the data below show, marginalized objects (in bold) also obligatorily follow lower adverbs, consistently with the claim that marginalization occurs in situ. The only exception is sentence (c), discussed below, where the object is marginally grammatical.
The marginal grammaticality of (38)(c) follows from the right dislocation of the adverbial sequence sempre bene, which appears unique in allowing for right dislocation. This is shown in (39), where different adverbial sequences are placed after a right-dislocated clitic-doubled object, thus leaving no doubt about their right-dislocated status. As the example shows, the only marginally grammatical sentence is (39)(a) involving the right dislocation of sempre bene. While the reason for this divergence remains unclear, its presence explains the pattern in (38). The object is marginalized in situ in all sentences. (38)(c) is grammatical because sempre bene is right-dislocated after the object. In all other cases, right dislocation is unavailable and the sentence is ungrammatical because the marginalized adverbs and the object do not follow the fixed base-generated order.
The base-generated order is also evident when considering marginalized adverbs and subjects, as in (40). Like marginalized objects, marginalized subjects may remain in their base-generated specVP position and follow the entire adverbial series, as in (40)(a). As expected, the subject may not precede any adverb except the right-dislocated sempre bene in (40)(b), in parallel with the object case examined above. Unlike objects, however, marginalized subjects may also precede the entire adverbial series, see (40)(f), possibly for reasons associated with case-assignment. (p.41)
Marginalized subjects provide the only exception to Cardinaletti’s claim that marginalization always occurs in situ, although even in this case marginalization in situ remains possible, as shown by (40)(a). More importantly, the properties that distinguish marginalization from right dislocation remain unaffected. The raised marginalized subject of (40)(f) is still c-commanded by T, hence still disallowing for clitic doubling. Furthermore, the higher position accessed by subjects in (40)(f) remains inaccessible to objects, as (38) showed, hence preserving the fixed order that characterizes marginalized arguments when compared with the freely ordered right-dislocated ones.
The distribution of marginalized phrases emerging from this chapter is summarized in (41). Marginalized objects follow marginalized subjects and lower adverbs, see (41)(a). Marginalized lower adverbs respect their base-generated order as in (41)(b). Marginalized objects precede marginalized sentential complements, as in (41)(c), again reflecting the corresponding base-generated order.
The examined data thus support Cardinaletti’s (2001, 2002) claim that marginalized phrases occur in situ, with the exception of subjects which appear able to optionally raise above lower adverbs. In the coming chapters, these results will be used to distinguish marginalized phrases from right-dislocated ones, which instead move outside the clause. This, in turn, will make it possible to accurately determine the position of contrastively focused constituents within the clause.
(1) ‘Marginalization’ is a direct translation of the Italian term ‘emarginazione’ used in Antinucci and Cinque (1977) in their seminal study of Italian right peripheral constituents. Its meaning, however, has changed over time. In both Cardinaletti (2001, 2002) and this book ‘marginalization’ is used in opposition to ‘right-dislocation’ to distinguish post-focus discourse-given phrases located in situ from post-focus discourse-given phrases right-dislocated outside the clause. This differs from Antinucci and Cinque’s original use where ‘emarginazione’ characterized any construction where discourse-given constituents obligatory occur at the right edge of the clause, independently of whether such positioning involves movement or in-situ realization. For example, the obligatorily displaced postverbal subjects of Italian interrogative clauses, like Cosa ha MANGIATO, Gianni (what has eaten, John) were considered marginalized independently of their actual position. An in-situ clause-final subject in a declarative clause, however, would not be considered marginalized because declaratives also allow for preverbal subjects. Consequently, the constituents analysed as marginalized in Antinucci and Cinque (1977) would today be analysed as either marginalized (i.e. destressed in situ) or right-dislocated depending on their final position. More importantly, the properties originally attributed to ‘emarginazione’ by Antinucci and Cinque do not necessarily hold under its new meaning. For example, the free ordering of marginalized constituents claimed by these authors does not apply to marginalized phrases under today’s interpretation of the term (Cardinaletti 2001, 2002); it actually identifies an important attribute of right-dislocated phrases (see Chapter 4).
(2) For example, the order <V S O> will be shown to be a legitimate order when V is focused and S and O are marginalized, yet the same order is ungrammatical when the entire clause constitutes a new information focus.
(3) Costa maintains that presentationally focused subjects can occur in specVP immediately before presentationally focused objects. He also notes, however, that such subjects must carry main stress, raising the doubt that his data have not been elicited under a context forcing sentence-wide presentational focus, since such discourse context would impose rightmost stress.
(4) Postverbal subjects are possible with unaccusative and directional verbs, where the subject is generated as an object and specTP can be filled by a silent directional particle, see Burzio (1986) and Pinto (1997) for discussion.
(6) Cardinaletti’s own examples (Cardinaletti 2001: 133) concern phrases introduced by the quantifier ogni ‘every’ under the assumption that phrases with this quantifier cannot be right-dislocated. This appears incorrect, see Section 18.104.22.168 for examples and discussion. Some of Cardinaletti’s examples also rest on the assumption that right-dislocated objects obligatorily require clitic doubling, but this assumption, too, will be proven inadequate; see Section 4.2.1 for discussion.
(7) Agreement loss with lower postverbal subjects is present in many regional varieties of Italian and also crosslinguistically, see Moravcsik (1978), Corbett (1979), Brandi and Cordin (1989), Barlow (1992), Saccon (1993), Fassi Fehri (1993), Manzini and Savoia (2002), and Samek-Lodovici (2002).
(8) Postfocal constituents can be ambiguous between a marginalized and right-dislocated analysis but this aspect only needs to be controlled for with grammatical sentences, lest data involving right-dislocated phrases are used to draw conclusions about marginalized ones, or vice versa. The sentences in (35) are ungrammatical, thus unacceptable under both analyses, hence also under a marginalization analysis, as required for the point concerning past-participle preposing just mentioned.