The Fine Structure Of The Left Periphery

Selection: The restrictions that a lexical item places on its arguments. (Usually considered a property of the complement) Verbs can select for different clause types: believe: declarative clause wonder: interrogative clause know either declarative or interrogative The fine structure of the left periphery. In L. Haegeman. (ed.), Elements of Grammar. Dordrecht: Kluwer. Main point: Just as there was evidence to split IP up into several projections (TP, AgrSP, etc.), there is also evidence that CP likewise needs to be split. The structure of CP proposed is: (2) a. b. Pat believes that Tracy left. * Pat believes who left. (1) (3) a. b. * Pat wonders that Tracy left. Pat believes who left. (4) a. b. Pat knows that Tracy left. Pat knows who left. ForceP 3 3 Force TopP* specifies 3 clause type 3 Top FocP hosts 3 topics 3 Foc TopP* hosts foci, 3 wh-phrases 3 hosts more Top FinP topics 3 marks 3 finiteness Fin ”IP” Part I. Conceptual arguments; introduction to the articulated CP THE FORCE–FINITENESS SYSTEM What does the complementizer do? It serves as the interface between two clauses: —The VP above, of which the embedded clause is a complement. —The IP below, which the CP contains. Selection must be local—if a head selects for a certain kind of complement, that complement must be its complement (not just somewhere inside its complement). CP must have, at the top, some specification at least distinguishing between declarative and interrogative clauses (the “force” of the clause). Similarly, at the bottom, for C to be sensitive to finiteness (such that certain kinds of C select for finite clauses, others select for nonfinite clauses), C must have some specification of (what kind of) finiteness (it embeds). These appear to be two different functions, and could conceivably be expressed as properties of two different heads in the “CP system.” This gives Rizzi’s split an initial plausibility, evidence will follow that we can put things in between the Force position and the Finiteness position, forcing us to the position that we must at least split CP up into two. THE TOPIC–FOCUS SYSTEM Another well-known use of CP is as the area to which left-dislocated things (generally topics and foci) appear. Viewed from the outside (above), the complementizer indicates what kind of clause it is (kinds include declarative, interrogative, comparative, adverbial, …). (5) Viewed from the inside (below), the complementizer shows a certain degree of sensitivity to the finiteness of the IP. For example, that occurs with finite clauses, for occurs with nonfinite clauses—but a pretty minimal specification, probably just ±finite. At first sight, English appears to use basically the same position for both, but languages differ in this respect, and very often treat the two types of sentences differently. a. b. Your book, you should give t to Paul (not to Bill). YOUR BOOK you should give t to Paul (not mine). Topicalization Focus fronting Brief side note on topic and focus: What topic and focus are exactly is a complex issue, and the way to precisely formulate their meanings has generated a great deal of literature. Some approximations: Topic: What the sentence is “about”. Sometimes paraphrasable with As for… (6) As for your book, you should give it to Paul (not to Bill). Topic is generally old information, known from preceding discourse. Focus: What is new in the sentence. One place focus can be found is in the answer to a wh-question, where the wh-word was in the question: (7) Q: What did Pat buy? A: Pat bought [F stamps ]. Focus is also very often used contrastively, as in: (8) Well [F I ] passed the test. (suggests a contrast between me and others, who we deduce did not pass the test). Complaint: Rizzi observes that “focalization” (fronting of a focus) is only available for contrastive focus in Italian, and notes that there is another kind of focus which is not fronted (left in situ, and indicated by stress) and which may or may not be contrastive. Rizzi speculates that such focus may move into the focus position covertly (between SS and LF); however, one problem with this is that this means that at LF the two kinds of focus (overtly fronted, covertly fronted) are not distinguishable. If the interpretation is derived from the LF representation, we are left without any explanation for why overtly fronted focus phrases have to be contrastive. In the recent literature on focus, several people have argued that contrastive focus is a separate kind of thing from the in situ type focus, and thus should be distinguishable at LF. For two prominent examples see É. Kiss (1998), Vallduví & Vilkuna (1998). Very interesting issues arise, many right at the heart of questions of how syntax, semantics, and pragmatics interact. (If you are interested, check the “old courses” section of my web site [http://www.bu.edu/linguistics/UG/hagstrom/] and look at the seminar on focus from Spring 2000). Wh-Criterion: Criteria: A phrase with feature [α] must be in a Spec-head configuration with a functional head carrying [α]. In earlier work, Rizzi proposed that wh-movement is driven by the Wh-Criterion, which essentially requires that a wh-phrase (carrying the feature [wh]) must end up in a Spec-head relation with C (also carrying the feature [wh] in questions).* * Italian: (9) a. b. Il tuo libro, lo ho letto ‘Your book, I have read it’ IL TUO LIBRO ho letto (, non il suo) ‘Your book I read (, not his)’ Focus preposing (focalization) There are two dimensions here (two “articulations” of the sentence): Topic–Comment Focus–Presupposition More or less. Actually Rizzi argues that T carries [wh] and that T moves to C (bringing [wh] along with it), where the wh-phrase can enter a Spec-head relation by moving to SpecCP. Clitic Left Dislocation (CLLD) (topicalization) Topic (what the sentence is about) Comment (what the sentence says about the topic) Focus (what is new in the sentence) Presupposition (what is old/presupposed in the sentence) And Rizzi assumes that these splits are effected by an X-bar structure. His motivation for TopP and FocP are at this point completely conceptual. (except one brief mention that there may be a particle in Gungbe which should be analyzed as an overt Foc head). Neg-Criterion: Next week, we will see another use of Criteria with respect to the Neg-Criterion, which will require that things with a [Neg] feature (like nobody) must wind up in a Spec-head configuration with the functional head Neg (also with a [Neg] feature). Rizzi extends these notions to Topic and Focus as well, using a Topic-Criterion and a Focus-Criterion to compel topics and foci to move into the specifiers of TopP and FocP, respectively. His idea is that it is tied to interpretation—e.g., if you want to be interpreted as a topic, you’d better be in SpecTopP. If TopP and FocP are part of the CP, yet FinP must be at the bottom and ForceP must be at the top (for selectional reasons), then it follows that TopP and FocP must be “sandwiched” between ForceP and FinP. FIRST POSITIONAL ARGUMENTS FOR A SPLIT-CP: DIFFERENCES BETWEEN TOPIC AND FOCUS Topics and foci give us our first bit of evidence that CP must be split up this way, as they seem to come between the positions of Force and Fin. TopP and FocP so far seem to be pretty much the same; but there is an important difference: Focus is quantificational, Topic is not. Italian: Quantification: (10) (11) (12) (13) di introduces non-finite embedded clauses—and follows topics che introduces finite embedded clauses—and precedes topics. a. Credo che loro apprezzerebbero molto il tuo libro. ‘I believe that they would appreciate your book very much.’ b. Credo di apprezzare molto il tuo libro. ‘I believe of to appreciate our book very much.’ a. Credo che il tuo libro, loro lo apprezzerebbero molto. ‘I believe that your book, they would appreciate it a lot.’ b. * Credo, il tuo libro, che loro lo apprezzerebbero molto. (‘I believe, your book, that they would appreciate it a lot.’) a. * Credo di il tuo libro, apprezzarlo molto. (‘I believe of your book to appreciate it a lot.’) b. Credo il tuo libro, di apprezzarlo molto. ‘I believe your book of to appreciate it a lot.’ … (di) … topic … (che) … IP Relative wh-words must precede topics, interrogative wh-words must follow topics (14) (15) (16) a. Un uomo a cui, il premio Nobel, lo daranno senz’altro ‘A man to whom, the Nobel Prize, they will give it undoubtedly.’ b. * Un uomo, il premio Nobel, a cui lo daranno senz’altro (‘A man, the Nobel Prize, to whom they will give it undoubtedly.’) a. * A chi, il premio Nobel, lo daranno? (‘To whom, the Nobel Prize, will they give it?’ b. Il premio Nobel, a chi lo daranno? ‘The Nobel Prize, to whom will they give it?’ … (Rel-wh) … topic … (wh-word) … IP Point: We can’t account for this if there is only one C, one SpecCP, to do everything. A quantificational structure has an operator and a variable. The variable is generally in argument position. The operator takes a set of possible values and evaluates the (part of the) sentence (in its scope) for those values. [Every student]i … [IP ti passed the test] Operator, variable, assigned a value by the operator ranges over students The interpretation goes as follows: The operator checks the truth of the IP for each of the values in its range. student A passed the test. student B passed the test. … student Z passed the test. T/F? T/F? T/F? For every student, the whole thing is true if every value it checks turns out True. For some student, the whole thing is true if at least one value it checks turns out True. For most students, the whole thing is true if most of the values it checks turn out True. and so forth. When Rizzi says “focus is quantificational” he means that focus phrases make use of a range of values which get assigned to variable (trace), and that the meaning requires reference to evaluations at different values. When Rizzi says “topic is not (quantificational)” he means that the value which the trace gets in the interpretation does not vary—it is still assigned a value by the topic phrase, but that value is constant. Rizzi refers to traces of this kind (which are otherwise like variables except that their values don’t vary) as null constants. Some empirical differences between topic and focus (in Italian) and some description of how this follows from Rizzi’s “FIQTIN” proposal: 1. Focus exhibits Weak Crossover effects (Topic does not) 2. Topic has a resumptive clitic (Focus does not). (17) a. (21) a. Giannii, suai madre loi ha sempre apprezzato. ‘Gianni, his mother always appreciated him.’ b. ?? GIANNIi suai madre ha sempre apprezzato (non Piero). (‘GIANNI his mother always appreciate, not Piero.’) Weak Crossover has elsewhere been proposed to be sensitive to true quantification. That is, a relation involving a null constant (non-varying) trace will not trigger WCO. (18) a. ?* Whoi does hisi mother really like ti? b. John, whoi hisi mother really likes ti. The proposed analysis of (18b) works like this: ti gets its value from whoi, which gets its (single) value from John. Who would be considered to be an anaphoric operator in this case; it gets its single value from a nearby higher referent and passes it along to its trace. The trace is thus a null constant, the relation between the anaphoric operator who and this null constant is not quantificational, and therefore not subject to WCO. (Rizzi in fact takes it a step further and enforces this for null constants—Null constants are only licensed by anaphoric operators of this type). Sidetrack: Rizzi wonders why Your book I bought is ok in English, given that its literal translation is ungrammatical in Italian (see next section: Italian would require a clitic). Rizzi’s explanation for why the Italian version is bad might appear to predict that the English one should be bad for the same reason. His way out is to say they have different structures, the English one being: (19) Your book [Op [I bought t]] where Op is an anaphoric operator getting its reference from the nearby your book, and the t is a null constant. Hence we expect to find no WCO in English in these constructions, and this is true: (20) John, hisi mother really likes ti. Italian lacks this kind of structure (or can’t make use of it because it has clitics and clitics are preferred over Op). (22) Il tuo libro, lo ho comprato. ‘Your book, I bought it.’ b. * Il tuo libro, ho comprato. (‘Your book, I bought.’) a. * IL TUO LIBRO lo ho comprato (non il suo). (‘YOUR BOOK I bought it (not his).’ b. IL TUO LIBRO ho comprato (non il suo). ‘YOUR BOOK I bought (not his).’ Short version: Quantifiers must bind variables, and clitics interfere (22) The trace of an A′-movement must be bound by either a quantifier or an anaphoric operator. (21) Long version: Rizzi makes some non-transparent assumptions about the way clitics work. He appears to be assuming that in the cases where there are clitics, the clitic starts out in object position and moves to where we see it, and the focus/topic phrases are just inserted up in TopP or FocP. Further, the clitic is a head and so (at least its final move) leaves an X°-trace and not an XP-trace. Only XP-traces can be variables. Starting with Focus (22), a focus phrase is quantificational. An important requirement on quantifiers is that quantifiers must bind a variable. It is not possible to have “vacuous quantification” (i.e. For every student x, x left is ok, but For every student x, John left is not). Rizzi refers to this under the name Full Interpretation sometimes (which essentially is the idea that everything that appears in the LF representation must contribute to the interpretation). In (22a), the focus phrase does not have a variable to bind. The clitic can’t be a variable, nor can its trace, and that’s all the options there are. In (22b), we can assume that the focus phrase itself moved into FocP and the focus phrase binds its own trace in object position. In (21b), there is an empty category in object position—but there is no kind of empty category that it could legitimately be. It can’t be a DP-trace (the trace of A-movement) because that would make it the trace of movement of the topic phrase to TopP (which is movement to an A′-specifier—it’s A′-movement). It can’t be a variable because there is no quantifier to bind it. It can’t be PRO because it’s an object. In (21a), the empty category there is the trace of the clitic. A note on Rizzi’s apparently strange terminology concerning empty categories: Early on (Rizzi 1986), Rizzi proposed thinking about traces (and the syntax as a whole) in a different way. Rather than saying that there are constraints on movement such that it can only do such-and-such a thing over so-and-so many nodes of a certain type, Rizzi proposed that there are constraints on the representation you end up with. One aspect of this was that empty categories of all kinds were not considered to be inherently a trace or a PRO or a pro—you look at where the empty category (ec) is in the final representation and look at what is around to “license” it. Based on that you can determine what kind of empty category it is. The problem with (21b) is that the ec can’t be any of the available types. Sidetrack: Rizzi observes that his story only holds up with bare quantifiers—everything, noone, etc. As soon as you add a restriction (many books, …) you do seem to be able to topicalize quantifiers. 3. Bare quantificational elements (noone, all, …) can be Foci (but not Topics) (26) (23) a. (24) (25) In order to save the story, Rizzi proposes that in these cases, the DP molti libri moves into SpecTopP, but that the quantifier is really molti ‘many’, which undergoes QR out of the DP, leaving behind a trace which can serve as a proper variable to be bound by molti. * Nessuno, lo ho visto. (‘Noone, I saw him.’) b. * Tutto, lo ho fatto. (‘Everything, I did it.’) a. NESSUNO ho visto. ‘NOONE I saw.’ b. TUTTO ho fatto. ‘EVERYTHING I did’ The problem with (23) is that quantificational elements are quantifiers and as such they need to bind a variable—but there is no variable in the topic structure. In (24), the trace of the focused quantifier can do double-duty; it is a legitimate variable both for the focus-related quantification and for the quantifier-related quantification. Side note: When we talk about contrastive focus, we are contrasting the focused element with “alternatives” to the focused element. In a case like MARY passed the test, we are contrasting Mary with, e.g., her classmates. When we focus a quantifier, though, it is reasonably to think that the alternatives are quantifiers: EVERYTHING I did contrasts everything with things like nothing, most things, at most 7 things, … Possibly related other side note: One might wonder how legitimate it really is to use the same trace to receive the alternative values from the focus phrase and to receive the alternative values from (each alternative) quantifier. How can we formally express what this means? Molti libri, li ho buttati via ‘Many books, I threw them away.’ ?P qp moltii ... legit. variable the position to which QR moves quantifiers. TopP qp DPj Top′ # qp ti libri Top ... # not a legit. variable ... tj ... Comment. This looks like something for which it would be very hard to give a proper semantic analysis. I have no idea what this is supposed to mean. It looks like (as best I can tell): For many x, x is a book and I threw x away. This is not the right meaning, though. It’s even clearer for most—Most books, I threw them away would be For most x, x is a book and I threw x away, but when you consider the vast number of things which aren’t books it’s very clear that it is going to wind up false, since Most things are books isn’t even true. 4. No limit on number of Topics (but only one Focus) (27) a. b. (28) Il libro, a Gianni, domani, glielo darò senz’altro ‘The book, to John, tomorrow, I’ll give it to him for sure.’ * A GIANNI IL LIBRO darò (non a Piero, l’articolo) (‘TO GIANNI THE BOOK I’ll give (not to Piero, the article).’) A Gianni, QUESTO, domani, gli dovrete dire. ‘To Gianni, THIS, tomorrow, you should tell him.’ Descriptively, topics and either/both precede and follow the focus: (29) (30) … C (Top*) (Foc) (Top*) … One way to think about this is: In X-bar theory, XP has only one Spec, but an unlimited number of things can adjoin to XP. So: Focus moves to SpecXP and Topic adjoins to XP. (31) Nevertheless, Rizzi goes for his view about TopP and FocP Forcing us to assume TopP can recurse (A Top° can take another TopP as complement ), while FocP cannot recurse. Why FocP cannot recurse —Luigi Rizzi Rizzi says that FocP is responsible for splitting up the focus and the presupposition (essentially, the new and the old/given). Suppose we have recursive FocP: Foc1 takes FocP2 as its complement. In the complement of Foc2 is given information, in the Spec of Foc2 is new information. But, FocP2 is in the complement of Foc1, and as such must as a whole be given information. Thus, we have an incompatibility: the Spec of Foc2 cannot simultaneously be new and old information, and hence recursion of FocP is ruled out by interpretive concerns. Comments. The purported inability to have two focus phrases is controversial. É. Kiss (1998) argues that you can have two focus phrases in Hungarian, and even the English example: No, it is false that John gave the book to Mary—BILL gave THE RECORD to Mary seems to have two contrastive foci (neither seems to be a contrastive topic). In this connection, it might be relevant to consider this: Rizzi tells us that focus and whwords cannot co-occur, and the two foci cannot co-occur. But it is also true of Italian that two wh-words cannot co-occur—you cannot ask Who bought what? in Italian (in that way). Of course, many languages do allow multiple wh-questions, including English. Perhaps this points to some Italian-specific restriction that is involved both in constraining wh-words and foci to one? 5. Topic is compatible with wh-words; Focus is not, either is ok with relative Op. (32) a. Un uomo a cui, il premio Nobel, lo daranno senz’altro ‘A man to whom, the Nobel Prize, they will give it undoubtedly.’ b. * Un uomo, il premio Nobel, a cui lo daranno senz’altro (‘A man, the Nobel Prize, to whom they will give it undoubtedly.’) a. * A chi, il premio Nobel, lo daranno? (‘To whom, the Nobel Prize, will they give it?’ b. Il premio Nobel, a chi lo daranno? ‘The Nobel Prize, to whom will they give it?’ … (Rel-wh) … topic … (wh-word) … IP Conclude: Relative wh-operators occupy SpecForceP (only place which can’t follow a topic). Wh-words in a wh-question occupy someplace lower, below topics. Focus is incompatible with wh-words regardless of the order: (33) a. * A GIANNI che cosa hai detto (, non a Piero)? (‘TO GIANNI what did you tell (not to Piero)?’) b. * Che cosa A GIANNI hai detto (, non a Piero)? (‘What TO GIANNI did you telll (not to Piero)?’) Suggests: Wh-words and focus compete for the same spot, SpecFocP. There’s only one SpecFocP, so you can only have one of {wh-word, focus}, Consistent with the above results, relative operators can co-occur with focus but the relative operator must precede the focus: (34) a. b. Ecco un uomo a cui IL PREMIO NOBEL dovrebbero dare (... non il premio X) ‘Here is a man to whom THE NOBEL PRIZE they should give (...not prize X)’ * Ecco un uomo IL PREMIO NOBEL a cui dovrebbero dare (...) (‘Here is a man THE NOBEL PRIZE to whom they should give (...)’) Recall from before… Relative wh-words must precede topics, interrogative wh-words must follow topics And here, the comprehensible part of the paper is pretty much over. The rest of the paper deals with various other phenomena in terms of headmovement, criteria, and the ECP. Ruling out a recalcitrant case: wh…Top… Earlier we saw that wh-words cannot be followed by a topic. If wh-words move to FocP, should we expect that the TopP below it would be able to house topics? Should we conclude that wh-phrases don’t move as high as FocP? We don’t want that; but perhaps it can be tied to a separate fact about Italian: In a normal wh-question, even a normal preverbal subject cannot intervene between the wh-word and the verb. We account for this by assuming that [wh] is generated on T, and then T moves to C (taking the verb past the subject) so that the [wh] on T (moved to C) can be in a Spec-head relationship with the [wh] feature of the wh-phrase in SpecCP. In the case we’re trying to rule out, there is a lower TopP below the FocP where the whword is. The [wh] feature that appears on T in all wh-questions still needs to be in a Spechead relationship with the wh-word (T has to move to Foc), but we know it hasn’t because the verb does not precede the topic. In a normal focus construction, a preverbal subject can intervene between the focus and the verb, so we must conclude that the [Foc] feature implicated in the Foc-Criterion (forcing movement of the focused phrase to SpecFocP) is generated right on Foc° (and need not be moved up there). In support of a TopP for non-argument preposing (vs. IP-adjunction)… 1) If it were just adjunction, why is it only to IP? Why not to other places in the clause where we see adjunction in other cases? 2) We can explain certain adjacency effects if we assume TopP is involved in nonargument preposing and we assume that Top° can interfere with the licensing of certain elements. Consider for, which bestows Case on the subject of an embedded nonfinite IP. A preposed adverb cannot intervene between for and the subject—assuming that there is a TopP when there is a preposed adverb, for no longer governs the subject—it is too far away. (35) a. b. ...for John to leave tomorrow. * ...for, tomorrow, John to leave. (36) a. b. With John unavailable on the weekend... * With, on the weekend, John unavailable... The fact that that is compatible with TopP comes from the fact that it’s Force and it’s above the fray. It has no relation to the embedded subject (it does not bestow Case), so it has no need to be local to it. (37) a. b. ...that John will leave tomorrow. ...that, tomorrow, John will leave. Side discussion: Crosslinguistic reasons to believe Case is relevant. West Flemish: With infinitives embedded by prepositional complementizers that aren’t bestowing Case (where the subject shows up in nominative Case anyway), preposed adverbs can occur between the complementizer and the subject. Brazilian Portuguese: Formal: subject of complement of pra ‘for’ nominative (preposing ok) Colloquial: subject of complement of pra ‘for’ oblique (preposing disallowed) Side side discussion: How about English adjacency requirements attributed to I C? If I moves to Fin, there’s no structural TopP position to allow preposed adverbs between the moved auxiliary and the subject. But if the pair below mean the same thing and indicate that in (38b) had has moved up to where if is in (38a), what could be wrong with the preposed adverbial in (38b)? (38) a. If (yesterday) John had done that... b. Had (*yesterday) John done that... Solution: I is prohibited from moving to Top° in English, so if you put a Top° in there, either I will have to move to Top° (prohibited) or I will have to skip over Top° (violating the Head Movement Constraint). The Split CP and the ECP ADJACENCY EFFECTS ON TRACES—where preposed elements block subject extraction. (Recall that a subject trace is not automatically lexically governed and is therefore very susceptible to ECP violations). The mechanism which allows extraction out of a non-governed position: (39) XP 3 DPi X′ Spec-head agreement: transfers Agr features for i to X° 3 Due to agreement, Xi now serves as a proper X° IP governor for ti. 3 ti I′ ... But, if there were an intervening TopP, the mechanism wouldn’t work. CONCERNING THE LACK OF ANTI-ADJACENCY EFFECTS IN FRENCH (40) French that-trace type effects are not alleviated by intervening topicalized adverbials. XP 3 DPi X′ Spec-head agreement: transfers Agr features for i to X° 3 Despite agreement, Xi cannot serve as a proper X° TopP governor for ti—because it doesn’t 3 govern ti. topicalized3 item Top° IP 3 ti I′ ... This is supposed to be at the root of: (41) a. b. ANTI-ADJACENCY EFFECTS—where preposed elements salvage subject extraction a. b. (43) TopP 3 3 Top° FinP 3 t′ 3 allowed Fin° IP in English Ø/qui 3 t ? (This is) a man to whom [liberty [we should never grant t t ]] * (This is) a man who [liberty [t should never grant t to us ]] (Fin° is the closest head to the subject trace—it would need to have agreement features for the subject, but it has Op in it already). (42) French has an overt agreeing complementizer (Fin): qui. English has Ø, which Rizzi (following Pesetsky) suggests is an affix that moves up to the verb embedding it. * (This is) an amendment which they say that t will be law next year. (This is) an amendment which they say that, next year, t will be law. It seems that topicalization licenses a head which can license the subject trace. A good candidate for the culprit is Fin°, since it governs the subject trace. Why is a preposed adverbial required? Suggestion: Without something in the topic-focus field, that is actually spelling out a conglomerate head—in such structures there is no split CP, there is just C with all of the features of Force and Fin in one head (enforced by Avoid Structure). That cannot agree, and so the closest head to the subject trace cannot be made into a proper governor. But if the topic-focus field is “activated”, then that becomes unambiguously Force (that is not compatible with Fin alone), and Ø becomes unambiguously Fin (not compatible with Force alone). Ø can agree, allowing the subject trace to be licensed. Never mind about the last couple of pages on “growing Agr”—it is not clear and isn’t featured in papers you are likely to see…