CAS LX 523 Syntax II More modern clauses tend to look more like this (at least): Spring 2002 January 14, 2002 Paul Hagstrom Week 1: Introduction, split-Infl, etc. (2) ForceP “CP” The enterprise (reminder): 3 ¥ Language faculty, part of being human (genetically specific). TopP* ¥ Comprises a system which highly restricts the range of possible human languages. 3 ¥ This system underlies all human languages, we seek to discover its properties. FocP —Looking in depth at the syntactic properties of individual languages 3 —Comparing properties across languages to see what the options are TopP* ¥ Once we see patterns, we 3 —state generalizations FinP —form a hypothesis (what causes the generalization?) 3 —check other predictions of the hypothesis AgrSP “IP” 3 —revise as needed TP • As the theoretical hypotheses develop, we see new things to check that we wouldn’t 3 have thought to check before. In the process, whether the hypotheses are right or NegP wrong, we have learned a lot about how language works. 3 ¥ We assume there is a fact of the matter, that the Truth is Out There. AgrOP We are just trying to figure out what it is. 3 ¥ Lots of proposals exist, some probably closer to the Truth than others. vP “VP” ¥ To evaluate them, we must see what they predict: 3 —Do they predict that all of the possible languages/constructions exist? VP English sentences must be SVO. I left. Taxes I can deal with, it’s the forms that bother me. According to many, the road to (2) began in 1989… —Do they predict that all of the impossible languages/constructions don’t exist? English sentences may have any word order at all. Pollock, Jean-Yves (1989). Verb movement, universal grammar, and the structure *Like I pizza. *Bill thinks Mary that left. of IP. Linguistic Inquiry 20(3):365Ð424. ¥ When two hypotheses both seem to predict the data, we get into aesthetics. —Which hypotheses are “simpler”, more “elegant”? Basic tasks: ¥ Present data showing that there needs to be another position in the clause —Which hypotheses make the greatest range of predictions? structure between negation and adverbs (named: AgrP). ¥ Explain the special behavior of have and be both in French and in English. (1) CP The clause structure of “the olden days” (say, 1986): • (The paper contains a couple of other points concerning NegP as well…) 3 IP CP: force (declarative, interrogative) 1. French: V moves to I; English: have/be move to I (I lowers to other verbs). 3 IP: functional morphology (tense, agreement) VP VP: lexical/thematic elements Assume: French and English share a common D-structure. Adverbs and not can’t move to the right. But this doesn’t seem to be quite good enough to describe/explain the data—progressively, they have been split up into further functional projections. (3) IP beaucoup ‘lots’ moves optionally. 3 subject I′ We use adverbs to see where (8) a. Pierre a lu beaucoup de livres. 3 the left edge of VP is. I’ll ‘Pierre has read lots of books.’ I (Neg) take them to be adjoined to 3 VP for now, like what b. Pierre a beaucoup lu de livres. not/pas VP happens to quantifiers c. Pierre lit beaucoup de livres. 3 undergoing QR. Cf. rien (adverb) VP below. # d. * Pierre beaucoup lit de livres. ... V ... Conclusion: All (finite) verbs move to I in French; English finite have and be do too. (4) a. * John likes not Mary. b. Jean (n’) aime pas Marie. 2. (Some) infinitives (can) move in French.

(5) a. * John kisses often Mary. Assume: ¥ Finite and non-finite sentences are the same except for [±finite] on I. b. Jean embrasse souvent Marie. ¥ not and ne…pas stand in the same structural position in tensed clauses and c. John often kisses Mary. d. * Jean souvent embrasse Marie. infinitives.

French negative object rien ‘nothing/anything’—must move, marks left edge of VP. (9) a. Ne pas être heureux est une condition pour écrire des romans. ‘Ne to not be happy is a prerequisite for writing novels.’ (6) VP b. N’être pas heureux est une condition pour écrire des romans. 3 ‘Ne to be not happy...’ rieni VP # c. Ne pas avoir eu d’enfance heureuse est une condition pour écrire des ... V ti romans. ‘Ne not to have had a happy childhood is a prerequisite for writing novels.’ (7) a. Pierre n’a rien mangé. d. N’avoir pas eu d’enfance heureuse est une condition pour écrire des Pierre ne has nothing eaten romans. ‘Pierre hasn’t eaten anything.’ ‘Ne to have not had a happy childhood...’ b. * Pierre n’a mangé rien. Conclude: Verb movement to I is optional for infinitives in French. c. Pierre ne mange rien. Pierre ne eats nothing But wait… ‘Pierre doesn’t eat anything.’ d. * Pierre ne rien mange. (10) a. Ne pas sembler heureux est une condition pour écrire des romans. ‘Ne not to seem happy is a prerequisite for writing novels.’ Ok, now infinitives can move, but only avoir and être, right? Predicts that lexical infinitives preceding adverbs should be ungrammatical b. * Ne sembler pas heureux est une condition pour écrire des romans. (‘Ne to sem not happy...’) But… c. Ne pas regarder la télévision consolide l’esprit critique. ‘Ne not to watch television strengthens one’s independence.’ (15) a. Paraître souvent triste pendant son voyage de noce, c’est rare. ‘To look often sad...’ d. * Ne regarder pas la télévision consolide l’esprit critique. (‘Ne to watch not television...’) b. Oublier presque son nom, ça n’arrive pas fréquemment. ‘To forget almost one’s name...’ So: Only infinitives of have (avoir) and be (être) can (optionally) move in French. Lexical infinitives cannot move in French. So: Any old infinitive verb seems to be able to move past adverbs. But: Remember from before, only avoir and être can get past negation. Quelle coïncidence! Both English and French have restrictions on moving lexical verbs, but not on moving have/avoir and be/être. (16) a. * Ne paraître pas triste pendant son voyage de noce, c’est normal. ‘Ne to look not sad...’

(11) a. Not to be happy is a prerequisite for writing novels. b. * N’oublier pas son nom, ce n’est pas un explot. b. ? To be not happy is a prerequisite for writing novels. ‘Ne to forget not one’s name isn’t worth writing home about.’ c. Not to have had a happy childhood is a prerequisite for writing novels. d. (?) To have not had a happy childhood is a prerequisite for writing novels. Summary of the puzzle. * (12) a. Not to seem happy is a prerequisite for writing novels. French: V[+fin] pas adverb _ English: V[+fin]i not adverb _ b. * To seem not happy is a prerequisite for writing novels. d. Not to get arrested under such circumstances is a miracle. not adverb V[+fin]‡ e. * To get not arrested under such circumstances is a miracle. V[Ðfin]* pas adverb _ pas V[Ðfin] adverb _ not V[Ðfin]* adverb _ 3. Verb movement to I isn’t really always verb movement to I. pas adverb V[Ðfin] not adverb V[Ðfin]

Verb movement for French infinitives is never obligatory. *Only have/avoir and be/être. Predicts that we can see adverb verb NP sequences (never allowed with finite verbs). ‡Lexical verbs (verbs other that have/avoir and be/être). Et voilà. French: All finite verbs raise past pas. (13) a. Souvent paraître triste pendant son voyage de noce, c’est rare. All nonfinite verbs can raise past adverbs. often to.look sad during one’s honeymoon that’s rare. Only nonfinite avoir/être can raise past pas. ‘To often look sad during one’s honeymood is rare.’ English: Only finite have/be can raise past not. b. Preseque oublier son nom, ça n’arrive pas fréquemment. Only nonfinite have/be can raise past adverbs. (see below) almost to.forget one’s name that ne happens not frequently. ‘To almost forget one’s name doesn’t happy frequently.’

(14) * Pierre presque oublie son nom. Pierre almost forgets his name. 4. The Split-INFL hypothesis: Introducing AgrP. (20) IP 3 ′ Hypothesis: NP I There is a position between negation and the adverbs at the left edge of VP 3 I where French verbs move to. This is “short” verb movement. Not all the way to I. 3 (pas/not) AgrP 3 French: Short verb movement is not lexically restricted (to être and avoir)— Agr VP short movement any old nonfinite verb can undergo short verb movement. 3 (adv) V (17) a. To often look sad during one’s honeymoon is rare. b. To almost forget one’s name doesn’t happen frequently.

(18) a. * To look often sad during one’s honeymoon is rare. *HMC b. * To forget almost one’s name doesn’t happen frequently.

(19) a. I believe John to often be sarcastic. b. I believe John to often sound sarcastic. 5. Verb movement and θ-theory—what’s special about have and be? c. (?) I believe John to be often sarcastic. d. * I believe John to sound often sarcastic. What differentiates have and be from other verbs? Pollock speculates… Conclude: English short verb movement is lexically restricted to have and be. Back in 1989, people did not generally believe that the subject started in SpecVP—rather, So, we have a correlation: the assumption was that the subject started (and usually stayed) in SpecIP. The subject needs to get a θ-role, and the verb provides it with one, directly assigned to SpecIP. English: Short verb movement of infinitives—only have and be. V
I of finite verbs—only have and be. What’s the difference between V
Agr/T and affix hopping? One adjoins V to Agr/T, the other adjoins Agr/T to V: French: Short verb movement of infinitives—any old verb. V
I of finite verbs—any old verb. (21) a. V b. Agr 33 We can say this like this: V
I is lexically restricted iff short verb movement is. V Agri Vi Agr

This would be predicted if V
I contained an instance of short verb movement: Pollock’s idea:Well, V needs to assign θ-roles. In (21b), V is encased inside Agr. In some That is, V
I is not a one-step jump, it’s short movement to a middle position, and then a languages, V can “see out” of this configuration and assign its θ-roles, but not in others. movement from there to I. Pollock’s hypothesis: French Agr is “rich” enough to transmit V’s θ-role. In French, (21b) can assign a θ-role. French Agr is “transparent”

English Agr is “not rich” enough to transmit V’s θ-role. In English, (21b) cannot assign a θ-role. English Agr is “opaque” So, if V has a θ-role to assign, and Agr is opaque (English), V can’t move to Agr. (25) a. Legg-eva-no Italian (if V doesn’t have a θ-role to assign, V can happily move to Agr: have, be) ‘they read (read-imperfect-3pl).’ b. Parl-er-ò And, if Agr is transparent (French), no problem, move away. ‘I will speak (speak-fut-1sg).’ What about the French restriction on nonfinite lexical verb movement to I? (26) a. ...werk-t-en Dutch Pollock’s hypothesis here: [Ðfinite] tense is (perhaps universally) opaque. (they) work-past-3pl b. ...werk-t-e Summary: (she) work-past-3sg ¥ There is a head position between negation and the VP (where adverbs are). ¥ Pollock proposes this is Agr (hence, AgrP). Conclude: In general, tense is closer to the verb root than subject agreement. • “IP” is more perspicuously renamed “TP” (TenseP) now, since it just has tense features. • Agr can differ in “opacity”—transparent Agr (rich inflection) allows θ-transmission. If the finite verb moves to one and then the other, it would look like this: ¥ [Ðfinite] T is always opaque. (27) F1P (28) F1P 1 1 Belletti, Adriana (1990). Generalized verb movement. Torino: Rosenberg & Sellier F1′ F1′ 3 3 Jean-Yves is right, we need to split up INFL. But it seems that he didn’t split it up right… F1 F2P [V+F2]+F1 F2P Mirror Principle. Morphological derivations (suffixation, prefixation, etc.) 1 1 ′ ′ directly reflect syntactic derivations... (Baker 1985, 1988) F2 F2 3 3 Practically speaking, this is the hypothesis that head-movement always adds morphology V+F2 VP t VP at the outer edge of the moving (complex) head. 1 1 V′ V′ (22) a. X … Y … Z 3 3 b. X … [Zi+Y] … ti t ... t ... c. [[Zi+Y]j+X] … tj … ti So, in [V+F2]+F1, F2 is closer to the verb and is the head that was moved to first. That is, something that is closer to the head morphologically was closer to the head in the F2 is the lower head. F2 is closer to the verb. The tense affix is closer to the verb. F2 = T. structure. TP is below AgrP. (23) NP V T Agr French a. je parl ai s So, we should really have clauses that look like this: b. tu parl ai s c. il parl ai t (29) CP speak PAST person+number 3 AgrP (24) NP V T Agr 3 a. je parl er ai TP b. tu parl er as 3 c. il parl er a VP speak FUT person+unmber Chomsky, Noam (1989). Some notes on economy of derivation and representation. Reprinted in (1995) The Minimalist Program, MIT Press. We won’t really pay too much attention to the details of anything except GB and MP, with some minor discussion of a GB/MP variant in terms of Optimality Theory. You’re both right. Pollock’s AgrP is actually object agreement. Belletti’s AgrP is subject agreement. Scope and NPIs (30) AgrSP 3 Scope and negative polarity items. subject AgrS′ 3 The term scope in syntax is generally used to mean c-command domain. The subject in AgrS TP SpecIP (or, for that matter, in SpecVP) c-commands the object, so the object is in the 3 scope of the subject. T (NegP) 3 Neg AgrOP Negative Polarity items are words which seem (at first approximation) to require negation: 3 AgrO VP (31) a. I didn’t see anybody. rU b. *I saw anybody. (adverb) VP Moreover, they seem to require a c-commanding negation. A little bit of history (32) a. I told Bill that I didn’t see anybody. This kind of generative grammar has progressed through several stages, sometimes names. b. *I told anybody that I didn’t see Bill. Aspects Standard Theory The subject position c-commands the standard position of negation, hence: Extended Standard Theory (EST) Revised, Extended Standard Theory (REST) (33) a. *Anybody didn’t see Bill. Government & Binding (GB) Lectures on Government and Binding (1981) The way people often think about this is that an NPI is licensed (allowed to be there) by a Barriers (1986) c-commanding negation. (Kayne 1994: Antisymmetry) Minimalist Program (MP) (sort of) This is close, but it’s actually not quite that simple, incidentally.. A Minimalist Program for Linguistic Theory (1993) Chapter Four (1995) (34) a. Did you see anybody? Minimalist Inquiries: The Framework (1998) b. Nobody saw anybody. Derivation by Phase (1999) c. If you see anybody, let me know. Beyond Explanatory Adequacy (2001) (35) a. He didn’t lift a finger to help me. and has had offshoots along the way (again, sort of) b. #He lifted a finger to help me. Lexical Functional Grammar (LFG) c. Did he lift a finger to help me? Head-driven Phrase Structure Grammar (HPSG) d. Nobody lifted a finger to help me. Relational Grammar (RG) e. #?If he lifts a finger to help me, I’ll be very surprised. At least English NPIs seem to be licensed also in conditionals, yes-no questions. A-positions and A-chains

The c-command facts also get kind of murky, since for this purpose, it actually seems like An A-position is an Argument Position, which generally refers to a place in the tree where specifiers can “c-command out”… This remains kind of a mystery. an argument (that gets a θ-role) can be found. Complement of V (object position), SpecVP or SpecIP (subject position). Non-A-positions are sometimes called A′-positions. (36) a. Nobody saw anybody. b. Nobody’s mother saw anybody. There’s also a distinction between movement of something into an A-position and into an A′-position. Movement into an A-position is generally movement into SpecIP (raising, Many interesting discussions out there about how to unify the environments where NPIs passive, unaccusatives). Movement into an A′-position are other movements, like wh- are allowed. movement, QR. (These are usually terms reserved for XP movement, not head movement). Incidentally, somebody seems to be a PPI (positive polarity item), so it goes both ways. A chain is basically just a list of the positions in which an element has been in the tree Somebody can be interpreted either as a specific person or not. The non-specific kind is the over the course of the derivation. one which seems to be a PPI. It seems like movement into an A-position (A-movement, resulting in an A-chain) and (37) a. Somebody has been here. movement into an A′-position (A′-movement, resulting in an A′-chain) have somewhat b. I saw somebody yesterday. different properties. (38) a. I didn’t see somebody yesterday. (specific only) For example, one of the phenomena that is supposed to be sensitive to the distinction are b Nobody saw somebody. parasitic gaps.

(41) a. Which papers did you file t without reading pg? Adverbs b. *These apples were eaten t without washing pg.

There are at least two kinds of adverbs, those that seem to be attached roughly around IP, and those that seem to be attached roughly around VP. Split VP, vP, and the position of AgrOP (39) a. The tornado probably ruined George. As the theory progressed, people split up VP as well, introducing a “light verb” (usually b. George probably has been ruined by the tornado. labeled v) in whose specifier the subject starts (so the subject is still “VP internal” in the c. *George is being probably ruined by the tornado. sense the vP counts as the VP, but it is external to the projection that has the object in it). (40) a. The tornado completely ruined George. (42) vP b. *George completely has been ruined by the tornado. 3 c. George is being completely ruined by the tornado. subject v′ 3 We’ll have lots more to say about adverbs later, but it’s worth knowing about the two v VP different positions. Manner adverbs are generally VP-adverbs, temporal adverbs are 1 generally IP-adverbs V′ 3 V object

and it is usually assumed that V moves to v (perhaps absolutely invariably). This is a later view, that came along around the same time as the Minimalist Program.

One VP has been split, however, a question arises about where AgrOP is. AgrOP in the standard structure is just above the VP, but there is some evidence which points to the view that AgrOP is actually between the vP and the VP. The two competing structures look like this:

(43) a. vP b. AgrOP 33 v AgrOP AgrO vP 33 AgrO VP v VP

Deciding between them has proved to be very tricky. McCloskey addresses this conflict a little bit in his article.