* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download GRS LX 700 Language Acquisition and Linguistic Theory
Esperanto grammar wikipedia , lookup
Malay grammar wikipedia , lookup
Lithuanian grammar wikipedia , lookup
Old Irish grammar wikipedia , lookup
American Sign Language grammar wikipedia , lookup
Germanic strong verb wikipedia , lookup
Old English grammar wikipedia , lookup
Swedish grammar wikipedia , lookup
Scottish Gaelic grammar wikipedia , lookup
Udmurt grammar wikipedia , lookup
Ancient Greek grammar wikipedia , lookup
Modern Hebrew grammar wikipedia , lookup
Polish grammar wikipedia , lookup
French grammar wikipedia , lookup
Navajo grammar wikipedia , lookup
Japanese grammar wikipedia , lookup
Macedonian grammar wikipedia , lookup
Kannada grammar wikipedia , lookup
Hungarian verbs wikipedia , lookup
Sotho parts of speech wikipedia , lookup
Latin syntax wikipedia , lookup
Italian grammar wikipedia , lookup
Georgian grammar wikipedia , lookup
Chinese grammar wikipedia , lookup
Russian grammar wikipedia , lookup
Portuguese grammar wikipedia , lookup
Turkish grammar wikipedia , lookup
English clause syntax wikipedia , lookup
Lexical semantics wikipedia , lookup
Pipil grammar wikipedia , lookup
English grammar wikipedia , lookup
Yiddish grammar wikipedia , lookup
Serbo-Croatian grammar wikipedia , lookup
Kagoshima verb conjugations wikipedia , lookup
CAS LX 522 Syntax I Week 8a. Adjunction and head-movement Our model of grammar Recall our model of grammar: we select items from the lexicon, put them on the workbench, pick them up and Merge them until we have a single object, then pronounce and interpret the result. pronounce Lexicon Merge Workbench interpret Adjoin In addition to Merge, there is another thing that we can do with two objects we pick up from the workbench. The new operation is Adjoin, and it will enable us finally to draw proper structures for sentences with adjectives, adverbs, and modifying PPs: John quickly ate the scrumptious cake on the lawn. Adjoin The operations Merge and Adjoin are two different ways to combine two objects from the workbench. Merge takes two objects and creates a new object (with the label inherited from one of them). Adjoin attaches one object to the top of another VP one. VP V eat DP it AP quickly VP V eat DP it Adjoin I generally indicate adjunction with a “double branch” to keep it clear what is adjoined and what is not. The concept here is that the VP node has been “stretched out” and the AP has been hooked into it. The AP occupies a strange position in the tree. It is not a sister, nor a daughter of VP. It is sort of in-between. It’s not fully dominated by VP, it’s only dominated by part of VP. AP c-commands the VP, but the VP doesn’t c-command the AP, though we won’t dwell on that here. Merge establishes a mutual c-command relation between two objects, Adjoin establishes an asymmetrical ccommand relation. VP VP V eat DP it AP quickly VP V eat DP it Good, more terminology Adjoining an object to another results in a multisegment node. So, there are two VP segments below, making up together the VP node. Often you will also see these called “multi-segment categories”, but we’ll continue to call them segments and nodes. IP VP I AP quickly VP V eat DP it Good, more terminology X dominates Y in just the same conditions as before, except if X is a multi-segment node, all segments of X must dominate Y for X to dominate Y. X excludes Y if no segment of X dominates Y. AP is dominated by one segment of VP but not by the other. AP is not dominated by VP. AP is not excluded by VP. X c-commands Y iff: IP (i) X excludes Y, and (ii) Any node that dominates X also dominates Y. AP c-commands VP. VP does not c-command AP. VP I AP quickly VP V eat DP it Adjunction The main intuitive idea: adjuncts are “loosely connected” and general serve as modifiers. Adjuncts are generally optional (no q-roles in any q-grids). They seem to be able to attach either to the right or the left. They seem to attach to maximal projections. IP VP I AP quickly VP V eat DP it Adverbs Adverbs (at least many adverbs, we’ll look at more later) generally are adjoined to the VP. Pat quickly ate the sandwich. Pat ate the sandwich quickly. IP DP Pat IP DP Pat I I [+Past] AP quickly VP VP V eat DP the sandwich I I [+Past] V eat VP VP AP quickly DP the sandwich Adjectives Similarly, adjectives seem to adjoin to the NP. the tasty sandwich. Pat’s tasty sandwich. DP DP D the NP AP tasty NP sandwich DP Pat D D ’s NP AP tasty NP sandwich Mysteries to live with… (In English, anyway) adverbs can often appear on the right; adjectives seem to be restricted to the left. Adjectives seem to have a fixed order: The big red fluffy sock. *The fluffy red big sock. Adverbs do too. Pat often happily eats a tasty sandwich. ?Pat happily often eats a tasty sandwich. PPs serve the same function PPs often serve to modify the event like adverbs, and are adjoined in the same way (on the right). Pat ate the sandwich on the hill in the rain. IP DP Pat I I [+Past] VP V eat VP PP in the rain VP PP on the hill DP the sandwich PPs serve the same function PPs can also modify nouns, like adjectives (again on the right). Pat bought the book with the shiny cover. DP NP D the NP book PP with the shiny cover Motivation One difference between Merge and Adjoin is that Merge happens because it has to, Adjoin happens because it can. We can think of Merge as always happening to check some kind of complement or specifier features. Verb and object merge to satisfy a q-role (which we could cast as a complement feature) I merges with VP to check [Inf] (or maybe just [V]) complement feature. C merges with IP to check an [I] feature (how we can encode the fact that C takes IP) Complements vs. adjuncts PPs in particular seem to be freely reorderable when they are adjuncts. The book with a red cover by Radford from CUP The book with a red cover from CUP by Radford The book from CUP with a red cover by Radford The book from CUP by Radford with a red cover etc… But consider book of poems. Here, of poems is a fundamental property of the book, it’s not optional. A book of poems is a different sort of thing than a book of cartoons. Complements vs. adjuncts And notice that of books cannot reorder with other PPs, it always has to be first. You also can’t have two of this kind of PP: the book of poems by Radford with a red cover *the book by Radford of poems with a red cover *the book with a red cover of poems by Radford *the book with a red cover by Radford of poems. *the book of poems of cartoons with a red cover So how might we handle this? Complements vs. adjuncts Easy: Suppose that of poems is a complement. That puts it close to the noun (structurally, and thus semantically), and there’s only one complement. DP D the NP N book NP PP with a red cover NP PP by Radford PP of poems One-replacement Pat bought the book of poems by Radford with the red cover, and Tracy bought… the one with the blue cover the (book of poems by Radford) with the blue cover the one by Chomsky with the blue cover the (book of poems) by Chomsky with the blue cover *the one of cartoons by Chomsky with the blue cover *the (book) of cartoons by Chomsky with the blue cover So, what can one replace? One-replacement So, one can stand in for any maximal NP. Notice that this gives us evidence for: Of poems as a complement vs. with a red cover as an adjunct The structural reality of the segments. DP D the NP N book NP PP with a red cover NP PP by Radford PP of poems Another mystery to live with… You can also say: And you can’t say the linguistics book, the poetry book. *I want the linguistics book not the poetry one. Linguistics and poetry here are acting like the complements were. How might we think of these? Right. But how could we build that structure? Tricky. We’ll have to wait and ponder that a little while later. X-bar parameters Many (most? all?) languages of the world have something like a basic word order, an order in which words come in in “neutral” sentences. English: SVO Akira ate an apple. Japanese: SOV John wa ringo o tabeta. John top apple acc ate ‘John ate an apple.’ X-bar parameters These two word orders work nicely IP with X-bar theory DP as it stands; the I difference can be I VP stated in terms of a Akira -ed simple parameter DP V which differentiates eat an apple languages as to whether they are head-initial or head-final. IP DP John VP DP ringo o I I -ta V tabe X-bar parameters Notice that in English, both V and I are headIP initial, and in I Japanese, both V and DP I are head-final. In Akira I VP fact, languages tend -ed to be consistent in DP V eat their headedness: Japanese has postpositions, C comes after IP in embedded clauses… English has prepositions; C comes before IP in embedded clauses… an apple IP DP John VP DP ringo o I I -ta V tabe X-bar parameters There are also languages in which the basic word order is VOS, although they are few in number. Malagasy: VOS Nahita ny mpianatra ny vehivavay. saw the student the woman ‘The woman saw the student.’ See how we might generate an X-bar structure of this? X-bar parameters By changing the order of the specifier and the X, we can get VOS order, and by changing the order of both (with respect to English) we can get OVS order. I IP [PAST] Malagasy: VOS Nahita ny mpianatra ny vehivavay. saw the student the woman ‘The woman saw the student.’ DP I VP ny vehivavay V nahita DP ny mpianatra Kana yanïmno bïryekomo Fish caught boy ‘The boy caught a fish’ DP kana DP I VP Hixkaryana: OVS IP I [PAST] bïryekomo V yanïmno X-bar parameters So by changing the parameters of headcomplement order and specifier-X order we can generate the following basic word orders: SVO (spec-initial, head-initial) (English) SOV (spec-initial, head-final) (Japanese) VOS (spec-final, head-initial) (Malagasy) OVS (spec-final, head-final) (Hixkaryana) And that’s all… The problem of VSO languages There are quite a number of languages, however, for which the basic word order is VSO. Irish, Welsh, and Arabic are among them. Try as we might, there is no way to set the X-bar parameters to get VSO order—we have a specifier (the subject) between the verb and its complement. French French presents a similar problem; consider the English sentence I often eat apples. The adverb often is an adjunct, attached at VP, as seen here. IP DP I I I [PRES] VP VP AP often V eat DP apples French In French the sentence is Je mange souvent des pommes. I eat often of.the apples ‘I often eat apples.’ The adverb souvent appears between the verb and its complement; there is no place to put it in this tree. Moreover, it should be basically in the same place as in English, given the structural similarity and the sameness of meaning. IP DP I I I [PRES] VP VP AP often V eat DP apples Movement Consider English yes-no questions… To form a question from a statement like: We prepose the modal should to the front of the sentence, before the subject. Bill should eat his peas. Should Bill eat his peas? Where is should in this sentence? Movement There is one position in our sentence structures so far that is to the left of the subject, the one where the complementizer that goes (C): Should Bill eat his peas? I said that Bill should eat his peas. This is not where we expect should to be, though. It is, after all, a modal, of category I. It is not a complementizer. Also notice that if we embed this question, should stays after the subject, and if is in C: I wonder if Bill should eat his peas. Movement All of this suggests that the way to look at this is that we start with the sentence… Bill should eat his peas …as usual, and if we’re forming a yes-no question, we follow this up by moving should to the position of C. If we can’t move it (in an embedded question, there’s already something in C: if), it stays put. Movement Given that things do seem to move around in the sentence (that is, they start where we’d expect them to but we hear them somewhere else), this gives us a way we might “save X-bar theory” from Irish and French. Let’s go back and look at French with this in mind… French Jean mange souvent des IP pommes. Jean eats often of.the apples DP I ‘Jean often eat apples.’ If we suppose that the French Jean I VP sentence starts out just like the [PRES] English sentence, we have the VP AP underlying representation souvent PP V shown here. mange What needs to happen to get the correct surface word order? des pommes French Jean mange souvent des pommes. Jean eats often of.the apples ‘Jean often eat apples.’ Of course—the V (mange) moves up to the I position. This always happens in French with a tensed/agreeing verb. This generally doesn’t happen in English. Hence, the difference in “adverb position” (really, of course, it’s verb position) IP DP I Jean V+I VP mange+[PRES] VP AP souvent PP V mange des pommes What happens when V moves to I? To show that V attaches to I, but that I remains primary, this is drawn in the tree structure like this. IP DP I I We say that V head-adjoins (adjoins, head-to-head) to I. The head formed this way is sometimes called a complex head, (it’s an I with a V adjoined to it). Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? We should also consider what happens to the VP from which the V moved. This too is still a VP, it must still have a head. We notate the original location of the V by writing t (standing for “trace” left behind by the original V), and we co-index the V and trace to indicate their relationship. IP DP I I Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? Since the VP is still a VP, it still gets a [V] category feature projected up from its head. So the trace is still a verb. In fact, there’s no reason to suppose that any of the features of the original verb have been removed given that [V] is still there. We write it as t, but its content has not changed. IP DP I I Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? What has changed is that the original verb is now related to a higher position in the tree, and for many purposes, the top copy in the tree is considered to be primary. What we have created by moving the verb is a chain of positions in the tree that the verb has occupied. IP DP I I Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? When we think of moved elements in tree structures, we will often need to consider the chain of positions; this is usually written like: ( Vi , ti ) referring to the two positions held by Vi and ti in the structure here. IP DP I I Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? Using indices like that is kind of reminiscent of what we did when talking about Binding Theory— and it’s not a coincidence. A fundamental property of movement is that the moved element must bind (c-command, and be coindexed with) the trace in the original position: Movement is only upwards. IP DP I I Vi mange VP I [PRES] AP V ti VP PP What happens when V moves to I? Great. So does Vi c-command ti? IP DP X c-commands Y iff: (i) X excludes Y, and (ii) Any node that dominates X also dominates Y. I I Vi mange VP I [PRES] AP V ti VP PP Auxiliaries English has two auxiliary (“helping”) verbs have and be, which cannot serve as the main verbs of a sentence but generally serve to indicate differences in verbal aspect (progressive, past perfect, …). The auxiliary verbs often appear in I. Radford has had us up until now drawing them as if they exemplify the category I. But really, these auxiliary verbs are verbs, they just have special properties. Among these properties: they can move to I. Auxiliary verbs The reason we can’t assume the auxiliaries have and be are objects of category I is simple: I am not singing. I will not be singing. I will not have been singing. Rather, it looks like the topmost one moves to I, so long as nothing else is in I. A word on auxiliaries The underlying structure of a sentence with an auxiliary verb would be something like this, where the auxiliary verb heads a VP, and takes the main verb’s VP as its complement. IP DP I I -ed VP V have VP V eaten … A word on auxiliaries The underlying structure of a sentence with an auxiliary verb would be something like this, where the auxiliary verb heads a VP, and takes the main verb’s VP as its complement. IP DP I I Vi have VP I [PAST] V ti VP V eaten … Why does V move to I? Notice that if there is something in I already, like a modal, then even an auxiliary verb doesn’t move up to I. John might not be eating apples. And moreover, the verb has no tense inflection. This all suggests that the view that it is the affix in I which causes V to move to I. The verb is happy not to move, but will move when it can in order to help I out. The movement is a requirement on I, not on V. English yes-no questions Now, let’s go back and think about English yesno questions, which we took originally to be motivation that movement occurs. Bill will buy cheese. Will Bill buy cheese? What’s happening here? Well, we saw earlier that it is reasonable to think that the modal will, which starts out in I, moves to C in questions. Willi Bill ti buy cheese? English yes-no questions Why does this movement happen? By analogy with the motivation CP for V-to-I movement, we will take C to hold a special (this Ii+C IP will+Ø+Q time silent, or perhaps DP I prosodic) affix that must be Bill joined up with I. This affix is the I VP “question” morpheme, of ti category C, which we can write buy cheese as Ø+Q. Ø+Q Incidentally, lots of languages have an overt question morpheme, which adds plausibility to our assumption that English has a question morpheme in C that is just null. Akira ga hon o kaimasita ka? (Japanese) Akira top book acc bought Q ‘Did Akira buy the book?’ English yes-no questions Also notice that if there is an overt question morpheme there in English (which happens in embedded questions), there is no need to move I to C: I asked if Bill will buy cheese. *I asked (if) will Bill buy cheese. I to C In English, anything that would be in I moves to C. So, modals and auxiliaries all “invert” around the subject: Will Bill buy cheese? Is Bill buying cheese? Has Bill bought cheese? But main verbs never raise to I in English. Consider then: Did Bill buy cheese? I to C Did Bill buy cheese? Why is there a do there? Before, we only saw do in sentences with not, inserted because the tense affix couldn’t “reach” the verb, blocked by not. What seems to be the case is that if I moves to C (that is, the past tense suffix -ed in this case), it also gets too far away from the verb (now Bill is between the suffix and the verb), and Doinsertion is required.