*!*Ghomeshi, Jila and Diane Massam. 1994. To Appear, Linguistic Analysis
tải về 302.24 Kb.
|
| of why middles are hard to get for type 1 cases).
Finally, the fact that type 2 resultatives are common cross-linguistically follows from a combination of facts. First, the caused-motion construction, being directly based on simple expressoins such as those in XXX above, is ubiquidous in languages around the world. Secondly, the verbs of surface contact make exceptional candidates for appearance in the caused motion construction since they entail in their basic senses that an agent acts on a surface. Our world knowledge tells us that it is typical for agents to act on surfaces such that something is caused to move on or off the surface; \bigskip
Eat and drink are not quite so straightforward, since they do seem to be transitive and yet occur as Type \#1. they actually look more like the second class of cases. ! Two types of verbs: Simple intransitives (bark, ring, sleep) Can appear with general resultative: {The dog barked the neighbors awake. Or with directional (with caused motion interpretation): {the dog barked the man out of the house. In both cases, neither non-subject argument is supplied by the verb. Transitives with non-obligatory second argument General: eat, drink Can also appear with general resultative or with caused-motion: {They drank the teapot dry {She drank him under the table. In both of these cases, the non-subject arguments are all added. Notice the eaten/drunk thing does not generally make for a good location, and so this argument is not generally linked to a directional. To the extent that you can get a location interpretation, you can get the corresponding sentences: {She ate the middle out of the bread. where second argument is/can be surface, and can appear as location wash, rub, sweep, cut Without the second argument appearing overtly: {He cut himself free \\ Cinderella scrubbed herself silly. With the second argument appearing as location: {He cut the debris out of the tree \bigskip
! Type 2: Verbs of surface contact; typical objects are surfaces; result must be directional PP; Obj of the PP corresponds to normal direct object; Verbs may be obligatorily transitive; readily form middles; exist in all? languages. \eenumsentence{ Tracy washed the soap out of her shirt. Pat rubbed the oil into the wood. Terry swept the leaves off the sidewalk. ! Claim: these are ! lexically derived via template augmentation from activities; two different but related senses a la prototype theory; pure surface contact is assumed to be basic. >From my Chapter 2 (motivated by other facts): wipe: Profiled (obligatory) roles prefer direct argument slots. Only if there are three profiled roles can one appear as an oblique (e.g. put). Unprofiled (non-obligatory) roles can appear as obliques. \medskip Q: why can't the surface argument appear as an oblique in the intransitive motion construction? *She washed into the bowl. A: because the washer is not compatible with being a theme (since one
Q: why can't these transitive verbs appear as simple intransitive resultatives of type \#1, with entirely different patient and result arguments, since only the agentive argument is obligatory? Maybe they can do just that: He cut himself free. Cinderella scrubbed herself silly. Admittedly not all cases work (*Tracy washed the electric bill through the roof), but then resultatives are highly idiosyncratic (pg 192). ]
Lexically derived accomplishment must denote unitary core event of direct causation. Examples similar to mine (e.g. The cook cracked the eggs into the glass) are unusual since two results are intended simultaneously: there is only a single causal chain with two results. The result that is added must be construable as part of a prototypical event with the name of that activity (see also Kiparsky in Complex Predicates volume: "Simple predicates refer to single events (and consequently, simple causatives refer to direct causation). That is, removal of something from the surface can be construed as a direct result of surface contact and thus as part of the prototypical event of surface contact. [Something like this seems to be right, and although I'm starting to feel like George (That is in my diss!), I did make the point about direct causation w.r.t. resultatives in the diss and book. In any case, it's not clear how this ability for accomplishment verbs to sometimes allow directional phrases is consistant with their discussion of break on page 6: A verb like break is said to have an event structure of the most complex type available (accomplishment) and therefore "cannot be further augmented via template augmentation." pg 6 [A question that is raised is how intransitive break is accounted for since augmentation is supposed to be monotonic, and they're assuming break is basically transitive. Annie raised this problem for them to me in another context] \bigskip
[How do the following fit into the above categorization? \eenumsentence{ "Save yourself rich at Snyder's!" "Rainey gaveled them to order." They seem more like Type 1: they are not verbs of surface contact (at least a isn't); I bet they are not possible in all languages; they do not naturally form middles. BUT, the verbs involved are transitive like Type 2. Maybe "gavel" is special as a denominal verb... ] Extended verb meanings are built up monotonically by template augmentation. Principles (applying both to lexically formed and syntactically formed): Each subevent in the event structure template must be identified
So, e.g., activities cannot be augmented via template augmentation unless there is an additional predicate to identify the additional subevent. [How are ditransitives supposed to be formed, since there is no additional predicate to do the RECEIVE predicate augmentation? Also what corresponds to CAUSE?] There must be at least one argument XP in the syntax per subevent in the event structure template. [Can one argument do double duty? ] Each arg XP in the syntax must be associated with an identified
*!*Ingram, David and William Thompson. 1996. Early syntactic acquisition in German: Evidence for the Modal Hypothesis. Language 72:1 97-120. I&T argue against Poeppel and Wexler's 1993 claim that children possess Full Competence. P&W had claimed htat children around the age of 2 show evidence that finiteness, verb agreement and verb movement are all in place (based on 282 sentences from one child). I&T argue for the Lexical/Semantic Hypothesis: early learning is conservative and based on lexical and semantic categories. They argue that early use of infinitives is not evidence of syntax, but instead is a reflection of the children's semantic categorization of infinitives as modals (involving the semantics of "will, can, should"). [interesting category in light of Wierzbicka's work on English infinitives.] I&T argue that children are very good at mirroring the input they hear: "what they hear is what you get." pg 97. P&W do support Borer and Wexler 1987 who argue that nonhead movement and the formation of A chains matures and is not available at the earliest stage. P&W state that the child can either use a finite verb in V2 position or a nonfinite form in sentence-final position, claiming that the two are in free variation with no semantic correlate. P&W argue that early use of inflection is evidence that agreement is in place (AGR). I&T argue that early inflected forms are initially unanalyzed, not learnd as roots plus affixes (citing Ervin 1964 and others). ``Research has continally shown that children are very good at replicating the word orders they hear..More critical is an analysis of syntactic productivity." pg 101. "Modal Hypothesis: German children in their early stage of acquisition use infinitives as main verbs in sentences that contain a modal interpretation, i.e. that some activity will, can or should occur." pg 102 I&T look at the 3 youngest Grm speakers in CHILDES. P&W claim that sentences with final infinitives and no
Rizzi (1994) argues that functional projections are only optionally prouduced in early child language: that the child may generate clauses starting with constituent lower in the tree than CP, AgreSP, TP and VP. (children do not show root infinitives with questions). I&T argue that this is also not viable, since it is a purely syntactic account and the use of infinitives creates a modal interpretation. In fact root infinitives of the child's sort are used in Child Directed Speech: maybe as a way of avoiding complex constructions with disocntinuous predicates. Perhaps "they do not disappear from child language because of the maturation of grammatical principles but..decrease in frequency over time as a result of the improving production and processing capacities of children." 116 I&T cite Slobin 1982: `Each language poses the child with a different set of problems to solve in discovering the notions to be mapped and the means of mpaping" (137). Modal Hypothesis is confirmed (statistical generalization) *!*Ingram, David. 1989. First Language Acquisition:
Aux appears in declaratives before questions. In questions, it isn't always inverted. Bellugi (1967/1971) on Brown's Adam: said inversion occured in Y/N Q's before Wh Q's. what he can ride in?: At 3;6 Adam produced only 8 inverted wh-q's out of 30 (27\%); 3;11 changed to 87\%. Kuczaj & Maratsos (1975): study on one child: said the same thing. Ingram & Tyack (1979): 21 children between 2;0 and 3;11 placed into periods A through F by MLU. 225 questions were collected from each child. ``at all times the children's use of inversions for wh-questions was as great as that for Y/N questions." pg 460. Not clear why the descrepency. ``the use of uninverted Aux is not as extensive as implied in Bellugi's stages and that they may be restricted to certain contexts." 460.: Period A: 7 subjects: 22\% aux's in Y/N, 44\% in Wh; 55\% of those were inverted in Y/N; 77\% in Wh. Period B: 3 subjects: 68\% aux's in Y/N, 70\% in Wh; 81\% of those were inverted in Y/N; 91\% in Wh. Period C: 5 subjects: 89\% aux's in Y/N, 88\% in Wh; 91\% of those were inverted in Y/N; 96\% in Wh. Kuczaj & Brannick: cross-sectional study of several children: the inversion of Aux in wh-questions occurs gradually for specific wh-words. Children show inversion for what Q's while having inunverted why questions." ``...this syntactic rule is initially relatively specific." K&B 1979:43. Brown (1968): before MLU stage III, the child's questions are seen
By Stage III: claim is that child has both subj-aux inversion (witness Y/N Q's) and Wh-movement (witness Wh Q's), but can't apply both at once. (but can John do what? would be expected to occur and doesn't). Maratsos and Kuczaj also argue that children acquire rules very gradually, restricting them at first to very specific contexts. Kuczaj and Maratsos (1983) claim that children use aux in declaritives before they acquire them in Y/N questions. Ingram: ``their accounts do not work from a theory of grammar and do not deal with how the child goes from specific learning to general rules." pg 464. *!*Israel, Michael. 1994. CSDL talk on Way Three early usages: Go your Path Construction: someone going their ride, wending their gate, makes his course, going their way, Getting or Keeping possession of ap th (this usage shrivels up so that in modern usage only FIND is left) Take your way?, hold your way?, snatch your way?, Find your way, Created Path Usage (literal path creation) Cut, Clear, Pave, Smooth, Manner starts off with verbs of motion, expands to include verbs of laborious motino, expands to include verbs of windy motion Means starts off wth path clearing (verbs of cutting, fighting, enabling) Finally we get verbs indicating some incidental activity
Production Principle: Imitate Others Comprehension Principle: Accomodate Deviatn Usages and Capture Generalizations.
*!*Jackson, Dan. 1995. The historical origins of the ``that trace effect" \eenumsentence{ *Who do you think that saw you? Who do you think saw you? Dan traces the evolution of that from a demonstrative pronoun. He argues that the loss of inflections in English led speakers
The trend was to include that in \ex{0 if the intended interp. was that of a RC (and to leave it out if a sentential complement reading was intended). 1) The spread of that into RCs caused an increase in ambiguity between the RC and complement clause readings of sentences like \ex{0. 2) At the same time, there was pressure to include that before subject relatives to avoid processing problems: {I saw the dog bit the girl \\ I saw the dog that bit the girl. These combined to motivate the grammaticalization of always including that for subject relatives and always omitting it for subject complements. In the case of "that" complementizer with objects: {Who do you think (that) Kathy likes? We still sometimes have an ambiguity 1) above: {Who did you know that Kathy likes? \\ RC: who was it that you know, that also likes Kathy \\ complement: which person likes Kathy But 2) did not hold, since there was no processing difficulty with: {I saw the girl the dog bit. I encouraged him to take a look at other complementizers: \eenumsentence{ *Who did he wonder whether knew it? Who did he wonder whether knew it? *!*Jurafsky, Daniel. A probabilistic model of lexical and Syntactic Access and Disambiguation. Cognitive Science 1995? Read: Cacciari and Tabossi (1988). The comprehension of idioms. Journal of Memory and Languaeg 27:668-683. DJ proposes a single computational architecture that models both access and disambiguation. It is based on a preference or coherence in interpretation: constructions rae
Bayesian access or evidential access: phonological, syntactic and semantic evidence, top down and bottom up is integrated to determine which constructions should be accessed. Conditional probablilty of each interp given the input is computed. Constituent expections: phrase structure skeletons Valency information: associated with verbs and other predicates Computing parallel parses can be very complex, but there are ways to
The model disambiguates on line by pruning unlikely interpretations. Construction Grammar is augmented with probabilities. Construction Grammar emphasizes the role of memory: there are many structures. Generalizations across constructions are captured via a type hierarchy., CORRELATIVE CONDITIONAL(pg 9): The more you look at it, the more unusual it gets. Every construction is annotated with a simple probability, estimated
To compute the probablity of NP -> Det N, the Treebank corpus was consulted to get a frequency for all NPs (52,627), and then for those NPs which consist of a Det and an N (33,108). The conditional probablity is 33,108/52,627 or .63. Serial models of lexical and idiom access have fallen out of favor because of Swinney (1979) and Swinney and Cutler (1979) evidence for parallelism. Syntactic access has been believed to be serial because of garden path phenomena, but DJ argues that this too can be treated by a parallel model, as long as you introduce some kind of pruning mechanism. He cites (Kurtzman 1985, Gorrell 1987, 1989; Boland 1991 and MacDonald 1993) as providing evidence for parallelism in syntactic processing. He goes through various models for lexical access and ends up arguing for one that seems to me like Bates and MacWhinney's competition model, although more explicit. Cacciari and Tabossi (1988) found htat different idimos are accessed at different rates. E.g., idioms are often not accessed until more than one word has been seen. DJ: Bayesian or evidential access algorithm. Conditional probabilities are computed for cosntruction given the evidence. Evidence can be syntacitc, semantic, and lexical, both top down and bottom up. Beam-search is used: every construction falling within a certain percentage of the most highly ranked cosntruction is accessed. (pg 15) Cacciari and Tabossi data is natural because phonetic evidence suggests both a lexical entry and an idiom, with hte leixcal entry being much more probable. A pruning algorithm may keep the idiom from being accessed. Lexical items with higher frequency will tend to have higher probability. MOre familiar idioms are accessed more quickly (d'Arcais 1993). Access of a cosntruction will be inversely proportional
She argues that fundamental changes take place in language development after 5, even within seemingly simple categories like determiners. Also, pronomialization and subordination change their function as the child moves from using them at the intrasentential level to the intersentential level. Carol Chomsky 1969 argued that children first use linguistic generalizations which hold for a great number of constructions in a language (e.g. subject is equivalent to logical subject; word order is canonical order; subject of a copmlemnt verb is hte closest NP preceding it). Others rebutted these claims though. ``I should like to submit that talented investigators will always be able to devise novel techniques and contexts which make it possible to demonstrate that a particular structure is understood either much earlier or much later than previous work had indicated. But this tells us relatively little about the general nature of developmental changes or about the function a given category may have for the child." pg 459. *!*Kay, Paul. An Informal Sketch of a Formal Architecture for Construction Grammar. draft. UCB Hi Paul,
A couple of quick comments: 1) you might want to explain what "reentrancy" is so it's clear how CG differs from HPSG. Constituency is separable in CG, but it's not necessarily separated as it is in LFG, is it? I mean, a construction could refer to a particular tree configuration and grammatical relations at the same time, no? Aren't inversion constructions or various (subject and non-subject) question constructions examples of that? 2) does LFG still use constraint equations? I thought they did away with them several years ago (because they feared they were not monotonic). It's possible they've been brought back... 3) you seem to downplay the significance of the distinction between LFG and CG in that CG routinely attaches semantic/pragmatic factors to the constructions. Is that the soul and reason d'etre for CG, though? 4) also, LFG, or at least Joan, doesn't seem quite sure how to handle constructions, although she sees the need for them in certain cases. Her work has focused much more on lexical items (as the name suggests). 5) I didn't quite understand what your sentence about "CG admits only actual words, phrases and sentences as elements of the modeling domain". Do you mean as opposed to traces? morphemes? *!*King, Jonathan and Marta Kutas. 1994.ms. A Brain Potential whose latency indexes the length and frequency of words. Bradley, Garrett and Zurif 1980: normal subjects show frequency effects in their RTs to OC but not to CC items, whereas Broca's aphasics show freqency effects to both word types. Gordon and Caramazza 1985: failed to replicate: there are technical difficulties with estimating freqency effects for very fast RTs. Van Petten and Kutas 1991: ERP studies: effects of word frequency and length are subtle, but there are larger differences between OC and CC items: OC items generate an N400 response dependent on cloze probability, repetition, and to some extent, frequency CC words do not generate an N400 response (``see King and Kutas in press for a notable exception"?) --$>$ Neville, Mills and Lawson 1992: CC items show a N280 response near Broca's area K&K note that less remarked is the fact that there is also a negative peak in same area somewhat later for OC items "N410" (Neville, Kutas, et al. 1986) K&K suggest that the N280 and N410 may be identical, that the latency is affected by lexical factors: frequency and length. OC classes: Nouns, Verbs, Adjectives CC: infinitival "to", Definite Articles, INdefinite Articles, NP Prepositions ("of"), VP Prepositions ("for"), Conjunctions, forms of BE RESULTS:
OC: N400 response and a N320 (the "N410 of Neville et al.": it may have been later in that experiment because words were presented vertically) CC: N280 and a broader negativity that is most obvious at later time points and continues to the point of overlapping with the ERP tohe following word (the N400-700): K&K in press have argued that this is due to the CC role in starting up constituents. In fact, they find the negativitity at 270 for the shortest and most frequent CC words, 300 msec for longer and less frequent CC words and 320 for longer and less frequent OC words They would like to argue that the N280 and the "N410" (actually N320) are really Lexical Processing Negativity (LPN). They suggest that the LPN effect becomes magnified by the time it can be used for RT experiments because freqency effects may impact processing at multiple points of time and the effects might be cumulative. What functional role does the LPN play during reading? It cuold have to do with syntactic processing (they allude to MacDonald's finding that frequency effects play a role in syntactic processing as well). But more likely, they argue is that it has to do with non-syntactic processes such as the control of eye movements. There is evidence that the area just above Broca's area, (where they actually find the LPN) is related to gaze control. Poor readers show changes in the shape and latency of the LPN. *!*Kiparsky, Paul. 1996?. Remarks on Denominal Verbs. Complex Predicates volume ed by A. Alsina, J. Bresnan, and P. Sells Contrasts his (and Bierwisch and Wunderlich's) semantic approach to constraints on word meaning to Hale and Keyser's syntactic constraints. Good paper for a class. Assumes decomposition into primitive elts with fixed combinatorial rules: Semantic Form (distinct from conceptual knowledge and Syntax, although interacts with both) Syntactic properties are predictable from word's meanings (cites Tenny, Krifka, Pinon, Condoravdi, Wechsler, Dor). Discussion of paint (must be intentional, more or less) Denominal causative verbs refer to generically intentional activies
"generically intentional activities" can be unintentional if performed while sleeping, walking etc. K gives examples like: \eenumsentence{ unintentionally painting the light switch an explosion painting the sky red ``In addition, lexicalized conventions of personification allow particular kinds of non-intentional agents to be assimilated to intentional ones": \eenumsentence{ but not Grm a disease kills people Grm but not Eng: a book wanting to be understood Direct causation is not "straightforward" (you can paint your house by having painters do it). K claims this is a language specific fact, since Finnish "build" cannot be used if builders built your house [see Goldberg 168-169] POINT: Conceptual knowledge is essential to the formation of lexical meaning. Bierwisch and Wunderlich draw on Formal and Cognitive (Jackendoff's) semantics. Some nice arguments against H&K are given. E.g. H&K claim that purely syntactic principles block {I gave the bush some fertilizer $\rightarrow$ *I bushed some fertilizer Principle: IOs may not be incorporated. Kiparsky: but why couldn't this use of *bush be derived from {I put some fertilizer on the bush Principle: [3.] If an action is named after a thing, it involves a
[But, with additional context, "canonical use" can change, e.g. Clark and Clark's teakettle?] E.g. to take means to use tape in any canonical way: to fasten, tie, bind, cover, support, record, or measure with tape. It cannot be used to mean to use tape as a paperweight [well, it can in the right context] It is not a canonical use of bushes to put fertilizer on them. LOcatum verbs: putting x in y is a canonical use of x LOcation verbs: putting x in y is a canonical use of y (not true of beach and sand: footnote 9) pg 10 Objects which have both canonical uses--to be put on something and to have something put on it-the denominal verb has both meanings: shelve, ice, index, string, tube. \bigskip
True vs Apparent denominal verbs: Real denominal verbs: 1. to box a present in a gift box (\# in a paper bag) to pocket change in one's coat pocket (\# handbag) to fence an area with barbed wire (\#with a mine strip) to star a stence (\# with a question mark) 2.
being shelf-like thing or paint-like thing. to shelve a book on a windowsill to paint an inflmaed throat with iodine K claims this follows from his principle 10 (3 above: canonical use) Apparent denominal verbs (can differ from nominal in stress patterns)
to land a hydroplane on water to dump garbage by the roadside to blanket an area with advertizing string him up with a rope anchored the ship with a rock hammer the desk with his shoe No syncronic noun-to-verb derivational relation (although psosibly one in the other direction) pg 18 \medskip
not both together: Principle [4.] The lexicalization constarint: A verb can inherently express at most one semantic role (theme, instrument, direction, manner, path) K gives the example of climb (citing Jackendoff). He claims it cannot mean both clambering (manner) and direction(upward) at the same time. But: {The skiier climbed the mountain (\#on a lift. \#to the bottom) Jackendoff had claimed, like Fillmore, that either one or the other OR BOTH meaning components could occur: not niether. K gives another example:tow, [with a truck? applied to a car?] \bigskip
Why are some verbs obligatorilly transitive? ANS: Because the agent is constitutive of the event, and [5] Constitutive arguments are not omissible An argument of a predicate is constitutive if it must participate in the entire event in a particular way. In push, the pusher is constititive, but not the thing-pushed (which need not move, or be affected in any other way). In roll, it is hte thing-rolled which is the constitutive arg not the rolelr (who need only initiate the event). [BUT, eat] Footnote 19: ``it is porbably not the case that all fixed transitives involve such constitive Agents, cf. such verbs as activate, free, blind, embitter, embolden. *!*Jackendoff, Ray. ms. (1994) Lexical Insertion in a Post-Minimalist Theory of Grammar. Lexical Insertion: X0 categories listed in the lexicon are projected into X-bar phrase structures J gives arguemnts that Phonological structure cannot be derived from Syntactic structure: there are various mismatches. ``Whatever we know about this [conceptual-intentional] system, we know it's not built out ofnouns and verbs and adjectives;" pg 4. ``We can say that conceptual structure interprets syntactic structure, or conversely that syntactic structure expresses conceptual structure." pg 4 ``I agree with Chomsky that conceptual structure is not a part of language...On the other hand, the SS-CS interface is part of language." pg 6 Representaional Modularity: ``the informational architecture of the mind segregates phonological, syntactic, and conceptual representations from each other. There can be no 'mixed' representations. Rather all coordination among these representations is encoded in correspondence rules... this claim is an explanded version of the traditional hypothesis of the autonomy of syntax, namely the idea that syntactic rules have no access to nonsyntactic features except at an interface" pg 8-9 Analogy to Autolexical syntax, and Tree-Adjunction Grammar. Lexical insertion acts to insert unanalyzed lexical items into
``..a lexical item is by its very nature a 'mixed' representation -- a $<$ PS, SS, CS $>$ tripel. THerefore it cannot be inserted at any stage of a syntactic derivation without producing an offending mixed representation." pg 11 Representatinoal Modularity view: 3 structures: phonological represnetation of /the cat/; syntactic NP represenation, CS {\small THING representation. ``These three structures do not just exist side by side. They are linked with each other by subscripted indices...the Det corresponds to the definiteness feature, the N corresponds to the type-constituent, and the whole NP corresponds to the whole {\small THING constituent." pg 12 ``...the lexicon as a whole is to be regarded as part of the interface component." pg 13 This is relevantly like HPSG, TAG and Congram. J argues that the phonological and sematnic interfaces are both at S-structure *!*Lakoff, George. 1986. Frame Semantic Control of
Ross (1967): {What did Harry go to the store and buy Lakoff (earlier) and Goldsmith (1985) argued that this was not a regular and. Instead, it was either a while or an in order to. Lakoff argues instead that it is a true conjunction: \eenumsentence{ What did he go to the store, buy, load in his car, drive home and unload. How many courses can you take for credit, still remain sane, adn get all A's in? Across=the-board extraction out of 2, 3, 5 or 1, 3 conjuncts. Cf. also comma intonation before and. \eenumsentence{ Sam is not the sort of guy you can just sit there, listen to, and not want to punch in the nose. This is the kind of brandy that you can sip after dinner, watch TV for a while, sip some more of, work a bit, finish off, go to bed, and still feel fine in the morning. Can extract out of questions, RCs, RNR. Type A scenerio: natural, expected course of events Type B scenerio: conventionalized expectation is violated
In A scenerios, the conjuncts which are "background states" do not need to involve extraction: {What problem did he sit there for a while and give up on? ("sit there for a while" is a background state) Type C scenerio: cause-result: {That's the stuff the guys in the Caucasus drink and live t be a hundred. He argues against syntactic explanation since frames are clearly involved. No semantic filter can help, since it can't make an ungrammatical sentence into a grammatical one. It is not a performance errror: these are not mistakes. \begin{quote Predication Principle: In a construction consisting of an isoltaed element and a structure expressing a propositional function, the propositional function is predicated of the referent of the isolated element {quote {
In \ex{0, John hit [] is a propositional function predicated of who. In \ex{1, John hit Bill and kicked [] is not a predication of who. {*Who did John hit Bill and kick? The first conjunct is not relevant. Tentative Constraints: A simple propositional function is a predication. If each member of a conjunction of predications can
Rules out: {*What kind of a sandwich did John make and eat an apple? because "[] eat an apple" cannot be predicated of "what kind of a sandwich" A conjunction of predicatiosn not meeting (II) is a predication if it is structured by a type A or B scenario. IN a type A scenario, the final conjunct must be prediated of the same elemnts as the conjoined predication. >From description of Deane (1991) article (in papers.new): There is a nice account of George's coordinate structure
\noindent ! Occur before the main event: \medskip \noindent Lakoff's A-scenerios [ ! preparatory action -- main action]: \eenumsentence{ He went to the store and bought something He came home and unloaded something. He picked up the phone and called someone He sat down and started typing a letter. This explains Lakoff's observation that A scenerios require extraction from final conjunct: final conjunct is main event. ! Scene-setters: conjuncts whose role is to describe the scene in which an event takes place: \eenumsentence{ Sam is not the sort of guy you can just sit there and listen to. Who did you stand in the parlor and tell jokes about? ! Internal Causes: \eenumsentence{ Which problem did he get bored and give up on? Who did he go berserk and start shooting at? Which part of your plan did he get confused and forget? \noindent ! Occur between conjuncts describing main events ! Incidental Events
\noindent ! Occur after main event \medskip
{How much can you drink and still stay sober ! Result conjuncts (Lakoff's C scenerios) ! Claim None of the above are central to the main flow of the narrative --
Erteschik-Schir: extraction is only possible from within categories which may be construed as dominant (whether or not they are in fact intended to be dominant in the context of utterance). Kuno: only those constituents in a sentence that qualify as the topic of the sentence can undergo extraction processes. Takami: extraction occurs from the phrase that carries new or "more important information" (i.e. more focal rather than topical information). Deane synthesizes all three accounts, giving them a common cause. "Extraction consists in the establishment of a long-range grammatical relation. An obvious prerequisite to establishing a relation between two concepts is that one be paying attention to both concepts at the same time... But in long distance extraction, the two concepts to be linked are separated far enough from one another that some means must be provided to focus attention on both. And what means would be more natural than if the two concepts were ones which commanded attention anyway? [i.e. topic and focus]... If extraction is to occur, the extracted NP must be inherently topical and the matrix phrase intrinsically focal, both at the same time." (pg 35) *!*Labelle, Marie. 1990. Predication, WH-Movement, and teh Development of Relative Clauses. Langaueg AcquisitionI(1) 95-119. key words: language acquisition, RC, French She notes that while children use initial Wh words in questions, they don't front WH for RCs until around 6 years. French doesn't allow prep stradning: {L'homme a\` qui j'ai parle\' \\ the man to whom I talked {*L'homme que jai parl\'e a\` \\ The man that I talked to Kayne 1983: P is not a proper governor in French Children use two strategies for relativization:
Labelle makes the point that children ages 3-6 produce RCs with fronted wh words (particularly what) appropriately, despite the fact that it is not obligatory (that can be used): \eenumsentence{ the boy whats crying is her brother I see a dog what's white I got everything what you got (Menyuk 1969) What explains the difference between French and English children? She argues that the difference is based on ANALOGY to
Both cosntructions are produced by children as early as 2;0, long before the first RCs are produced. Rondal (1979): 2;6 pseudorelatives; 3;6-relatives without complementizer; 4;2-relatives with compelmentizer. Both constructions are predicative in character, introduced by que or qui and not by a WH word, learned very early and are very frequent. ``when children start producing RCs it is possible that htey take a predicative construction that is well established in their language and use it to restrict the range of an NP. It is reasonable to think that chidlren simply transfer the grmamar of the predicative and cleft constructions to restrictive RCs. *!*Lambrecht, Knud. 1984. Formulaicity, Frame Semantics, and Pragmatics in German Binomial Expressions. Language 60 4. 753-797. N und N
(note lack of determiners) Must be conjoined. Therefore cna't be produced by: NP --$>$ NP und NP Productive N and N are united into a single semantic complex ZZ Three types: Lexicalized and irreversible Novel but semanticaly motivated seantically unmotivatd but pragmatically constrained
*!*Lambrecht, Knud and Maria Polinsky. 1997. Typological Variation and Sentence-Focus Constructions. CLS 33. \begin{tabular{lcc Sentence Type & Arg in Focus & Predicate in Focus \\ Argument Focus & + & - \\ Predicate Focus & - & + \\ Sentence Focus + & + \\ {tabular SF(thetic) and PF(categorical) are distinguished by TOPICALITY PRESUPPOSITION: assumed status of a referent as a topic of current interest in the discourse. in AF: subj or some other constituent is in focus Why didn't she come to work today? \eenumsentence{ Her SON is responsible (AF) Her son had an ACCIDENT (PF) Her SON is sick (SF) SENTENCE FOCUS CONSTRUCTION: Sentence construction formally marked as expressing a pragmatically structured proposition in which both the subj and predicate are in focus. The focus domain is the sentence minus any topical non-subject arguments. has been called "news sentnece" "neutral desctription" "all new utterance" "thetic sentence" "event-reporting sentence" Given an SF construction, three questions arise: questions of FORM, questions of FUNCTION, *the question of hte relationship between the two* (what this paper is about). The PRINCIPLE OF PARADIGMATIC CONTRAST: SF constructions have the form they do because they are to be minimally distinct from corresponding PF constructions within a given language. SF marking crucially involves DETOPICALIZATION of the nominal constituent involved. How can the subj argument be coded in such a way that it will not be interpreted as a topic? 1) cancel formal properties conventinoally associated with topics 2) code the SF subject with features conventinoally associated with FOCUS
SF constructions tend to lack a formal opposition betwen subj and obj (tend not to have NP VP structure). They ignore distinction between SF and AF. PF: Truman DIED. TRUMAN DIED. (requires iconic focus accent on predicate;
SF: TRUMAN died. *TRUMAN DIED (can't have SF interpretation: lack of accent can be explained via reference to contrast with PF) They compare: \eenumsentence{ (I saw) a black BIRD \\ broad scope accent: does not require that speaker saw a black thing be presupposed. (I saw) a BLACK BIRD \\ if blackness of bird is unexpected or somehwo remarkable (I saw) a BLACK bird \\ usable only if speaker seeing a bird was presupposed, and the fact that it is a black one is asserted. \\ The accent is motivated ICONICALLY (I saw) a BLACKbird. \\ compound accent: motivated by the need to distinguish d from a: motivated by Paradigmatic contrast The fact that c and d have the same prosodic contour is an instance of prosodic homophony. \begin{tabular{llll ADJ + NOUN &&& NOUN + VERB \\ a. black BIRD (broad scope) & - & Johnson DIED & (PF) \\ b. BLACK BIRD (broad scope) & - & JOHSON DIED (PF) \\ c. BLACK bird (narrow scope) &-& JOHSON died (AF) \\ d. BLACKbird (broad scope) & - & JOHSON died (SF) {tabular Note both examples in d resist separation of parts: \eenumsentence{ *Black-as-sootbird *JOHNSON, after a short illness, died. Ways to neutralize Subj=Obj distinction crosslinguistically: English: prosodic inversion Italian: syntactic inversion: \eenumsentence{ si \`e rotta la MACCHINA \\ ``The CAR broke down" (SF) Ho rotto la MACCHINA \\ ``I broke the car" (PF) Subject in SF in a has position and accent of focal obj in b. Russian allows both syntactic inversion and prosodic inversion: \eenumsentence{ pticy POJUT (PF) \\ birds. NOM.PL sing.PRES.PL \\ ``The birds are SINGING pojut PTICY (SF) \\ sing.PRES.PL birds,NOM.PL \\ ``There are BIRDS singing PTICY pojut (SF) \\ PTICY pojut \\ ``The BIRDS are singing" Detopicalizatoin in SOV lngs: the SF subj may be placed in imediately preverbal (focus) position as in Jap and Kor; or the SF subj may be placed postverbally (Lat, Armenian, Chukchi). VSO: most verb initial languages use preverbal position for topic, so SVO order would have a PF interpretation. But if this is not possible as in Irish, SVO order can serve to mark SF as predicted. If SVO is used for topic comment, then the focus position available for objects is used (e.g. in Malagasy). In some languges, SF subj is morphologically incorporated into the verb (demonstrating the cohesion between subj and v) (Sasse on Boni): {a\'ddige\'ee~juudi \\father-may~died \\ "my father died" Since objects only are usually incorporated, the incorporation of the subject obeys their generalization. Suspended subj-verb agreement: Jespersen ``In Danish the verb was here put in the singular before a plural word...in Eng, there is hte same tendency to use there's before plurals..in Italian, too, one finds v'e\` instead of vi sono..." See chart on page 15 for cross-linguistic differences. *!*Lambrecht, Knud. 1994. BOOK. Chapter 1. Cambrdige University Press conversational pragmatics :interpretation of propositions in relation to conversational settngs, conversational implicature. lexical pragmatics: e.g. deixis discourse pragmatics: presupposition and assertion: structuring of propositions into portions which a speaker assumes an address already knows or does not yet know identifiability and activation: a speaker's assumptions about the status of mental representations of discourse referents in the addressee's mind at the time of an utterance topic and focus: speaker's assessment of the relative predictabilty vs unpredictabilty of the relations between propositions and their elements in the given discourse situations information structure packaging in the clause Allosentences: semantically equivalent but formally and pragmatically divergent sentence pairs such as active vs passive, canonical vs topicalized. Like the biologists question, "Why are there so many kinds of
L: syntax directly reflects communicative needs, but is not exhaustively accounted for in terms of its communicative function in discourse. My CAR broke down Mi si \'e rotta la MACHHINA (It.)\\ to me itself is broken the car J'ai ma VOITURE qui est en PANNE (Fr)\\ I have my car that is in breakdown Sentence focus: the domain of 'new information' extends over the entire proposition, including the subject; either introduces a new referent or announces an event involving a new referent (latter are, Event-reporting sentences) Focus Marker: sentence accent: formal indicator of the focus structure of the sentence My car is not the TOPIC. The speaker is the topic, "what the utterance is about" In all three lgs, pragmatically unmarked is SVO, pragmatically unmarked sentence accent is clause-final (or near final). Thus the unmarked information structure is: Topic-Focus. (Discussion of what it means to be marked: page 20ff: basically greater distributional freedom) Unmarked for Narrow Focus(contrastive reading): She likes Germans. Marked for Narrow Focus: It's Germans she likes. \medskip
(1) above contrasts with: {My car broke DOWN. which has predicate focus (everything except the subject is focused). \medskip
(2) above contrasts with: {La mia macchina si \'e rotta By placing the subject after the verb, Italian repsects the unmarked prosodic sequence in which the constituent carrying the main sentence accept occupies final position. Italian is able to maintain the topic constituent in its unmarked initial position, but at the expense of the unmarked, canonical syntax. That is, the unmarked prosodic sequence is maintained. \medskip
(3) is a clefted construction, which Fr makes extensive use of because of a powerful grammatical cosntraint against focused subjects.: Ma VOITURE est en panne is illformed, prosodically. \medskip
The grammatical patterns in 1-3 can only be understood
\medskip
The communicative needs do not directly determine the form, since such an extreme view would imply that sentences expressing similar propositional contents in similar communicative situations must have similar forms across languages, which is contradicted by the 3 examples above. MOTIVATION is what is important. *!*Lasersohn, Peter. ms 1996. Pragmatic Halos ``exactly" is said to tighten the pragmatic halo: allow less pragmatic "slack." ``The townspeople are asleep" is said to also have a pragmatic halo: true evenif not all of the townspeople are alseep ``Circular" is not scalar (doesn't occur with "very"), but it combines with "perfectly": "perfectly"'s job is to limit the amount of pragmatic slack. ``When exttreme precision is not requierd, people acept utterances which deviate in minor ways from teh truth. But the degree of deviation which is allowed is not determiend solely by the pragmatics of the situation of utterance, but in part by the appearance of particular words within the utterance itself. We might call thsee words SLACK REGULATORS." pg 4 The slack regulartors are designed to narrow the range of close to teh truth a sentence is. But the examples that are given and are clear are scalar. The non-scalar predicates are more difficult (slack operators by hypothesis can operate on non-scalar predicates.) ``close enough to the truth" will normally mean "close enough not to obscure pragmatically relevant details or distinctions" invokes Grice here. [certain words or phrases themselves have smaller halos than others,w ithout an overt slack regulators. "2:58" has a narrower halo than "3 o'clock," "199" has a narrower halo than "a couple of hundred" and generally pragmatic reference points have larger halos] [Halos are defined on both predicates and arguments and the effect is assumed to be the simple cross product. But the choice of predicate or argument can affect the size of teh halo of hte other. What counts as "children" differs depending on whether the predicate is "may go to R-rated movies unattended", "eat solid food", "shoudl be seen and not heard." Similarly, "It is warm in Urbana today" depends on normal Urbana weather, and normal weather for "today."] [ You also assume that predidcates are continuously defined on arguments (see Figure 1), but it's not clear what this would mean for non-scalar predicates.] [It's interesting to think of slack "loosening" cases. Could you extend the analysis to generics arguing that the construction acts as a slack enhancer? Is "approximately" a candidate for a slack enhancer, too?] In the last PP on page 15. I think you might have meant to refer to 43a not 46 in second sentence. *!*Leinonen, Marja. 1985. Impersonal Sentences in Finnish and Russian: Syntactic and Semantic Properties. Slavica Helsingiensia 3 Impersonal constructions: no agent implied or expressed; subjectless; no verbal agreement, or fixed 3rd person verbal agreement. [seems to be family resemblance: not all cases discussed have all three properties, although prototypes, e.g. weather-type expressions do] ``Existential" sentences: Partitive "subjects": don't control verb agreement. E.g. ``There came no complaints." ``...different syntactic constructions bringin nuances that are associated with the given `abstract grammatical sentnece models' (Russkaja grammatika 1980: 354)" pg 15. ``Linguists workign with Slavic languages have been notably prone to accept the idea of variant sentential structures as basic patterns, or models, in a language. .. Lately, G.A. Zolotova has propounded for Russian a semantic-syntactic classification of sentence patterns or models, each with its own syntactic structure and a core propositional meaning..." pg 20. [The "core meanings" that are given, however, are totally vague and seem designed to cover many different construction types: e.g. "subject of state" for 7 different syntactic patterns] Long list of various patterns and "core meanings" follows: pgs 20-24, followed by certain (meaningful?) generalizations: { state predicates with N-Poss, N-Dat, N-Ac Prep N quanitity with N-Gen existence or its negation with N-Gen, etc. Alludes to a derivational relation between certain sentences [it's becoming less and less clear what view of grammar this really is] State of a Location and an experiencer are the semantic prototyes for impersonal patterns (pg 27). Other cases include: -sja passives without agents (`it is planned to organize a club'); passive participial predicates + infinitive (`It is forbidden to smoke'); predicative adverbs + N-Acusative (`Traces are visible'); certain verbs expressing usualness, happening [raising to subject verbs?], etc. Interesting discussion of Russian passive(s). Here core meaning seems to be equated with verb class and case roles (pg 33). One core meaning is posited for two syntactic structures with two meanings (defined as ``sentence patterns: structural meanings with the same case roles but different orientations and aspectual/actional modifications). [This is definitely constructional; ``core" doesn't seem to be used consistently throughout] Pg 35: ``different syntactic case configurations... reflect a difference in meaning" ``if one verb may appear with different syntactic cases, these also reflect differences i the semantic interpretatoins, despite the core meaing remaining the same. Thatis, different orientations are different views on the situation, be it in terms of aspect or prototypical role properties, e.g. responsibility/control for the action." Restrictions on passives in Grm, Eng, Russian (Weiss & Girke 1978) *!*Levin & Rappaport. 1995. Unaccusativity A the Syntax-Lexical Semantisc interface. Chapter 6. The Problem of Locative Inversion. MIT Press. certain versb of appearence appear, arse, issue verbs of existence exist, flourish, thrive inherently directed motion came, will go had arrived existence involving spatial configuration with inanimate subjects protruded, hang manner of motion rode, strides, bounded, walk less frequently: verbs of sound emission rattled passive verbs: were recovered, has been mounted, could be seen, heard *shop, talk, smile, complain: internally caused agentive activity verbs (unergative) *break, melt, dry: unaccusative change of state Somtimes activity verbs with animate subjects (unergative) are ok: The following are attested (pg 224): \eenumsentence{ ...around them chattered and sang as many girls... On the third floor worked two young women Behind the wheel lounged a man At one end in crude bunks slet Jed and Harry.. Sometimes agentive verbs of manner of motion occur with a locative (not directional) PP: \eenumsentence{ Above them pranced the horses... They were profusely covered with leaves and flowers among which ran, flew, crawled, fled, pursued or idled an extraodinary variety of animals, birds and insects. Verbs of (light, sound, substance) emission: \eenumsentence{ throug the enormous round portal gleamed and glistened a beautiful vallue.. On one hand flashes a 14 caret round diamond; on the other hand sparkles an 8 caret stone flanked by the diamond-studded initals WN. In the hall ticked the long case clock Over a Bunsen burner bubbled a big earthenware dish of stew. Verbs of body-internal motion: \eenumsentence{ Black acros the clouds flapped the cormorant in this lacey leafage fluttered a number of grey birds *!*Liddell, Scott K. Real, Surrogate and Token Space: Grammatical Consequences in ASL. Chapter 2 Langacker 1991: Current Discourse Space: mental space comprising those elements and relatoins construed as being shared by the spaker and hearer as a basis for the communicaton at a given moment in the flow of discourse. \subsection{Real Space a person's conception of their current directly perceivable physical environment: a grounded mental space; doesn't refer directly to physical entities, but to conception of physical entities. exists independently and apart from discourse: dependent on perception. (Slightly different than Gilles' "reality" space, which need not refer
Deictic elements such as you, here, this typically make reference to Real Space. 1.PRO: 1 handshape toward referent; 2. certain verbs can be directed toward specific body
tải về 302.24 Kb. Chia sẻ với bạn bè của bạn:
|