*!*Ghomeshi, Jila and Diane Massam. 1994. To Appear, Linguistic Analysis
tải về 302.24 Kb.
|
| principle to other verbs so that applying the same
combinatory principle to new verbs becomes progressively easier. ``Pathbreaking" verbs tend to be generic verbs: typical grammaticalized markers of tansitivy in many langauges. They do not have "high transitivity" according to Hopper and Thompson criteria. She proposes that their "high transitivity" = marked transitivity. The unmarked case is when the transitivity is inherent in the verb, whether or not the tense is present oor past, the mood realis or irrealis, whether the object is animate or not, etc. Some nice tables with first verbs. *!*Orgun, Orhan. ms. Turkish direct object (non)incorporation facts Word order facts: new topic focus predicate background Accusative objects can occur without case marking. They must immediately preced the predicate. He looks at the general cases: book-reading Orhan notes that nonreferential constituents tend to occupy
Stress facts are like Persian: V tends to be stressed (unless a particular argument or adjunct is focused), but in N-V, it is the N that is stressed. Orhan notes that hte V can bear contrastive stress but that lexical compounds cannot bear contrastive stress. He argues that the position of the bare N is motivated by parsability concerns since the nominative case is unmarked. peace happiness brings, both peace and happiness are unmarked. Principle: unless case marking gives sufficient clues, SOV order should be used. [But that doesn't explain how cases without subjects are parsed. It also doesn't explain why adverbs cannot intervene] Causativization treats the NV sequence as transitive: therefore the N seems to be the DO. (The causee of a transitive verb appears in dative case) [nice!: check Persian] Various suffixes, particles and function words can intervene: question particle: Ali today book-int read-past Is it a book/books that Ali read today? Particles expressing contrast, similarity and reduplicated Ns also intervene: Ali kitap bile okur Ali book even read-aor Ali kitap falan okumaz ali book etc. read-neg-aor Ali does not read books and the like Ali kitap mitap okumaz Ali book (redup!) read-neg-aor Ali does not read boks and the like. It is therefore not true, Orhan argues that the NV has any "morphological integrity." Also, the N can be syntactically complex: three book. *!*Payne, Doris L. Argument Structure and Locs of Affect in the Maasai
Claims: the DO bears TWO semantic roles: possessor and affected. cites several other cases in which argumetns require two roles. verb+construction is required: It is easy to differentiate intransitive and transitive verbs
Morphology exists for many variations in argument structure including impersonal passives, middles, antipassives, causatives, dative and instrumental applicatives. THE EP Construction: Possessor i smarked as object on the verb (for 1st and 2nd).
She.me.scream child.nom ``My child is screaming"
take person.nom woman.acc skin.acc "The person will take the woman's skin."
it.me.see goat.nom hand.acc ``The goat will see my hand (to my detriment or benefit). *!*Pine, Julian M, Elena V. M. Lieven and Caroline F. Rowland. 1998. Comparing different models of the development of the English verb category. Linguistics 36 (4) 807-830 >From Abstract: Great deal of lexical specificity in children's early use of morphology, auxiliaries, pronoun case marking and SVO order: lack of overlap in the verbs to which different morphological markers were applied, lack of overlap in the verbs with which different aux's verbs were used, disproportionate use of "I"; lakof of overlap in the words which served as subjects and DOs of transitive verbs. [they don't address the possibility that this is just what the child wants to talk about, not what she is capable of talking about] ``There has been a growing awareness in recent years of the shortcomings of models of grammatical developemnt based on the gradual extension of cognitive semantic categories. First, there is the problem that children's early grammatical knowledge does not appear to be restricted in the way that such models would seem to predict (Maratsos 1982, 1988; Maratsos and Chalkley 1980)." pg 3 [early N's for example include walk, kitchen, minute as well as ball] ...Second,...'agent' cannot be used as a bootstrap into subject in ergative languages. Third..children are capable of acquiring linguisitc distinctions which have little or no semantic base from very early in development...(e.g. count/mass, N/V) [yikes!] Valian (1986): the semantic heterogeneity of early lexical items implies the child is dealing with syntactic categories such as Det, N, NP, from teh beginning. Tomasello (1992): lexical constructivist account Problems with T.: -ing is generalized very early, pronouns seem
With MLUs of 2, English speaking children drop subjects less often than do Italian children (evidence of pro drop?).: Valian (1986; 1991). Radford's (1990) maturational model of acquisition: strict division between lexical and functional stages: predicts pro drop and no case marking in early multiword speech (Valian is critiquiing this). Different morphological markers initially applied to distinct sets of verbs (Bloom, Lifter and Hafitz 1980, Clark 1996). Reasonable to assume child has knowledge of abstract pattern if 1) a relatively large number of instances of the pattern are attested and 2) the pattern is defined with a particular verb (assuming, I suppose, that the verb always appears in the pattern). Experiment: 12 children: vocabs of 100 words on MacArthur CDI: therefore were in early stages of multiword speech. (aged 1;3.7 - 2;0.18) 2 20 minute recording sessions Productivity: verbs marked both with and without a particular morpheme/verbs marked with the morpheme: tested whether this was significantly different from zero. 8/12 children showed productivity with the progressive morpheme. Only two children were productive with the past tense and
Also, very little overlap in the verbs with which different inflections were used: i.e. little evidence for category-general marking. Any semantic "flavor" in acquisition may be epiphenominal: it may reflect the semantic distribution bias in the input, not the learner's bias. AUXILIARIES (main verb do and aux were differentiated on distributional grounds) aux + verb pairs were considered distinct ignorning tense, polarity and word order. can, do, be, have accounted for 90.3\% of all aux+verb combinations. Overlap = no of verb types occuring both with aux 1 and aux 2/no. of verb tpes occuring either with aux 1 or aux 2. None of the children showed overlap that was significantly different from 0: ``these results provide very little support for the attribution of syntactic auxiliary verb and verb categories to these children" pg 17 Another approach was taken: Searched for simply Aux + X: Used Braine's (1976) statistical criterion for productivity of positional pattern based on the Binomial theorem (6 different instances in the same order and 0 instances in a different order or 8 instances in the same order and 1 instance in a different order. [Shouldn't sample size play an important role here--perhaps 40 minues is not enough time!] Four chidlren did have one productive positional pattern: "can't + X" ("can't find it") or "don't + X" ("don't eat pen"). ``These data suggest that children's use of aux + verb combinations can be accounted for in terms of a mixture of rote-learned and lexically specific knowledge and would thus seem to count against a syntactic performance limitation account [like Valian's]" (pg 18). (also shows that children *are* generalizing quite early). All children had an "I + X" pattern that met Braine's criterion, but none showed contrastive position of "me". ``These findings count against a strong version of the Verb-Island Hypothesis since they suggest that although children's knowledge of pronoun case-marking may be lexically specific, they also have at least some lexically specific knowledge of SVO word order which generalises across different verbs." pg 20. Subjects did occur overtly often, but 44\% of the time, they occurred with one verb, the copula; also "I" was the most common subject; and subject NPs and object Nps showed little overlap (raises doubts about the generality of children's knowledge of SVO order) According to Braine (1987, 1988b): in artificial language learning: under serial presentation conditions, subjects readily learn the positions of words or phrases w.r.t. a marker. However, they have great difficulty in learning arbitrary dependecies between *classes* of words. Pine et al suggest that early grammatical knowledge may consist of a complex set of relations between high frequency markers and lexcially and morphologically defined slots which reflects the interaction between a distributional learning mechanism constrained in the way that Braine (1987) suggests....This would imply that verb-island effects are just a speical case of more general lexical learning effects." pg 26 *!*Pinker, Steven. 1986. Productivyt and Conservatism in Language Acquisition. In William Demopoulos and Ausonio Marras (eds) Language Learning and Concept Acqusition: Foundational Issues. New Jersey: Ablex Publishing. He gives evidence for conservatism including: Chidlren learning fixed word order lngs do not
Children never invert or put negation after main verbs in questions. Children do not use auxs in ungrammatical sequences Contra Conservatism: errors: overgeneralization of inflection to exceptional stems fail to use obligatory closed class morphemes fail to invert subj and aux (sometimes, while inverting
verb subcategorizations (long list of exceptions given (from Pinker 1984) performed an experiment with Jess, Annie Zaenen and some others: taught children novel verbs by performing an action for which there was no Engish verb and narrating the action with a single grammatical context (active or passive). Children were induced to produce the form and also were tested on comprehension. The children were between 4 and 5 1/2. [Pretty darn old!] These chidlren produce the alternative forms productively Also showed productivity in creating "to" forms after
*!*Prince, Ellen. 1996. draft. Constructions and the Syntax Discourse Interface. University of Pennsylvania ``the association between form and function, is, for all intents and purposes, an arbitrary one. " pg 2 ``the Construction Grammar notion of 'construction,' which inextricably unites form and function and which is language-particular, is compatible with what I shall propose but is not sufficient to account for all the relvant phenomena, since ...[form and function are not inextricably united...and some generalizations are universal and not language particular." pg 3. Generalizations: form-function associations are arbitrary (single discourse function
! Universal construction-template: can be mappen onto different syntactic forms in different languages. \noindent ! Left Dislocation: 3 different discourse functions: \eenumsentence{ [My sister got stabbed in her bed] The landlady, she went up.... \\ removes discourse-new entities from syntactic position disfavored for discourse-new entities and creates a separate processing unit for them ``She's going to use three groups of mice. One, she'll feed them mouse chow...Another, she'll feed them veggies." \\ the entity designated by the initial NP stands in a salient partially-ordered set relation to some entity or entities already evoked in the discourse ``This is Alice Freed's copy. My copy of Anttila, I don't know who has it \\ same as above [?] + the open proposition is info saliently on the speaker's mind, and the initial NP represents the instantiation of the variable and the new information." The same discourse property in \ex{0a, simplifying processing by avoiding introducing a discourse-new entity in a position disfavored for discourse-new entities accounts for the following syntactic forms, too [not enough explanation here]: \eenumsentence{ 'Run-on There are a lot of Americans, they can't read. Resumptive pronoun RC: There are a lot of American's who they can't read. Also, `analogous' syntactic forms in different languages with different discourse functions. e.g. Ward 1995: Italian analog of English existential there sentences have the discourse function associated with Eng presentational there sentences and vice versa. [see sentences in paper, page 5] Resumptive pronouns in Eng: the head is sufficient for the hearer to evoke the discourse entity; the RC only adds new information about that already evoked entity. Resumptive pronouns in Hebrew: (Erteschik-Shir 1989): used when a `restrictive-set' reading is intended: precisely where the discourse entity cannot be evoked until the RC is processed: \eenumsentence{ $<$ in the context where H knows that S bought a dress$>$ \\ Here-is the-dress that-I-bought [e] $<$ H knows S had 3 dresses in mind originally$>$ \\ Here-is the-dress that-I-bought it. ``no matter how `minimalist' the theory of syntax, we still find the literature replete with mentions of the 'passive construction,' the 'relative clause construction,' ...etc...Chomsky 1993, after stating explicity that constructions no longer have any place in syntactic theory, goes on to refer to sets of linguistic objects as constructions no less than 8 times. Thus constructions have clearly not been eliminated from syntactic metalanguage..." pg 5 ``A discoure function associated with a syntactic form in one language may be borrowed into another language where it is associated with a syntactic form of that language just in case the two forms..are mapped onto the same abstract Construction-template by the bilingual speaker." pg 7 Same form: [cp NPk [c' [ip NPj [Vi [vp tj ti ...tk]]]]]: Topicalization (or "focus movement"):
Focus-frame (presupposed) is an open proposition; focus identifies the instantiation of a variable The open proposition conveys that a certain entity has a certain attribute; the focus reprents the value of that attribute, new to the hearer The Hearer knows that hte entity has the attribute in question Yiddish movement. {Egg creams, you want, bananas you'll get.(Ross 1967: 267) Less constrained: The info in the open proposition is known or inferrable to the Hearer: not limited to values of an attribute: may represent discourse entity, etc. Prince: speakers have borrowed the discourse function of a syntactic form in Yiddish and associated with the syntactic form called Focus Movement in English. ``Yiddish must have a syntactic form which can be construed as, in some sense, `the same form' as English Focus Movement and which has the same discourse function as Yinglish Yiddish Movement" pg 10 `` The syntax of Yinglish Yiddish Movement...is significantly different from that of Yiddish Focus-movement...The most obvious difference is that the subject in ..Yinglish..is preverbal, while the subject in the Yiddish sentence follows the finite verb." pg 12 (notes this is because of V-2 constraint) ``..what exactly has been borrowed? Certainly not the syntax... a discourse template function..cannot stand alone...The form chosen...is that of English Focus-Movement and this choice, while intuitively obvious, is not at all clear unless we adopt something along the lines of the Construction-templates mentioned earlier.. pg 12. *!*Rappaport Hovav, Malka and Beth Levin. 1995. The Elasticity of Verb Meaning. In IATL 2: Proceedings of the Tenth Annual Conference of the Israel Association for Theoretical Linguistics and the WOrkshop on the Syntax-Semantics Interface. Projection Principle theories: projectionist theories vs Constructional Theories ``In projectionist frameworks, verbs with multiple meanings have multiple lexical representations...If this kind of variation is the rule rather than the exception, this approach leads to a significant proliferation of lexical representations. However, this kind of variation is not merely associated with individual verbs but rahter with classes of verbs." pg 2 (But then they suggest rules) A lexical entry of a verb is an association of a constant with a lexical semantic template (the 'thematic core' of Pinker 1989). Re: section 2: Comparison with Constructional Theories ``Proponents of the constructional approach have suggested that they
``...both theories distinguish between two aspects of verb meaning. We can call these aspects the structural and the idiosyncratic..." pg 4 I think you misrepresent what's been done in the way of restricting which verbs can occur in which constructions (pg 5). It's the subject of much of chapter 2 of my book (particularly 2.4 to the end) and also of chapter 5. It's not at all the case that any construction is supposed to appear with any verb. That may be what Hoekstra had in mind (to be honest, I've had a hard time figuring what he had in mind), but it's clearly false. In talking about the ditransitive, the resultative, the "caused-motion" and the "way" constructions (chapters 6-9), I also suggest pretty specific semantic constraints that further constrain which verb classes can occur with each construction. The lexicon is not supposed to be just a list of concepts at all (pg 4)! It is highly structured by non-monotonic inheritance links of various types that I discuss in chapters 3 and 4. Also I wouldn't say that the meanings of verbs are atomic exactly, they just aren't amenable to decomposition in terms of a finite set of primitive elements. They do have some internal structure: they specify which participants are obligatorially expressed, and they make reference of all kinds of frame-specific knowledge. Finally, constructions reside in the lexicon (or better, the constructicon) alongside words, so it's not clear it makes sense to contrast the lexicon with "syntax" the way you seem to on page 4. Manner verbs: activities Result verbs: achievements or accomplishments: the complex
Assumptions: Templates may be freely augmented up to other possible templates
``because of the template assocaited with a verb like break cannot be augmented further, no other achieved state or location can be added to a sentence with break unless it further specifies the change already lexicalized in the verb as in Kelly broke the dishes to pieces. [But Kelly broke the eggs into the bowl. is fine.] [You don't intend *'s on the first examples of 23 a and b, do you?] ``The acse study here is a first attempt at making explicit
BRISSON, C 1994. The Licensing of Unexpressed Objects in ENglish Verbs in CLS 30. 90-102. Is cited as a constructional approach. The main claim seems to be: an argument can be left unexpressed if it understood as 'prototpyical' pg 9. *!*MacDonald, Maryellen C. Neal J. Pearlmutter, and Mark S. Seidenberg. 1993. The Lexical Nature of Syntactic Ambiguity Resolution. ms. USC. Swinney 1979, Tanenhaus, Leiman & Seidenberg 1979: multiple senses are accessed. Swinney claimed senses exhaustively accessed; others claimed only most common meanings were accessed. Some other studies argued that context did have an effect: at least other words in a sentence could prime a particular meaning (Simpson 1981, Tabossi et al. 1987). Swinney's 2 senses were roughly equi-frequent. Simpson (1981): reviews studies suggesting that frequency affects the order in which the meanings of ambiguous words are accessed. Highest frequency senses seem to be accessed first. Assumptions: frequency can affect access; contextual info can result in activation of only a single meaning; Syntactic processing has been viewed very differently: Frazier's Minimal attachment Studies suggesting lack of context effects in parsing reduced relatives (vs main verb interpretation)(pg 13): Ferreira & Clifton 1986, Rayner et al. 1983, etc. Counterarguments that context effects were obscured by poor stimulus construction (e.g. McClelland, Tanenhaus, etc). M: very similar empirical theoretical and methodlogicical issues have arisen in both lexical and syntactic domains: the role of frequency, the role of context, whether modular or interactive. ! Claim: the parallels derive from the fact that the syntactic ambiguities are based on ambiguities on the lexical level. Argument structure: capture facts about correlations between syntactic and semantic info: specify categorial info about complements, and thematic role info. 'This conjunction of syntactic and semantic info in one representation captures interdependencies betwen these types of info that are important for language professinga nd ambiguity resolution (Boland & Tanenhaus 1991; FCarlson & Tanenhaus 1988, Stowe 1989). X' theory is put into the lexicon: 'The lexicon is thus the level at which syntactic processing takes place'. Since it is clear that levels are not independent, connections between levels of lexical represtnation can be exhititory or inhibitory. Ambiguity resolution is a constraint-satisfaction problem. (Parallel is drawn to Bates and MacWhinney model). Pieces of syntactic structure can be activated to differing degrees. Early on M seems to criticize the idea that meanings are fixed entries that can be 'accessed'; instead she suggestss that meanings are activated to varying degrees. (But her own representations seem fairly fixed.) M notes that not all contexts are equally constraining: The next word in the sentences is [blank] vs I drink my ocffee with cream and [blank]. She seems to attribute the lack of context effects found by Swinney 1979 to the fact that the contexts were not constraining enough. TRUE?? She notes that lexical info seems to dominate over effects of contextual constraints in lexical processing ('bottom-up' effects of Marseln-Wilson 1987). This is attributed to the fact that contexts are less effective at preselecting a possible alternative. \medskip
MV/RR ambiguity: Three types of verbs: enjoyed is used much mre often as a past tense than as a past participle: shold therefore be easier to comprehend in a past tense construcion than a past participle. examined more equaly balanced between past and pp reviewed is used more frequently as a past participle than
Findings after reanalyzing previous studies: all effects were in the predicted direction (not clear how strong the findings were, though). Relative frequency of particular argument structures should have a similar effect: push is more likely to be transitive; move is more frequently intransitive. \smallskip Kawamoto's (1988, 1993) model suggest that
\medskip
! NP/S ambiguity Minimal Attachment predicts that NP interpretation should be favored, because S is more complex phrase structure. Can test frequency effects by looking at verbs which statistically prefer the S arg structure, though. Such studies have had mixed results. M offers an explanation: 1) Trueswell 1993 note that many supposedly S-biased words such as pray, agree were actually biased toward some other arg structure. 2) Low frequency verbs (e.g. hint) that take an S complement typically prefer to have the that complementizer present, whereas high frequency verbs (e.g. know) often omit that. Target sentences normally omit that to make the sentence more easily ambiguous: which creates oddness effects for verbs like hint. \medskip
predicts: depends on your fav theory of phrase structure. Speakers often prefer PP complement attachment. M: different factors come into play here: the type of P involved (one that often marks a complement, e.g. with, into or one that often marks a modifier. Also, does the verb obligatorily, or frequently require a PP complement? Context effects seem to be most powerful in PP-attachment \medskip Juliano and Tanenhaus 1993 show that that at the beginning of a sentence is more likely to be a determiner, and almost never a complementizer; but that after a verb is more likely to be a complementizer and not a determiner. In reading-time experiments, subjects were extremely sensitive to these contingent frequency effects. ambiguity. *!*Macken, Elizabeth, John Perry and Cathy Haas. 1993. Richly Grounding Symbols in ASL. CSLI Report No. 93-180. `` The use of richly grounding symbols [RGS's] does not imply that ASL lacks the features of structure and productivity that are the hallmarks of language. SEcond, the meaning of RGSs may be conventional. The correct contrast is between symbls whose structure has a cognitively robust relation with the structure of that which is symbolized, and hence easily inferred and/or remembered, and symbols whose connection with their meaning is arbitrary. Third, richly grounded meaning is found in the way we modify and combine symbols, as well as in the basic symbols themselves." pg 16. ASL is a dual-representation language: combines both arbitrary signs and RGSs. They note this can help explain Bellugi's observations that citation forms often take twice as long as spoken words, but the time to express a whole proposition is the same. *!*McNally, Louise. Bare plurals in Spanish are interpreted as properties. ms 1995. Universitat Pompeu Fabra. Hi Louise, I just read through your paper, and I find it very clear and well written. I am the most interested in your analogy to incorporation, which I find intriguing. I think it's probably common for languges that allow incorporation to also allow either BPs or BSs (bare singulars) with a property interpretation. In fact, it may be that this is a logical (historical) precursor to incorporation. E.g. Persian has a productive pattern of both BPs or BSs, and the incorporated cases seem to be cases in which the BS and the verb have formed an especially tight semantic bond, which is then mirrored in the tighter syntactic bond. (So I wouldn't necessarily set the two possibilties up as two independent ways to satsify a single pragmatic function :pg 13). I wonder what the relation between BSs and BPs is. You could imagine only BSs occuring in truly incorporated forms because of word-formation constraints (derivation preceeding inflection). But Persian allows free unincorporated objects to be either BS or BP: both with a pretty darn close generic interpretation. I guess I'm just asking myself this out loud: what is the sematnic difference in that case? ``[This] analysis suggests why BPs resemble incorporated nominals... without requiring that they be incoporated in any interseing syntactic or morophological sense...incorporated nominals..and BPs would resmble each other in their sematnic type---both interpeted as properteis (as van Greenhoven 1993, 1995 proposes--and therefore likely to combine with predicate denoations via similar semantic operations...speakers of languages without morphological or syntactic incorporation constructions would find use for other linguistic devices (such as BPs with their special semantics) that would allow them to approximate the pragmatic functions of incorporation. pg 13 *!*Miller, D. Gary. 1993. Complex Verb Formation. Philadelphia: John Benjamins Publishing. A Complex Verb : ``oen that has undergone some sort of derivation to alter the form, meaning, or argument structure of the base verb (or verb 'root'). For the most part, only affixation is treated..." ``syntactic morphologists generally relegate the `adjectival passive' to the lexicon." ``This work takes the radical position that it is not accidental that most data admit of either a syntactic or lexical, PS or non-PS analysis, and that the reason for it is that both are likely to be valid - under diffrent circumstances. Reasons why all words are often claimed to be listed in lexicon: lexical insertion has (at least in the past) been
derived words tend to acquire idiosyncratic meanings e.g. transmit: transmission (Aronoff 1976) Listing whole words would account for the INHERITANCE of irregularity by a derivative (Aronoff 1976). Lexicalist Hypothesis: (Chomsky 1970): syntactic rules cannot make refernce to any aspect f the internal structure of a word This position is summarized by DiSciullo & Williams (1987:49) as follows: \begin{quote Words are 'atomic' at the level of phrasal syntax and phrasal smeantics. The words have features, or properties, but ehse featuers have no structure, and the relation of these features to the internal composition of the owrd cannot be relevant in syntax - this is the thesis of the atomicity of words, or the lexical integrity hypothesis of Chomsky (1970), WIlimans,..and numerous others {quote Strong Lexicalist Hypothesis: fully inflected forms (provided by lexical rules) are placed in phrase structure by lexical insertion. Weak LExicalist Hypothesis (Interpretive Morphology): inflectional operations are applied in the syntactic component (or later). ``Radical departures from the various lexicalist positions are found in Sproat (1985), Baker (1988) and Hale and Keyser (1991)": Phrase STructure Morphology: 1) there is no separate morp0hological component 2) there are no Word Formation Rules 3) all morphology is essentially syntactic in the sense that it is governed by grammatical rather than lexical principles and parameters \subsection{Derivation vs Inflection Derivation is semantically based, recursive, and non-suspendable;
Bybee (1985) views derivation and inflection on a continuum. Bybee's COntinuum of Expression Units: lexical - derivational - inflectional- free grammatical - syntactic (Miller doesn't buy it)
(Beneficiary) Goal args fail to undergo Middle Formation in both Grm and Icelandic, irrespective of their morphological case marking. Lexical case-makring on a verb's object is not a relevant factor (contra Fagan 1992). The fabric washes well. *He helps with difficulty. Help takes a goal or ben argument Evidence for ben/theme distinction:
*!*Mohanan, Tara. 1994. Wordhood and Lexicality: Noun Incorporation in Hindi. Submitted to NLLT NI: noun combining with a verb to form a complex verb: Mithun 1984, Hopper and Thompson 1984, Sadock (1980, 1991), Baker(1988), DiSciullo and WIlliams(1987) and Rosen (1989) are cited as general refs. Agreement facts argue that the N must be treated syntactically as an argument on par with the other args of the verb (I think this is Baker's take, too). At the same time, though, NI is lexical in various ways. Most NI sentences are actually ambiguous between an NI reading, and a simple verb reading. (e.g., "Anil will sell books," OR "Anil will do book-selling.") nothing may intervene between N and V in NI: not subject, locative adjunct, or negative. (word order is usually free in Hindi) [certain things can intervene in Persian] N cannot be modified, and modifier cannot be stranded either: this is not predicted by Baker's account of head-movement, or by Sadock's coanalysis account. [same is true in Persian] neither wh-words nor pronouns can be the N in NI: ``*what-selling" ``*this-selling": M attributes this fact to the fact that wh-words and pronouns are maximal projections, and N of NI is necessarily zero-level. [would also seem to be ruled out by the naming pragmatic constraint, though] neither the N or the V can be gapped in a coordination construction: ``*Anil does horse buying and Ram does -selling." ``*Anil does hourse-and Ram elephant-selling." This is claimed to follow from the Lexical Integrity Hypothesis: the categorial representation formed inthe lexical modeule is not visible to principles of phrasal organization. [M points out that adopting this principle effectively rules out a head-movement analysis like Baker's or a coanalysis like Sadock's as well]. neither N or V can be conjoined in NI: this is attributed to the fact that lexical categories cannot be conjoined in Hindi. N doesn't bare an overt case clitic: but M claims N is in nominative case: is nominative case is unmarked? how do we know it's there at all? Explanation for this is that clistics apply to phrasal categories, not zero level categories. evidence also comes from phonology: stress contours NI can be input to lexical derivational N formation: via derivational suffix -vaalaa Semantic/Pragmatic constraints on NI: N must be generic or ``non-referential": it can only
\bigskip
Agreement facts: 1) Verb agrees with the object iff the object is NOM, and the subject is not NOM. 2) NOM case can only be determined after phrasal concatenation: because phrasally concatenated model can assign DAT case to the subject 3) V can agree with incorporated in NI SOLN to the paradox: separate two dimensions of syntactic info: grammatical functions (NI involves PRED and OBJ) and grammatical categories (V0 -> N0 V0). *!*Naigles, Letitia, Henry Gleitman and Lila Gleitman. 1993. Children Acquire Word Meaning Components from Syntactic Evidence. In E. Dromi ed., Language and Cognition: A Developmental Perspective. Human Development Vol. 5 104-140. New Jersey: Ablex Publishing. Frames:
NP V NP PP: The elephant comes the giraffe towards the ark. NP V NP: The zebra goes the lion. NP V PP: The lion puts in the ark. NP V: The zebra brings. 51 2 yr. olds; 23 3yrs; 20 4 yrs; adults were given sentences such as those above, as well as grammatical sentences, and asked to act out what the sentence described. Frame Compliance: scene acted out was in accord with the syntactic frame, not the already known lexical semantics. Verb Compliance: scene was acted out in accord with the previously known sematnic implications of the verb itself Other: the enactment is "wrong" in some other way. \noindent Results: Mean Percent Frame Compliance as a Function of Frame Type and Age \begin{tabular{lllll Frame & 2 & 3 & 4 & Adult\\ NP V PP PP & - & 78.75 & 79.20 & 75.85\\ *The elephant comes the giraffe towards the ark & & & &\\ NP V NP & 69.15 & 70.45 & 71.65 & 50.00\\ *The zebra goes the lion & & & &\\ NP V PP & 60.40 & 63.85 & 70.45 & 30.40\\ *The lion puts in the ark & & & & \\ NP V & 65.37 & 40.00 & 44.55 & 34.14\\ *The zebra brings & & & & \\ {tabular Children show more Frame Compliance than adults, probably because they performed "repair" on the ungrammatical sentnces where possible, interpreting, The zebra goes the lion as The zebra goes TO the lion. Frames with PPs were more likely to be Frame Compliant because ``the additional semantic info in the PPs simply makes it easier to take a guess at what the whole sentence might mean." pg 128. They assume that use in a syntactic frame reflects an additional component to the verb's meaning (doing a little syntactic bootstrapping of their own, as they put it). The assume: ``The number of nominal phrases and the number of thematic roles in a clause must line up one-to-one (theta-crterion; see Chomsky 1981)." (129) The note that "word formation rules" are more likely to add structure than to tear it down. Where do the presumed correspondences come from? They assume ``there is sufficient cross-linguistic similarity in these linking rules to get the learning procedure started. For instance, the best guess from cross-linguistic inquiry is that, in all, languages, the number of NPs in a snetence at teh surface is a straightforward function of the logic of the verb concent... there is an overwhelming tendency, corss-linguistically, for agents to appear as subejcts and themes as direct objections, with other arguments appearing in oblique cases." (pg 136-137) *!*Nishimura, Yoshiki. ms 1995. Agentivity and Causation in Cognitive Linguistics. Describes a prototype category of agentivity and causative constructions. (clearly written) *!*Pesetsky, David. ms. Zero Syntax [Adele: read Woodbury 1987 pg 688ff on discussion against placeholders in syntax: cited in Knud's book] *!*Polinksy, Maria and Isaac Kozinksy. 1992. Ditransitive Constructions in Kinyarwanda: coding conflict or syntactic doubling? CLS 28. ed. by C. Canakis, G. Chan and J. Denton. 426-439. Coding Conflict: absolute identity in the marking of two or more distinct grammatical, semantic or pragmatic functions by free or bound segmental elements and/or by suprasegmental elts. They argue against syntactic doubling, wherein two of the same GR are found within a clause. They look at Kinyarwanda and argue that the patient of ditransitive applicatives is a DO, the recipient being an IO. Range of applicatives ("ditransitives"): { Possessor Ascension: \\ man hold woman child \\ the man is holdingh te woman's child {Benefactive: \\ woman read child book \\ the woman read the book to the child. {Instrumental Advancement: \\ woman wash child soap\\ The woman washed the child with soap. {Morphological causative: woman wash-CAUS girl child \\ THe woman made the girl wash the child. Prototypical ditransitives: the verb semantically requires a recipient: give, show, offer, demand, request, ask, send, teach, explain, promise: these require no verbal marking (neither do benefactives). Evidence that the patient is DO, the recipient IO: Conjunction Reduction: "The cow kicked the woman's child and the child *the woman cried." The patient has more controlling properties than the advanced/ascended NP. *!*S. Riehemann, 1997 ms. Idiomatic Constructions in HPSG. In JP Koenig, gert Webelhuth, Andreas Kathol (eds). This paper explicates how phrasal idioms can be represted directly in HPSG without assuming that their meaning is present in any (or all) lexical items within the idiom. This paper represents an extremely valuable contribution to the theory, since it provides a natural way of accounting for the intuitively phrasal but cleary flexible nature of idioms. It would also be a simple extension to account for meaningful, phrasal, flexible constructions of more general types such as locative inversion or the argument structure constructions discussed in Goldberg 1995. Below I offer some suggestions, mostly for clarification of the main points: --I've never heard of the "cat among pigeons' idiom. The author
--I don't understand the argument based on examples in (3) that "loose ends" is not an argument of "tie." Is the author assuming that subjects are not arguments? It would seem that the raised or passivized argument of a verb is still an argument, which would make the examples in (3) unremarkable on a lexical account. --SR cites several problems with lexical approaches: -would seem to predict that the idiomatic meaning of the word coould appear elsewhere. -would seem to have problems with the fact that parts of idioms often specify adjectives, adjuncts, and determiners not normally subcategorized for. --Nice dicussion of Pullum 1993 who suggests that idioms are represented semantically by meaning postulates basically. -SR notes that inference is too strong, would predict idiomatic meaning available for: The cat and the dog are out of the laundry bag. (-> the cat is out of the bag) - Also, psycholinguistic evidence about the processing of idioms suggests the literal interpretation is not generally accessed first [should give a citation here: e.g. Ray Gibbs 1993 Poetics of Mind]. --"OSU" in footnote 5 should be unabbreviated. --The idea proposed in footnote 8 is interesting and has precedent in
--While the proposal is exciting and interesting, it could stand to be explicated a little more clearly. SR proposes Underspecified Phrasal Semantics (UPS approach), following Copestake 1994. phrases have set-valued feature WORDS, containing all of the words in the phrase --It is claimed to be a virtue of the theory that the words of
--The words have their ordinary meanings, ``overwritten during compilation by skeptical default unification" [a word about what this is would be helpful], or maybe they don't because ``all this can be precompiled..[so] that the literal meanings are not present anywhere in the parse." pg 8 note 6. (see below *) --Only the semantic relationships between the words are specified: beans is the undergoer of spilling, which indirectly fixes the syntax to some extent: allows modification, ``movement". --It is noted that the words in the set do not all have to be sisters when they appear overtly. They are related semantically, not syntactically, but does this violate some kind of locality condition nonetheless? Maybe not. -- Nondecomposable idioms are given without specifying meanings for individual words *and* by specifying syntactic form. By not specifying the semantic relationships among the words, it is clearly necessary to specify the syntactic relationships if the idiom is to be expressed at all. However it's not clear to me that this follows from the proposal, since it would seem to be possible that only a set of word forms could be specified which might then occur overtly in any configuration whatsoever (The kick was a bucket). Is there some general mechanism that avoids this? -- ``The necessary information about the literal words is available and can be used to restrict syntactic variation" --Is info about literal words available (see above *) --assuming it is in fact available, how does that restrict the syntactic variation exactly? -- Idiom families are said to be handleable because the literal meanings are accessible [again see *] ( The feature CXCONT is used in teh ERGO grammar to encode the semantic contribution of constructions.) *!*Rosen, Sara Thomas. 1989. Two Types of Noun Incorporation: A Lexical Analysis. Language 65 2. With Mithun and DiSciullo & WIlliams, she argues that NI is a lexical process (against Baker and Sadock). Two types of NI: Compound NI: involve simple compounding: one argument of the simple verb is satisfied
the arg structure of the simple verb is changed to become intransitive no stranding of modifiers no doubling outside verb Classifier NI: can still take an object: the incorporated N is semantically like a classifier of the object. Simple and complex verb's arg structure is the same: can still take an object Languages with this type of NI are said to allow 'stranding' of modifiers (null-head modifiers). doubling outside of verb Compounding shoudl be invisible to the syntax. [This typology doesn't seem to allow for transitive CPs which are not
R notes that there is a subject/non-subject asymetry: with deep subjects not incorporating: instrumentals incorporate in Nahuatl, means phrases incorporate in Niuean, and some locatives incorporate in Samoan--non of these phrases is ever a DO. Goals and benefactives never incorporate in any language. *!*Sag, Ivan A. ms (1994 Dec) English Relative Clause
Argues that the move away from construction-specific rules has led to a sacrifice of precision. ``this method [of multiple inheritance hierarchy] has a significant untapped potential forthe treatment of grammatical ocnstructions as well, a point well appreciated by the closely related ongoing work within the tradition of Construction Grammar..." pg1. Notes to Ivan: Dear Ivan,
pg 3, 1st PP: explain "sponsorship" or maybe footnote this. pg 5: The Korean verb marking is nice motivation for the verb being
pg 6: To make contact with ConGram, you could mention Jean-Pierre's thesis as another reference for multiple inheritance of lexical items. The sign is defined just like the construction in ConGram: a pairing of form and meaning. The difference in terminology is kind of telling, though, since we've heretofore largely concentrated on the phrasal patterns to almost the exclusion of the lexicon (but see Rhodes' CLS paper on morphology in Congram, JP's work), whereas work in HPSG has come at form-meaning pairings from the Sausurrian "sign" angle. Pg 9: Work in ConGram has discussed multiple inheritance for phrasal patterns (e.g. George's "There"-constructions paper in WFDT, Fillmore & Kay ms, chapter 4 of my book). (You might spell out "daughters" for DTRS once) Pg 10: middle PP: you mention that [VAL-D elist] is a default
This question is of some interest to those of us working in ConGram, since Chuck, Paul and JP adopt complete inheritance because it is more constrained, whereas George and I prefer normal or default inheritance, since you can capture more types of relations. pg 20: treating "that" as a finite verbal head of the RC is a stretch for me. pg 26: Quirk's observation is not specific to postnominal modifiers: \#arrived guests recently arrived guests It seems that change of location predicates are not generally good as modifiers (at least without additional adverbial help): \#moved in couple recently moved in couple pg 26: when you say, "we should be able to explain linguistic systems
do you mean that the classification in terms of sorts is supposed to be a type of general categorization? If so, then normal mode inheritance may be more accurate, since people seem to categorize in terms of prototypes and extensions from them. Default inheritance seems to capture that better than complete inheritance. It may be outside the scope of this paper, but have you thought about how to handle resumptive pronouns in RCs? It seems they might cause problems for the Valence Principle. Typos:
add period to PP pg 8: 1st PP: "phrases...as feature STRUCTURE that serve" pg 10: you sometimes use "hd-valence-ph" and sometimes abbreviate to "hd-val-ph" pg 11: footnote 9: "one important CHANGES" *!*Sag, Ivan A. and Janet Dean Fodor. 1994. Extraction
They debunk putative evidence for WH-traces as syntactic entities: Wanna Contraction: "wanna" does not contract if there is a trace intervening between want and to. They note that a nominative trace does not prevent contraction: {
Bresnan proposed that the 's is a proclitic on beneath
! Floated quantifiers: {*How satisfied do you think they were all [ ]?
Psycholinguistic evidence: antecedent activation occurs at the verb and at the trace position. Following Fodor (1993) they point out that no pyscholinguistic finding could even in princpile qualify as evidence for empty constituents unless it were established that the data pertained to the syntactic processing or representation of the sentene and not the semantic. The parser detects a gap, but that doesn't mean that the parser detects a phonologically null syntactic entity. They argue conversely that Pickering and Barry's putative evidence against a syntactic gap is also not conclusive. *!*Schmerling, Susan. 1973. Subjectless Sentences and the Notion of Surface structure. CLS 9 577-586 \eenumsentence{ Seems like the class always wakes up five minutes before the bell rings. Guess I should have been more careful Going to lunch? differ from fragments in that they require no preceeding discourse. Perception verbs: \eenumsentence{ Looks like an accident Sounds like another ghost Smells like hot metal Tastes like almonds Feels like real silk. Modals:
`` Wastebasket class":
(Almost no analysis here) ``there is some elusive element of spontaneity and implusiveness involved
\eenumsentence{ ??Got a lot of nerve. Got a lot of nerve, doesn't he?/I'd say/eh? *!*Slobin, Dan I. 1997. The Orgins of Grammaticalicizable Notions: Beyond the Indidivual Mind. In Dan I. Slobin (ED) The Crosslinguistic study of Language Acquisition: Vol. 5. Expanding the Contexts. Mahwah, NJ: Lawrence Erlbaum Associates. Dan argues that there is a cline bewteen open and closed class lexical items and that we do not need to posit innate mechanisms to account for the cline. *!*Slobin, Dan I. 1982. Universal and Particular in the acquisition of Language. In Eric Wanner and Lila R. Gleitman (eds.) Language Acquisition: The State of the Art. New York: Cambridge University Press. 129-170. Different languages differ in the ease of acquisition of different subsystems. Word order and adpositions in particular are studied. Operating Principles: predispositions to perceive speech
>From previous work: ``Surface marking which is `overt' and `clear' is more easily discovered and acquired by the child" ``...postposed markers (sufffixes and postpositions) are more salient than preposed markers" Here proposes that sentence processing is facilitated by local cues: signals to underlying menaing which occur at localized points in the sentences. E.g. inflections to indicate grammatical relations are easier than word order. Also, separate clauses are easier than particles which conflate several notions or nominalizations which condense entire clauses. ``On lexical level, semantic clarity of forms and a minimum of lexical diversity facilitate acquisition" *!*Sweetser, Eve. 1987. Metaphorical Models of Thought and Speech: a comparison of historical directions and metaphorical mappings in the two domains. \eenumsentence{ report $<$ to carry back (again) (Latin) refer $<$ to carry back (again) (Latin) Traugott (1985), Traugot and Dasher (1985): mental state, speech act $<$ physical location/motion Sweetser: Explanation: Both mental processing and conversations can be understood in terms of a metaphor for travel: \eenumsentence{ walk over/walk about think about/think over talk over/talk about speech act senses $<$ physical location/motion Don't find: mental state $<$ speech act Traugott (1982): expressive $<$ textual $<$ propositional [Use {THE car of the 21st C $<$ the car $<$ that car {epistemic $<$ root ] Particular mental state is often a preconidtion for the felicitous performance of a speech act: \eenumsentence{ one must have mentally certified (made certain) before one can certify its truth to others mental observation is a precondition for observing ``making a remark" mental recognition is a precondition for giving a person the floor Decision-making is mentally spearting the chosen option from the rejected option. {decide (Lat decaedo ``cut off from") Categorization is putting concepts into sets. \eenumsentence{ distinguish (Lat. Di(s) ``apart") confuse (Lat con+fundere ``pour together, mix") I can't tell them apart I can't tell one from another Reasoning as building a logical object: \eenumsentence{ suppose $<$ Lat sub+ponere ``put under" hypthesize $<$ Gr hypo+thesis ``under + putting" Argument As Physical Combat: \eenumsentence{ concede $<$ Lat concedere ``to yield, to give up" insist $<$ Lat, ``stand in/on, stay" [notice it has a force-dynamic element] convince $<$ Lat con+vincere ``(intensifier+concquire" TODAY: one is gaining/losing ground in an argument can hit below the belt someone brought out hte heavy altillery *!*Thrasher, Randolph Hallet Jr. 1974. U of Michigan dissertation Morgan (1972) discourse fragments in 5 minutes: require previous discourse context. Present cases are distinguished from the following in NOT requiring special context: \eenumsentence{ Must be broken Probably hasn't gotten home from work yet Long discussion in chapter 1. Bolinger The Imperative in English (no date given): \eenumsentence{ like it? care to come along? tell the differnce? \eenumsentence{ Gotta go now See you next Tues Too bad about old charlie No need to get upset about it Been in Ann arbor long? Ever et a chance to use your Dogrib? Ever get to Japan, look me up Good thing we didnt' run into anyone we know Last person I expected to meet was John Wife wants to go to the mountains this year. pg 5 Subjects (I, you, there, it and aux's, if, articles are dropped. ommission from initial position
\eenumsentence{ Who are you rooming with? \\ Boy I went to high school with [ok? seems topical...] Why do you have to go home? Son's sick. Hints at relation of htis type of deletion with that found in newspapers, telegrams, signs. pg 21-23. Main difference: place of deletion is more restricted in present case. T argues for actual deletion because reflexives can occur in fragment. ``the choice has been narrowed to deciding between generating subjectless fragments
Reasons for creating them as part of grammar: 1) they refer to grammatical elements ! Article deletion: ``articles in any other position than at the front of the sentence do not delete" pg 36 Modification of N sometimes encourages deletion: \eenumsentence{ *? Boy is at the door. (a) Boy who says he's a student of yours is at the door. (The) Boy who called earlier is at the door. [In these cases the modification makes it clear that speaker is INTRODUCING the boy] Dummy but not anaphoric it and there delete \eenumsentence{ TOo bad you lost More than one way to skin a cat. Not all dummy subjects delete. Hedges, quantification are suggested to help deletion. ``it is also obvious that the totally unexplored are of conversational expectation is involved. ...if your guest commented that he was thirsty or really hot..": (there's) beer in the refridgerator. Avoidance of ambiguity: Deletion is blocked if remaining fragment looks more like a simple NP or an imperative than an S. pg 41: {*(I'll) try to leave as soon as possible Stressed elements do not delete. {*Hey Mike. Turn now. \subsection{chapter 3: I and you \subsection{chapter 4: no title
Possessive pronouns can be deleted in conversation: \eenumsentence{ Bill. Wife's on the phone. Got to go. Kids are waiting in the car. Bill's not coming in. Son just called to say he has the flu. pg 83 {(my) head is really throbbing. pg 86 Permission statements or reports of permission allow my,your to delete:
"Pronoun must be initial position" Notice not just subject below: {Wife and kids getting used to life in Ann Arbor? pg 83 \\
For some speakers, all of the possessive determiner deletion is implossible. ``the variation gives support to Roos' squishiness principle. In fact, there seems to
Hierarchy: { aux $>$ dummy subj $>$ if $>$ subj pronouns $>$ articles $>$ possessive pronouns Not quite true to say INITIAL position: \eenumsentence{ Bigger they are, harder they fall. if you don't get started, light'll turn red on you ]item when you get a minute, like ot have a word iwth you This time next month, won't be able to buy beef at any price After class, you going to be free can't go otu like that. Hair's a mess, sweatshirt's on inside out, socks don't match, and your face isn't even washed. pg 89 ``Rapport deletion" (pg 99): {Gotta match?
tải về 302.24 Kb. Chia sẻ với bạn bè của bạn:
|