- Количество слайдов: 41
The ontology of syntax and the ontology of thought Wolfram Hinzen Durham University ’Interdependencies‘ Munich, 21 January 2008
From theory to ontology (Quine) ► Jackendoff (2002): § The ontology of syntax: ►N, V, X’, XP, gender, … § The ontology of semantics: ►THING, EVENT, CAUSE, presupposition, predicates, , … ► Mutual autonomy?
Tripartite parallel architecture of linguistic processing Phonological formation rules Syntactic formation rules Conceptual formation rules Interface rules
The ‘immediacy model’ (Hagoort 2004, 2007) ► Parallel processing in distinct ‘syntactic’ and ‘semantic’ components ► No stage of processing has a representation of sentence-level information where syntactic form is mapped compositionally to semantic content. § syntactic units do not function as units of semantic interpretation ► Direct mapping of an input to a discourse model, by-passing syntax or else integrating it directly with non-linguistic sources of information.
Form and content in neurolinguistics the ‘distinction between syntax and semantics [is] based on extensive theoretical and empirical work’ (Newman&Ullman 2001) ► ‘the distinction between syntax (sentence form) and semantics (sentence meaning) is fundamental to our thinking about language. (…) Our findings strongly indicate that a part of Broca’s area (BA 44, pars opercularis) is critically implicated in processing syntactic information, whereas the lower portion of the left inferior frontal gyrus (BA 47, pars orbitalis) is selectively involved in processing the semantic aspects of a sentence’ (Dapretto et al. , 1999). ►
Typology, I: Gil (2000) ► ‘overwhelming evidence for the distinctness syntax and semantics’ of § Undecidability of truth value in First Order Predicate Calculus § a virtually infinite number of potential semantic categories with no effect on syntax ► ‘Semantics before syntax’ § positing semantic categories does not add to the ontological complexity of the description.
Typology, II: Haspelmath (2007) ► Non-existence of pre-established categories for language typology § Language comparison cannot be category-based § Must be ‘substance-based’, as substance is universal. ►Phonology: focus on phonetics ►Syntax: focus on semantics
Philosophy of language: semantics=‘reference’ FORM [Pegasus [is [a horse]]] naming CONTENT ref 1 ref 2 ref 3 A ‘State of affairs’
Standard minimalist architecture Compositional mapping LEX (=atoms) SEM 1 PHON 1 SEM 2 PHON 2 Sound PC PHON 3 SC SEM 3 ‘C-I systems’ ‘propositions’ ‘Thought’
The ‘Strong Minimalist’ Thesis Language is an optimal solution to conditions imposed by linguistic interfaces (Chomsky 2001) ► Two empirical hypotheses ► 1. ‘C-I incorporates a dual semantics’: (i) extended argument structure and (ii) discourse properties. ► Language ‘seeks to satisfy the duality in the optimal way, EM serving one function and IM the other’. 2. the ‘C-I interface requires the A/A’ distinction’ ‘if expressive potential is to be adequately utilized’. ► The syntax provides it (via the device of inheritance).
A more optimal model: all you add is Spell -out (Hinzen, 2006, 2007) List of atoms Phase-1 Phase-2 PHON 1 PHON 2 SM PHON 3 (SEM)
Does semantics come for free? ► Semanticity begins where adaptation begins ► But every human semantics is highly systematic and has a specific and restricted formal ontology. § Hard to see alternatives for it: Force Tense Event Participant
Problems of semantics Specificity ► Empiricity: ► § formal description is not to provide evidence for the descriptions used. Metaphysical status of ‘meanings’ ► Does the world out there really reflect the ontology of language? ► Do we have independent access to the semantic ontology? ► Insisting on a distinction between ‘syntax’ and ‘semantic’ on purely conceptual grounds? ► § Compare ‘mind’ and ‘body’.
In theory ► Our intuitions as to what ‘syntax’ and ‘semantics’ are, arise through a mixture of pre-theoretic intuitions and philosophical traditions. § The legacy of ‘anti-psychologism’ § The ‘externalist’ paradigm ‘Form’ vs. ‘content’ in ancient and medieval metaphysics
In practice ► Syntactic categories are standardly defined in a way that takes their semantic dimensions into account. ► Syntactic structures fundamentally sensitive to semantic distinctions. § Huang’s (1982) analysis of wh- movement in Chinese ►John thinks Mary bought what ►John wonders Mary bought what ►John remembers Mary bought what
A new heuristic: Perfect syntax-semantics alignment ► Syntax as the skeleton of thought § provides the forms of a possible thought (Hinzen 2006 a, 2007 a). ► Restrictions: § Lexical semantics § Associative thought § Discourse-driven semantic effects that are not fully systematic ► John fell. Bill pushed him. § Semantic effects that do not add to a system’s basic categorial ontology.
Basics: What is syntax? ► Def. : A internal computational system that assigns structural analyses to expressions. § Many two-term idioms: ►turncoat, fatcat, slapdash § No three-term idioms (Namiki 1994): ►turncoat brigade, Christmas tree cookie ► Emergence of a part-whole analysis on economy grounds (Snyder&Roeper 2003). ► Sentences are trees not chains.
► Fixed hierarchical sequence of projections reflecting semantic distinctions (Cinque 1999) § § § Evidently, I was probably once married I evidently was probably once married ? I was evidently probably once married ? ? I was probably evidently once married ? ? ? I was probably once evidently married ► Conditions imposed by the sensorimotor interface § E. g. linearity (Kayne 1994; Moro 2000) ► Constraints of computational economy (single- cycle generation) (Hinzen&Arsenijevic 2008)
How establish the alignment hypothesis? ► A logical problem ► Hypothesis predicts its own untestability. § Conceptual arguments § Refuting proposed syntax-semantics misalignments
Argument from generativity ► The ‘space of possible thoughts’ doesn’t have non-arbitrary, finite boundaries, any more than language does: thought is generative ► Generativity depends on combinatoriality ► If syntax is minimalist, how could the syntax of thought be different?
Argument from lack of independent access ► Carstairs-Mc. Carthy (1999) on the sentence/NP distinction ► Talking about semantic distinctions may merely be a fancy way of talking about formal-syntactic distinctions. § Pesetsky&Torrego 2001: ‘It is a truism that nouns pick out “persons, places, things and abstract ideas”, while verbs denote things like events and states. We suggest that this truism is simply a slightly less formal way of observing the difference between items whose features are lexically specified and items whose Tfeatures are lexically specified’
Argument from logicality ► Thoughts are structured by such relations as predication, modification, restriction, etc. ► These are linguistic notions and have been studied empirically in linguistic terms ► Being logical ones, how can there be a parallel combinatorial system in thought that has them too, but is different?
Argument from form-meaning correlations predicate composition adjuncts predication ‘bare’ Small Clauses events VP/v. P tensed events TP events placed in discourse CP kinds N reference DP event-participant argument adjective property etc.
On the origin of truth in language [Tom] [Fire!] [the Eiffel tower] [Caesar’s destruction of Syracuse] [You idiot!] [You, an idiot] [to be an idiot] [For John to love his mother] [if he is handsome] [John pale] [who has no money] Names Exclamatives NP NP Expressive Small Clause Appellative Small Clauses TP CP with non-finite Tense Modifier Small clause predication Relative clause
The growth of reference in acquisition ► The narrowing acquisition path (Roeper 2003) 1. 2. 3. 4. 5. 6. 7. 8. 9. Bare noun=pure kinds ► Want cookie; Milk § I like cats NP generic § § § § Non-specific NP existential I saw cats in the yard NP member of a set I saw a cat NP expletive definite (Marinis 2000) That's the way to play; all got into the hospital and left DP individual specific I saw that cat I saw John DP part-whole John’s leg DP discourse-familiar object I have a hat. The hat is green. Unique specific object focal stress I have THE hat Specific
Imperfections in the design of the syntaxsemantics interface? ► Jackendoff (1999): the syntax-semantics match is as ‘imperfect’ as that between syntax and phonology. § Is phonology compositional? (Wagner 2006)
Many-to-many relations between categories ► N: can denote § § objects (dog), situations (concert), places (region) times (Tuesday) ► V: can denote ► P: can denote § situations (to rain), § states (to sleep), § events (to jump) § times (on Tuesday) § places (in the house) § properties (in luck) (normally denoted by adjectives).
But intuitively impossible mappings exist A NP/DP like ‘Mary’s smile’ can’t locate this event in time, nor assert the fact that it took place (‘Mary smiles’). ► Impossible to use a VP, like ‘smiles’ or ‘kills Bill’, to refer to an object, like a smile or a killing of Bill. ► Nominals that denote an event/process also contain a VP (Fu et al. , 2001) ► § ? While Sam’s exploration of Easter Island was impressive, Amy’s doing so was a real surprise. § *While Sam’s trip to Easter Island was a fiasco, Amy’s doing so was a benefit. § While the removal of evidence purposefully is a crime, its removal unintentionally is not. § *His explanation of the problem fortunately to the tenants did not cause a riot. § John’s fortunate/*fortunately removal of the evidence saved me
Syntactic reflexes of semantic categorization ► Place-denoting nouns can appear in functional syntactic positions: § on top of the shelf or in front of my house ► Others cannot: ► Propositional attitude-denoting Ns take complement clauses, unlike nouns denoting objects, situations, times and places: § E. g. , on friend of § the fear that she won’t come) § *the dog that she won’t come). ► PP denoting properties are probably a result of idiomatization: § I’m in (*a/*the) luck. ► Overlap between denotations of NPs and PPs is imprecise, and disappears if PPs are just richer NPs (Asbury 2005).
Syntactic complexity tracks conceptual-semantic complexity: Verbs ► Vs can denote § states (e. g. , ‘lives’) ► Saddam lives; *Saddam is living; *Saddam lives in two months § unbounded activities (e. g. , ‘runs’) ► Saddam is running; *Saddam runs in two months § achievements (e. g. , ‘reaches’) ► Saddam is reaching his hideout; Saddam reached his hideout in two months; *the hideout’s reaching by Saddam § accomplishments (e. g. , ‘constructs’) ► Saddam is constructing a hut; Saddam constructed a hut in two months; the hut’s construction by Saddam
Syntactic complexity tracks conceptual-semantic complexity: Nouns ► Ns can denote § Abstract objects (‘justice’) ►*the justice; *justices § Masses (‘sand’) ►*one sand; *sands; *a sand § Countable individuals (‘glass’) ►a glass, half a glass, three glasses § Animate objects ►Animacy markers etc.
Jackendoff on Thematic Roles ► The same syntactic function (e. g. ‘direct object’), seemingly allows for a wide spectrum of Thematic roles: § a. Emily threw the ball. (Theme/Patient) b. Joe entered the room. (Goal) c. Emma emptied the sink. (Source/Patient) d. George helped the boys. (Beneficiary) e. The story annoyed Harry. (Experiencer) f. The audience applauded the clown. (? ? ? ) ► But syntactic behaviour of the relevant NPs isn’t the same: § The ball was thrown by Emily. vs. ? The room was entered by Joe. § The sink empties quickly. vs. *The clown applauds cheerfully. ► Methodological problem of deciding empirically what the Thematic Roles are except on a syntactic basis.
Syntactic constraints on thematic roles Thematic arguments relate to their head in a way adjuncts do not ► Verbs don’t have many arguments, and even conceptually highly relevant roles, like PATH, never show up as arguments ► universal list of thematic roles is short, possibly as short as the maximal number of verbal arguments. ► § Restrictions on the number of Thematic roles correlates with a restriction on argument structure types (Hale&Keyser 2002; Hinzen 2008 b) ► Thematic roles are structural notions, perhaps corresponding bi-uniquely to phrase-structural configurations
Coercion ► Syntactic coercion: § The clown laughed the children. § A wolf/Wolf came in. ► Semantic coercion (Pustejovsky 1995) § John began the book ►‘Begin’ selects for an event-complement, and a semantic type-mismatch results.
A syntactic solution to semantic coercion What is an event-description if not the denotation of, specifically, a verb phrase (‘writing a book’, ‘reading a book’, etc. )? ► Syntactic solution: ► -began V ► the book Event-denoting V-complements blocked with unaccusatives: § § § We began reading the book/We began to read the book *Reading the book began/*To read the book began We began the book/*The book began
Evidence against VP-insertion? ► Lower VP modification: § We finished eating the meal slowly § We finished the meal slowly ► Expected under V-incorporation: § We caused Bill to die slowly § We killed Bill slowly ► Passivization: § ? The book was begun to be written by the author § The book was begun by the author ► Expected under V-incorporation: § ? Bill was caused to be dead by us § Bill was killed by us
Is coercion compositional? ► Pylkkanen et al. (in press) compared MEG effects for: 1. COERCED: The client knew that the seamstress began the dress after her vacation 2. ANOMALOUS: The client knew that the seamstress captivated the dress after her vacation 3. CONTROL: The client knew that the seamstress sewed the dress after her vacation ► Neural N 400 generator (M 350) not responsible for composition problems in (1).
Evidence from f. MRI and ERP studies ► *He spread his warm bread with SOCKS. (Kutas&Hilyard 1980) ► *Dutch trains are SOUR (van Berkum&Hagoort 2004) ► *Trees can GREW (Ni et al. , 2001) ► *Max OF PROOF theorem (Friederici 2003) ► Gennari&Poeppel (2003) on conceptual semantic complexity of ‘love’ vs. ‘destroy’.
Phases as complete referential expressions = a unit of computation, to be completed before a new phase starts.
The DP phase REFERENT Zhe san ben shu DEM three CL book ‘These three units of (the kind) book’ DP Spec D’ D Num. P Num’ Num Cl. P Cl’ Cl NP PARTITION predicate KIND
Independent evidence for phases Phase ► [CP what did [TP John [v. P what [VP destroy what]]]] Edge Which of the papers that he gave Mary did every student [v. P ___ ask her to read ___ carefully]? *Which of the papers that he gave Mary did she [v. P ___ask every student to revise___]?