Evaluating Solution (1) What’s the predictive power? Does it rule out the following? Country X middle Country Y AM-left/PM-right Country Z Men-left/Women-right
Evaluating Solution (1) Each listing is a stipulation, thus no predictive power. Must generalize and make predictions!
Generative Approach, Solution (2) Principle: within a country, drive on x side only. Parameter: x = left/right Australiax = left Chinax = right Singaporex = left Taiwanx = right USAx = right etc.
Evaluating Generative Solution (2) Pretty good, but…. 1) each listing still a stipulation 2) a parameter always a disjunction
Evaluating Generative Solution (2) Research question: can we get rid of the parameter and the listings? The research is now theory-driven, rather than data-driven, as the data have been accounted for.
Evaluating Generative Solution (2) Expanding the scope of data: side of the road + side of the driver
Driving on the Left Right
Driving on the Right Left
The driving side is always the opposite of the driver side!!
Generative Approach, Solution (3) Principle: on Planet Earth, drive on the left, if the driver seat is on the right; otherwise, drive on the right.
Evaluating Generative Solution (3) Wow, no listings and no parameters!! But, wait! There’s still a disjunction.
Evaluating Generative Solution (3) Principle: on Planet Earth, if the driver seat is on the right, then drive on the left; otherwise, drive on the right.
Evaluating Generative Solution (3) Let’s again expand the scope of data: driver + passenger + center of the road
The driver is always closer to the center of the road!!
Generative Approach, Solution Ultimate Principle: when driving on Planet Earth, stay closer to the center of the road in relation to the front seat passenger.
Evaluating GG Solution Ultimate Does it allow a functional explanation? Yes, it does! Being closer to the center of the road affords the driver the best range of vision with the least physical strain
Evaluating GA Solution Ultimate It’s simple and elegant, but is it complete?
Evaluating GG Solution Ultimate Consider 建國高架橋下迴轉道 US Postman’s jeep And, of course, Myanmar!
Evaluating GG Solution Ultimate …the two kinds of linguists need each other. Or better, that the two kinds of linguists, wherever possible, should exist in the same body. (Fillmore 1992:35)
Evaluating GG Solution Ultimate Lessons from Myanmar and Pirahã.
Evaluating GG Solution Ultimate It’s simple and elegant, but how many countries do you really need to observe to derive it?
2) Contrasting LFG and TG 1.Motivation 2.Phrase structures 3.Grammatical features 4.Theta roles & linking 5.Summary & examples
Under the Generative Grammar, there are many competing frameworks: TG (incl. GB, MP…) LFG HPSG etc.
They share the same goal, but differ in: 1) what is “simple” exactly? 2) the right balance between descriptive adequacy and theoretical elegance Consequence: somewhat different architectures some different primitive notions
TG Principles: X-bar scheme for DS (spec rule) XP → YP, X’ (comp rule) X’ → ZP, X Parameters: (spec rule) YP > X’ or X’ > YP (comp rule) ZP > X or X > ZP
Extremist View (Kayne 1994) : Universal X-bar scheme with fixed order: spec > head > complement No PS parameters in DS!
TG DS → movements → SS
TG That, I don’t know t. John was kisses t.
LFG Single level c-structure Language-specific PSR allowed X-bar scheme as default
LFG That, I don’t know. John was kisses. No DS, no movements. WYSIWYG.
3. Grammatical Features e.g., case, number, person, etc.
TG Features grow on trees. Mary has kissed John [3/sg/nom] [3/sg/nom] …. …..
LFG Features & Grammatical Functions form an independent f-structure C-structure Mary has kissed John
4. Theta roles & linking
TG Theta roles are assigned to tree positions. kiss [x y] Mary has kissed John
LFG Theta roles, or argument roles, also form an independent a-structure, which is linked with the predicate’s f-structure kiss
5. Summary & examples
TG (1) John, Mary has kissed. kiss [x y] DS Mary has kissed John 3/sg/nom 3/sg/nom …. …..
TG (2) John, Mary has kissed. Movements John Mary has kissed t 3/sg/nom 3/sg/nom …. …..
TG (3) John, Mary has kissed. John Mary has kissed t [3/sg/nom] [3/sg/nom] …. ….. Feature checking
TG (4) John, Mary has kissed. SS John Mary has kissed t
LFG (1) John, Mary has kissed. c-structure John …. Mary has kissed ….
LFG (2) John, Mary has kissed. kiss f-structure
TG vs. LFG In a nutshell (1) TGLFG MovementsYesNo Grammatical functions NoYes
TG vs. LFG In a nutshell (2) TG: tree-centric theta roles and grammatical features are all part of the tree LFG: parallel planes argument structure, functional structure, and constituent structure are all independent
An Overview of LFG 1.Lexical entries 2.Phrase structure rules 3.C-structure 4.F-structure 5.Correspondence between c- and f-structure
1. Sample lexical entries time N flies V
2. Sample phrase structure rules S → NP : SUBJ VP VP → V NP : OBJ NP → N
3. Sample c-structure S NP :SUBJ VP N V time flies
4. Sample f-structure S NP :SUBJ VP N V time flies
5. Correspondence between c- and f-structure S NP :SUBJ VP N V time flies
Some of LFG’s Motivations 1.Lexical integrity 2.Non-configurationality 3.Movement paradoxes 4.Lexical processes over movements
1.Lexical Integrity Lexical Integrity Hypothesis (Huang 1984) No phrase-level rule may affect a proper subpart of a word. Ex: I like singing and dancing → *I like [sing and dance]-ing. You speak and I do too. → *He is a singer and I do too.
TG Mary went. Mary /ed/ go Affix Hopping Violating lexical integrity.
LFG Mary went. Mary went Maintaining lexical integrity.
2. Non-configurationality English is a configurational language, where grammatical relations (e.g., SUBJ, OBJ) are largely encoded by the configuration of the constituent structure. There are, however, non-configurational languages, where grammatical relations are largely encoded by morphological means.
Language Typology 101 V i : S V t : A P V i : S V t : A P Case can be marked structurally or morphologically! → Accusative → Nominative (unmarked) language Ergativelanguage → Absolutive (unmarked)
English Subj Obj Mary has kissed John John has kissed Mary Case marked by structural configuration.
Malayalam Case marked by affixes. Yes, I speak Malayalam.
Malayalam 1. Kutti aana-ye kantu(SOV) child.NOM elephant-ACC saw 2. kutti kantu aana-ye(SVO) 3. aana-ye kutti kantu(OSV) 4. aana-ye kantu kutti (OVS) 5. kantu kutti aana-ye (VSO) 6. kantu aana-ye kutti (VOS) Case marked by affixes. ψ
Malayalam (TG) kutti aana-ye kantu (LFG) kutti aana-ye kantu Which is simpler? (lots of movements!) (fixed DS, fixed order) (no DS, no ordering) (no movements!)
Malayalam F-structure for all six word orders
Warlpiri The two small children are chasing that dog. wita-jarra- kurdu-jarra- small-DUAL-ERG child-DUAL-ERG ka-pala wajili-pi-nyi pres-3duSUBJ chase-NPAST yalumpu maliki that.ABS dog.ABS rlurlu ψψ
Warlpiri Word order: Free Constraints: 1) 1st position must be a constituent 2) 2nd position must be T (AUX) Examples: 1) [that.ABS dog.ABS] NP T chase children-ERG small-ERG 2) [dog.ABS] N T children-ERG chase small-ERG that.ABS 3) [chase] V T children-ERG dog.ABS small-ERG that.ABS 4) *[T] T chase small-ERG children-ERG that.ABS dog.ABS 5) *[small-ERG dog.ABS] *C T chase children-ERG that.ABS
Warlpiri TG (same as English) NP T VP Consequence: lots of movements Prediction: Warlpiri, like Eng, has VP Test: [chase dog.ABS] VP T children-ERG Result: Warlpiri has no VP! *
Warlpiri LFG TP → C T C* C T C... Typology: X-bar vs. W-star Cause: morphology competes with syntax
3. Movement paradoxes 1.a. *The theory does explain. b. The theory does explain that mass is energy. c. That mass is energy, the theory does explain t. 2.a. *The theory does capture. b. * The theory does capture that mass is energy. c. That mass is energy, the theory does explain t.
3. Movement paradoxes 1.a. You are not a student. b. Are you not a student? c. You aren’t a student. d. Aren’t you a student? 2.a. I am not a student. b. Am I not a student? c. * I aren’t a student. d. Aren’t I a student?
3. Movement paradoxes 1.a.* 他最擅長. b. 他最擅長語言學. c. 語言學，他最擅長 t. 2.a.* 他最拿手. b.* 他最拿手語言學. c. 語言學，他最拿手 t.
3. Movement paradoxes TG: mismatches are unexpected, because the source and the target of movement must be identical. LFG: mismatches are expected, because there is no movement and mapping between two planes (e.g., c- and f-structure) is not one-to-one.
4. Lexical processes over movements Participle verbs (present, perfect, passive) in English may convert to adjectives. 1. a very disturbed market. (passive) 2. a well-prepared student. (perfect) 3. an all smiling bride. (present) Particle V → A
4. Lexical processes over movements happy [x] TG was happy John Prediction: V[x] → A[x], x undergoes movement
4. Lexical processes over movements True for passive and unaccusative verbs disturbed [x y] TG was disturbed the market
4. Lexical processes over movements Not true for unergative verbs prepared [x] TG John has prepared well V[x] → A[x], x undergoes no movement
4. Lexical processes over movements happy LFG John was happy Prediction: V[x] → A[x]
4. Lexical processes over movements True for all intransitive participle verbs. disturbed LFG The market was disturbed
3) CONCLUSION The air-mattress metaphor
Corpus Approach vs. Generative Approach and Movement vs. Grammatical Functions One-Soon Her 何萬順