CAS LX 500 Topics in Linguistics Language Universals Classification Fall 2000 December 12, 2000 Paul Hagstrom Week 15: Summary Caveat: In many cases, it is not entirely true to say that a language can be classified in one of these groups—often a language will be dominantly in one An overview of what we’ve seen… category but show characteristics of another category in a small part of its grammar. This strengthens the case for a “universal inventory” of language Underlying plot: There are lots of languages, and they differ from one another. components. but they do not differ randomly. The range of “possible human languages” is not unbounded (structurally). What we have done Morphology: Analytic MANDARIN Polysynthetic SIBERIAN YUPIK is explore a sample of how languages can differ and how they can’t. Agglutinative KOREAN Fusional RUSSIAN (agr/case) Classification: 4000Ð8000 languages, which can be grouped in various ways. (Structural characteristics, historical relatedness, areal proximity) Basic word order: SOV most common JAPANESE SVO very common ENGLISH Structural characteristics that hold of all languages: absolute universals VSO common IRISH Structural characteristics that hold of most languages: universal tendencies VOS rare MALAGASY Conditional characteristics: If L has X, it will have Y: implicational universals OVS very rare HIXKARYANA OSV very rare ? The topics we’ve covered: V2 common GERMAN 1. Classification and implicational universals, innateness explanation 2. Hierarchy of color terms, hierarchy of vowels Case-marking: Nominative/Accusative SA/O most common ENGLISH 3. Pidgins and creoles, Nicaraguan Sign Language Ergative/Absolutive A/SO less common DYIRBAL 4. L1 acquisition; phonological perception, babbling, one word stage S=intransitive subject A=transitive subject O=transitive object 5. Optional , (verb raising), SLI, Korean negation 6. Case errors during L1 acquisition, Tense and agreement in French Markedness Unmarked features are the most statistically common (≈“default”). 7. Structure-building in L1 & L2 acquisition (syntax: V2, VPISH) Marked features are the most statistically rare. 8. L2 acquisition; Case-drop in Japanese and accessibility of ECP. The vowel /a/ is in pretty much every language there is. The vowel 9. accessibility hierarchy /π/ is in relatively few languages. /a/ is unmarked relative to /π/. 10. Ergativity, Mahajan’s P-incorporation account Often, this turns into an implicational universal; if a language has 11. Optimality Theory, syllables and prosody, sonority hierarchy the vowel /π/, it will have the vowel /a/. If a language has nasalized 12. Passamaquoddy and Mohawk unstressable syllables vowels, then it will have non-nasalized vowels; hence [+nasal] is 13. Question formation, wh-movement, wh-islands, Irish English Q-float marked (relative to [Ðnasal]). 14. Particle movement in Japanese and Sinhala 15. Semantics of questions, briefly 16. hierarchy Universals 17. Functional projections; AgrP in French 18. hierarchy and more functional projections Patterns have been observed, looking at a wide range of languages. Some implicational 19. Universal base order universals (If a language L has X, it will have Y), but some more powerful universals in terms of hierarchies (In an ordered list, if a language has a feature X, it will have all of the less marked features on the hierarchy as well). Major “take-home” points of each topic Why do languages conform to universal implications? Color and Vowel Hierarchies (Berlin & Kay 1969, Crothers 1978)

Several reasons: ¥ Functional pressures Color foci: If a language has, e.g., green, it will have red, white, and black. —Some sounds are hard to make; the harder a sound is to make, the less likely it is to be part of a language’s purple phonological repertoire. (1) white < red < green < blue < brown < pink —Choice of vowels maximizes phonetic distance in order black yellow orange to aid perception. gray —Color receptors in the human eye might make certain divisions of the color space more differentiated than others Interestingly: Color foci were reliable, crosslinguistically, across informants, across trials. in terms of perception (and you name colors you can Color boundaries were extremely unreliable… differentiate) —Subjects tend to be “more important to the message” ☞ It is important to look at the data in the right way to see the universals. than objects, hence would come first (Often, this requires consideration of the abstract syntactic structure, for example) ¥ Genetic origin —If the parent language had a feature, it’s often likely Hierarchy for vowel inventories (based on 209 languages studied by Crothers): that languages descended from it will have the feature. • “Language instinct” (2) /i a u/ > /π/>/´/>/ø/>/\/ > /e/ > /o/ (roughly) —Part of the makeup of the human brain leads us to have language, but it is structured in a particular way that Dispersion. only allows certain languages. ¥ Colors appear to be arranged in the color space so that their foci are maximally distinct. ¥ Vowels appear to be arranged in the color space so that their foci are maximally distinct. A priori, we don’t know which is the “root cause” for any given universal, but the most interesting ones are those which give us insight into the “language instinct” (“Universal Grammar”) that humans come with. (These are the ones we’ve mainly been looking at) Pidgins and Creoles (Bickerton) Implicational universals: p → q. Pidgin. Speech forms which do not have native speakers, primarily used as a means of Meaning: There are languages with p & q, ¬p & q, ¬p & ¬q communication among people who do not share a common language. Unsystematic. but (nearly) no languages like p & ¬q. Creole languages (creoles). Has native speakers, acquired with a pidgin as input. Because Good ones ¥ three of the four possibilities must be attested the input has no discernible structure, children have to impose structure (of human ¥ one must be unattested (since it would be a counterexample) language, as dictated by UG) on the input—and they do. Children’s language (creole) is ¥ attested languages should be reasonably numerous. systematic, and includes innovations not found in the pidgin or the substrate languages. Moreover, the innovations are substantially similar across creolization situations. ¥ All languages have vowels Non-implicational absolute ¥ If a language has distinct 1/2 reflexives, it has 3 reflexives Implicational absolute Articles: HCE ¥ Nearly all languages have nasal consonants Non-implicational tendency Definite specific NP, presupposed the boat da [some Salish languages don’t] Indefinite specific NP, asserted a certain boat wan ¥ If a language is SOV it is probably postpositional Implicational tendency ¯ nonspecific NP some boat ¯ [Persian is SOV with prepositions] TenseÐModalityÐAspect (TMA) systems: Basic syntactic structure (of course, more said about this in the later classes) Order: Tense Modality Aspect HCE (3) a. She will leave. bin Tense particle expresses ¥ past for stative (was hungry) b. She will not leave. ¥ past before past for action verbs (had walked) c. She left. This one is the oddball. go Modality particle expresses ¥ irrealis (futures will eat and conditionals if we eat) d. She did not leave. stei Aspect particle expresses ¥ nonpunctual (progressive, habitual, iterative) So, we have evidence for the following positions in our sentences: Sentence : (4) subject tense/agreement (not) verb (object) fo ¥ One for realized actions (I managed C1 to stop, I saw C1 John leave). go ¥ A different one for hypothetical actions (I wanted C2 John to leave) (5) XP (6) IP 3 3 Spec X′ NP I′ as verbs (no to be): 3 She ¥ Bill intelligent (vs. Bill is intelligent) 3 X Comp IVP Questions: will 1 ′ ¥ Yes-no: no word order changes, often optionally a sentence-final particle. V ¥ Wh: Question word first, no other word order changes. 3 VNP Serial verb constructions: (common) eat lunch ¥ Multiple verbs with one subject, at most one object, one tense/aspect/negation (7) IP Past tense Ðed is a suffix, so it is pronounced (no coordinators and or subordinators that or intervening pauses) 3 with the verb to make “ate” NP I′ Nicaraguan Sign Language: (Kegl et al) She a little like French: à + le → au Arguably a pure creole situation—pidgin is unsystematic hodgepodge of home signs, 3 → creole is what the kids in the schools for the deaf wound up speaking (innovative, IVP de + le du → increased fluency). Also indicates that signed languages are governed by the same internal -ed 1 à + les aux language system as spoken languages (signed languages are real languages). Creole V′ doesn’t seem to share all of the same properties as the spoken creoles (though it seems to 3 have serial verb constructions), but some of this may have to do with modality VNP (agglutinative morphology, for example). eat lunch

Critical period: “She ate lunch” Children younger than 7 years old generally were able to acquire the creole; older signers (8) a. I know [ that she will eat lunch ]. were not able to “fill in the holes” in the grammar like the younger acquirers. b. I wonder [ if she ill eat lunch ]. c. Will she eat lunch ? Conclusion: It doesn’t take language to make language (something must be innate). d. What will she eat? This gives us evidence for another position (in fact another XP) before the subject: (9) subject tense/agreement verb object (10) CP (11) CP We also have some reason to believe that the subject doesn’t start in SpecIP but rather 1 1 starts in SpecVP and moves to SpecIP: C′ C′ 3 3 (13) a. All the students will leave. (14) IP b. The students will all leave. qp CIP C+Ii IP c. * The students will leave all. NP I′ that/if 3 will 3 i NP I′ NP I′ the students 3 she 3 she 3 IVP IVP will 3 ti VP NP V′ will 1 1 !1 V′ V′ all t V 3 3 i VNP VNP leave eat lunch eat lunch

(12) CP Word order variation and syntactic parameters 3 ′ NPi C It seems to be possible to describe the structure of all languages in these structural terms. what 3 That is, the structure is basically universal, but certain parameters can differ from C+Ij IP language to language. will 3 NP I′ One parameter: Order of O and V in terms of order of V and NP under V′ she 3 t VP (15) SVO: … SOV: … j U U 1 V′ V′ ′ V 33 3 VNP NPV V ti verb object object verb eat And possibly another parameter: Order of S and VO in terms of order of I′ and NP.

Lexical categories: Noun, verbs, adjectives, —“open class” items. (16)SVO/SOV: … VOS/OVS: … (Things you can make new ones of—Xerox, Xeroxed, etc.) U U IP IP Functional categories: Complementizers, tense, modals (might, can, should, could), 33 (the, a). NP I′ I′ NP subject RU RU subject So: CP, IP are functional projections. VP is a lexical projection. (“projection” = “XP”). (17) CP A lot of languages, for example French, So, this view says that the trees are basically provided by UG (CP, IP, VP) and that qp move the verb up the tree to combine languages differ only in the order of things in the underlying tree and what they move. C IP with Infl. NOTE: This generally happens ¥ verb-object order (order of branches under V′) [SVO, SOV] (that) qp with finite verbs—often NP I′ nonfinite verbs do not move. ¥ verb-subject order (order of branches under IP) [SVO, VOS] she qp ¥ whether V moves to Infl [English vs. French] Infl+Verb VP ¥ whether Infl moves to C [SVO, VSO] [pres 3sg] qp ¥ whether something has to be to the left of C [V2] tells Adv VP 3 In French, you can see that finite verbs move and nonfinite verbs do not by looking at the t V′ order of negation (pas) and the verb. 3 t NP the truth (22) IP qp ′ (18) a. * She tells always the truth. (19) a. Elle dit toujours la vérité. NP I b. She always tells the truth. b. * Elle toujours dit la vérité. subject qp I+V NegP (20) VSO: CP Some languages move the verb to Infl "qp 1 Neg VP 3 and move the whole thing higher: VSO vs. SVO: not C+Infl+V IP 1 qp 1 pas tV NP buys 3 object NP I′ z------m John 3 t VP (23) subject Vfinite+Infl neg object 3 (24) subject (Infl) neg Vnonfinite object t V′ 3 An example of a “head-final” language (NP precedes V under V′, VP precedes I under I′, t NP IP precedes C under C′). lunch (25) Chelswu-ka chayk-ul sa-ess-ni ? Korean (21) V2: CP Some languages (like German) do this too, Chelswu-NOM book-ACC buy-past-Q 3 but also require that something move ‘Did Chelswu buy a book?’ yesterday C′ to the left of C (leaving the verb in 3 “second position”). (Like in English (26) CP C+Infl+V IP wh-questions) C′ buys 3 qp NP I′ IP C in CP, pronounce TP before C John 3 t VP qp ni 3 NP I′ in IP, prounounce NP before I′ t VP Chelswu-ka qp 3 VP I in I′, pronounce VP before I t V′ V′ ess in V′, pronounce NP before V 3 t NP qp lunch NP V chayk-ul sa The “Split INFL” hypothesis—functional heads TP and AgrP

In French, we found that: All finite verbs appear left of negation (raising to INFL). Nonfinite auxiliaries can appear left of negation, Nonfinite main verbs cannot appear left of negation.

(27) V/Auxfinite Neg t Auxnonfinite Neg t *Vnonfinite Neg t

But that: All finite verbs appear left of adverbs Nonfinite auxiliaries can appear left of adverbs Nonfinite main verbs can appear left of adverbs

(28) Ne pas arriver souvent en retard, c’est important. ne not arrive often late it is important ‘It is important not to often arrive late.’ There is a position between the adverb souvent and Neg pas where the verb can move.

The hypothesis: “IP” is really two projections “TP” (tense) and “AgrP” (agreement) The morphology on the verb suggests that AgrP is the higher of the two because the tense morphology is closer to the verb stem:

(29) NP V T Agr (30) NP V T Agr a. je parl ai s a. je parl er ai b. tu parl ai s b. tu parl er as c. il parl ai t c. il parl er a speak PAST person+# speak FUT person+#

If the moves to one and then the other, it would look like this:

(31) AgrP (32) AgrP 1 1 Agr′ Agr′ 3 3 Agr TP [V+T]+Agr TP 1 1 T′ T′ 3 3 V+T VP t VP 3 3 V′ V′ 3 3 t ... t ... So, in [V+T]+Agr, T is closer to the verb and is the head that was moved to first. The optional stage (around 2;0): a. There are main with finite verbs. Mirror Principle. Morphological derivations (suffixation, prefixation, etc.) b. There are main clauses with nonfinite verbs. directly reflect syntactic derivations... c. Children move the verb as required in the adult language for each type of verb.

First Language Acquisition French kids around 2;0 [+finite] [Ðfinite] pas verb 11 77 (would be adult like) Children universally seem to go through a number of ‘stages’ as they acquire language: verb pas 185 (adult like)2 ¥ Preliminary stage (0-1;0) (prelinguistic development, babbling) ¥ Single word stage (1;0Ð1;6) ¥ Two word stage (1;6Ð2;0) (simple sentences, vocabulary explosion) German kids around 2;0 [+finite] [Ðfinite] ¥ Simple sentences V-final 11 37 (would be adult like) ¥ Complete sentences (embedding) V2 197 (adult like)6 Phonological perception Conclusion: Children know about tense, and they know what effect tense has on word order. What’s different between kids and adults seems to be that they don’t At a very early age, infants seem to be able to discriminate both native & non-native always produce finite verbs, but when they produce a nonfinite verb, it is contrasts. As they grow older, they become less able to do so (presumably as they tune treated like a nonfinite verb would be in the adult language. their phonological system to the language they are learning)

h h Other languages for which there is evidence of an OI stage: All Study on Hindi/English contrasts /ba/~/da/ /d a/~/t a/ /ta/~/Ta/: studied to date (Danish, Dutch, English, Faroese, Icelandic, Norwegian, Swedish, ¥ nearly all the kids 6Ð8 mo. could discriminate non-English contrasts Faroese). Also, French, Irish, Russian, Brazilian Portuguese, Czech, Hebrew. ¥ 8-10 mo. kids showed decrease in ability (mixed performance) ¥ by 10Ð12 mo. kids do much worse, like adults Very Early Parameter Setting. From the earliest observable stage, that is, from the time that children produce multi-word utterances, they have correctly set the basic Babbling inflectional/ structure parameters. These include the V-to-TNS (verb raising) parameter, the V2 parameter, and parameters of basic word order , including the relative ¥ 4Ð6 months, isolated consonant & vowel type sounds order of the verb and the object. ¥ 6 months, reduplicative CV syllables. ¥ 10 months, wider range of syllable types (CVC, VC), Optional Infinitive Stage? Tense is optional. syllable sequences with more varied members. (Everything else [UG, language-particular grammar] is known to the child)

Two-word stage (1;6Ð2;0) or so Maturation. The hypothesis put forward by Wexler is that this is an example of maturation, analogous to losing one’s baby teeth and growing new ones. The part of First evidence of syntactic development (one word couldn’t give us evidence of this). grammar that requires tense in an adult utterance is not in place yet in the child’s grammar, Tremendous vocabulary explosion. though pretty much the rest of it is.

Maturation and optional infinitives (Wexler et al.) Null subjects: In many languages it is possible to leave the subject out if it is recoverable (where English would use a like he, she, they). Often this has to do with the fact Re: interaction between (tense marking, agreement) and syntax (word order). that the verb is marked with the person of the subject. These languages are null subject languages. Kids around 2;0 produce a lot of null subjects, but they are much more frequent with nonfinite verbs, and even in non-null subject languages you can get null subjects in Nearly all of the time kids use ACC, they’ve also used a nonfinite verb. nonfinite clauses. Further evidence that kids know how tense and agreement works (and Idea: If kids leave out either TP or AgrP, the verb is nonfinite, how it interacts with null subjects). but only if they leave out AgrP is there a Case error.

(33) a. I want [— to leave]. Leaving is to be done by me. Optional Infinitive Stage? Tense and/or agreement is optional. b. I want [Bill to leave]. Growing trees in first language acquisition Null Subject/Optional Infinitive generalization: Only non-null subject languages seem to show optional infinitives around age 2;0. (No OI’s in Italian, Spanish, Catalan, Tamil, Another explanation for the Optional Infinitive stage (other than that tense and/or Polish). Hebrew: only allows null subjects in 1st/2nd person, non-present, and everywhere agreement is optionally omissible in an otherwise full syntactic tree): Kids start with else we find OI’s in kids’ speech. “small trees” (VP only) and gradually add structure. (We might augment this with a separate TP stage and an AgrP stage—the L2 results in the next section suggest this) Specific Language Impairment (SLI). At least a certain form seems to be explainable as an “extended optional infinitive” stage—kids who persist in using nonfinite main clause (37) a. VP b. IP verbs longer than usual. Further evidence of a maturation-type account? 33 NP V′ NP I′ Other reflexes of OI stage? Korean kids make an error with one form of negation which my 3 I 3 happens at about the same age as OI’s happen in other languages. Same root cause? V NP Infl VP make a house 3 t V′ Subject Case errors 3 VNP Default case: In sort of syntactically impoverished environments (no verb), English: ACC. color me Russian, Dutch, German, Faroese would all use NOM in such environments (NOM default). c. IP d. CP 33 (34) a. (Who did it?) Me. (35) Der, den habe ich gesehen. NP I′ NP C′ b. It’s us. It’s me. He, him have I seen what 3 where 3 c. Me too. ‘Him, him I saw.’ Infl VP C IP d. Me, I like pizza. 33 NP V′ NP I′ English kids sometimes make errors in the Case of the subject, in these other languages, my 3 I 3 kids don’t seem to make errors. V t Infl VP making 3 ′ (36) a. I get Bozo. (A 2;3) d. Him fall down. (N 2;3.14) t V 3 b. Me get John. (A 2;3) e. Her have a big mouth. (N 2;2.6) VNP c. My see that. (A 2;3) put it

Finiteness vs. case Stage (c) is the best evidence for this view; kids producing their early wh-questions subject Finite Nonfinite consistently produce non-nominative subjects. On which one is right, the jury is still out. he+she 436 75 him+her 4 28 % non-NOM 0.9% 27% Second language acquisition as tree-growing (Vainikka & Young-Scholten) (39) a. John ga sono hon o yonda. b. * John ¯ sono hon o yonda. John nom that book acc read ‘John read that book.’ c. John ga sono hon ¯ yonda. Transfer of parameter settings from L1. Native speakers of SOV languages learning German (naturalistically) start speaking German as SOV (essentially correctly, modulo Two different effects from the same universal principle. L2 native speakers of English V2), while native speakers of SVO languages start speaking German as SVO (incorrectly) learning Japanese, untrained in the rules for Case marker dropping, nevertheless perform and then move to SOV. correctly—they allow dropping the object Case marker, but not the subject Case marker.

Tree growing. L2A starts at a VP stage? No verb raising (verbs follow negation and From what the students did see, it would have been difficult to see the generalization: adverbs), no modals (which start in INFL), no agreement, no complementizers (which start ¥ Nominative markers can be dropped on objects of certain verbs. in C), no wh-movement, verbs follow negation and adverbs. • Topic markers (on logical subjects) can be dropped—just not subject markers. ¥ The textbook just said that dropping case markers reduces emphasis, no hint Next stage TP stage? Verb movement, some auxiliaries and modals, but still no at the restriction. agreement, complementizers, or wh-movement. The conclusion is: The students were not taught this, nor were they given evidence Next stage AgrP stage? CP stage? Verb raising common, auxiliaries and modals sufficient to figure it out. Nevertheless, they appear to adhere to it. common, agreement acquired, some embedded clauses with complementizers, complex wh-questions. Universal Grammar constrains first languages to be languages which respect the ECP. It appears that UG also constrains what languages are “available” to L2 learners. Infinitives? L1 kids show correlation between finiteness and position—L2 adults don’t. L2 learners: When they use the finite form, it is treated syntactically as finite. The Noun Phrase Accessibility Hierarchy Where a finite form is required, they treat the verb as finite, but sometimes use the nonfinite form of the verb. Where a nonfinite form is required, they use the nonfinite form correctly. (40) The man [(who) I met]. head noun ----m z---- in L1A, when a verb is finite (underlyingly) it is pronounced and treated as finite, and when a verb is nonfinite (underlyingly) it is pronounced and treated as nonfinite. Rel. clauses categorized by what role the head noun would play inside the relative clause: in L2A, when a verb is finite (underlyingly) it is treated as finite, but sometimes Subject relatives: The man [who — gave me a newspaper] pronounced as finite and sometimes pronounced as nonfinite. When a verb is nonfinite Object relatives: The man [(who(m)) I met — yesterday] (underlyingly) it is pronounced and treated as nonfinite. Indirect object relatives: The man [(who(m)) I gave the newspaper to —] Object of a preposition: The man [(who(m)) we have been arguing about —] Genitive: The man [whose house I bought —] Universal principles and Case drop in Japanese Object of a comparative ( -er than): The man [who I am taller than —] Syntactic research has identified a principle that appears to be operative in all languages, Subject > Direct Object > Indirect Object > Object of P > Genitive > Comparative Object the Empty Category Principle. In English, it accounts for (38)—moving a wh-subject out from under that is ungrammatical. This is the Noun Phrase Accessibility Hierarchy. The subject is “most accessible” (easiest) for relativization. (38) a. Who did John predict ti would win the election? b. * Who did John predict that ti would win the election? The object of a comparative is “least accessible” (hardest) for relativization. In Japanese, the ECP allows dropping of the object Case marker, but not the subject Case If a language can relativize position X in the hierarchy, marker. it can relativize all positions higher (to the left) on the hierarchy. Nominative/Accusative vs. Ergative/Absolutive Telling apart the NOM and ACC in Slavic. Old Church Slavonic: plurals have different forms in the masculine. S = intransitive subject A = transitive subject O = transitive object Czech: plurals have different forms in the animate masculine. Russian plurals have different forms in the animate. Nominative/Accusative: S and A (NOMINATIVE) vs. O (ACCUSATIVE) Ergative/Absolutive: S and O (ABSOULUTIVE) vs. A (ERGATIVE) Deciding in Czech whether something can be a possessor (“decision tree”). * means impossible as a possessor. The different one is usually marked with an overt case marker: Agreement marking can also be nominative/accusative or ergative/absolutive. NPs qp Ellipsis processes also tend to allow only interpretation within the group. qp *Neuter Masc, Fem (41) [I returned] and [— saw him] If — is interpreted as I, NOM/ACC pattern qp S A *Plural Singular qp (42) [I saw him] and [— returned] If — is interpreted as him, ERG/ABS pattern *Inanimate Animate O S qp *Not uniquely Uniquely identifiable Languages are often not fully ergative or fully nominative. Dyirbal is syntactically identifiable ergative, and generally ergative, but in the first and second person plural pronouns, the overt markings are NOM/ACC. English is almost everywhere NOM/ACC, but there are tiny There seem to be several hierarchies and they interact (they are independent hierarchies): pockets of ERG/ABS: Employee, Employer, Escapee (-ee on S and O, -er on A). Nominal hierarchy: pronouns > proper nouns > kinship terms > common nouns Animacy hierarchy: human > nonhuman animates > inanimates Animacy hierarchy Gender hierarchy: masculine > feminine > neuter Number hierarchy: singular > plural The type of noun can affect whether it is marked according to a NOM/ACC system or Definiteness hierarchy: uniquely identifiable > not uniquely identifiable according to an ERG/ABS system.

Arabana Thargari Dyirbal Syllables and sonority 1st/2nd person NOM/ACC NOM/ACC NOM/ACC 3rd person NOM/ACC NOM/ACC ERG/ABS Syllable types:Every language has CV syllables. human NOM/ACC NOM/ACC ERG/ABS If a language has VC syllables, it has V syllables. animate nonhuman NOM/ACC ERG/ABS ERG/ABS The more complex the syllable edges (CVCC, CCVCC, …), the less likely. inanimate ERG/ABS ERG/ABS ERG/ABS Vowel length possibilities: Every language has short vowels. That is, there seems to be an animacy hierarchy such that nouns which are “more If a language has super-heavy vowels (3 timing slots), it animate” are more likely to be marked according to a NOM/ACC system, whereas nouns has heavy vowels (2 timing slots) which are “less animate” are more likely to be marked according to an ERG/ABS system. No language has super-super-heavy vowels (4 timing slots) Syllables and weight. A heavy syllable (e.g., a syllable with a long vowel) is represented (45) Sets of syllabic segments as having two weight units (moras), and a light syllable has just one weight unit (mora). a. vowels Lithuanian, Bulgarian b. vowels, sonorants English, Gonja σσ c. vowels, sonorants, obstruents Imdlawn Tashlhiyt Berber 13 µµµµ = mora (46) σσσ RgU RgU RgU Light σ Heavy σ µµµ 111 (43) σσσ ald g gi gi µ µµ µµ Berber Berber Berber nucleus English English g onset gr g g Bulgarian ta ta tap (47) Sets of moraic segments ta taa tap a. vowels Khalkha Mongolian, Yindiµ Light σ Heavy σ Heavy σ coda b. vowels, sonorants Lithuanian, Tiv Open σ (CV) Open σ (CV) Closed σ (CVC) c. vowels, sonorants, obstruents English, Latin, Cairene Arabic

The syllable inventory of a language seems to make reference to a hierarchy of sonority. (48) σσσ Rgu Rgu Rgu Sonority: 1. voiceless stops µµ µµ µµ voiceless fricatives obstruents 11 11 11 2. voiced stops ai al ad 3. voiced fricatives English English English 4. voiced nasals Lithuanian Lithuanian voiced laterals sonorants (m, n, r, l) Khalkha Mongolian 5. voiced r-sounds 6. voiced high vowels vowels Jespersen: In every group of sounds there are just as many syllables as there are clear 7. voiced mid vowels relative peaks of sonority. 8. voiced low vowels Sonority Sequencing Principle: ¥ Moraicity constraint (what sonority is required to be a mora) Between any member of a syllable and the syllable peak, only sounds of higher sonority ¥ Syllabicity constraint (what sonority is required to be a syllable nucleus) rank are permitted.

(44) most sonorous a (σ constraint is always at least as high Syllable Contact Law (some languages show effects of this, not all) µ i as , since a syllable nucleus is In any sequence Ca.Cb, there is a preference for Ca to exceed Cb in sonority. \ both a mora and a syllable.) m  syllabicity constraint (σ) ENGLISH Optimality Theory s d Optimality Theory. Grammar is a set of ordered constraints… least sonorous t  moraicity constraint (µ) ENGLISH (49) NO-EPENTHESIS (‘no adding consonants or vowels’) Faithfulness Prosodic structure and word stress NO-DELETION (‘no deleting consonants or vowels’)

(50) ONSET: *[σ V… (‘syllables must start with a consonant’) Markedness (51) a. word word NO-CODA: *…V]σ (‘syllables must not end with a consonant’) 31 FS FW F 11 3 Underlying representation: /kæt/ σσ σσ cón tèst tém pest “kæt” violates NO-CODA, satisfies ONSET, satisfies NO-EP and NO-DEL. “kæ” satisfies NO-CODA, ONSET, and NO-EP, but violates NO-DEL. The intuition about stress in words is that it indicates groupings of syllables (the feet), each “kæt\” satisfies NO-CODA, ONSET, and NO-DEL, but violates NO-EP. group having a prominent syllable (a head).

The idea of Optimality Theory: Which one you say is a function of which constraints are Basic parameters of word stress: more important in your language. Boundedness: Feet in a language can either be bounded (at most 2 syllables), or Underlying form Constraints, in order. Check the first one first, unbounded (any number of syllables). stored in your lexicon anything that survives is checked on the second, … Foot dominance: The side of the foot where the head is (the strong element). /kæt/ NO-CODA ONSET NO-DEL NO-EP kæt * ✔✔✔ Bounded left-dominant feet are trochees (X .), bounded right-headed feed are iambs ( . X) kæ ✔✔* ✔ kæt\ ✔✔✔* Quantity sensitivity: Whether feet count moras separately or just look at syllables. Heavy Q-insensitive (no difference between H and L) The hand Possible ways you Light Q-sensitive (H can’t be in a weak position of a foot) points to could pronounce it Q-determined (every foot needs to dominate an H) the winner (“candidates”) Iterativity Whether each word has one foot or repeating feet Capturing universals in Optimality Theory Re-ranking these constraints (24 possibilities) only yields 7 patterns. Directionality Whether the “first” foot is at the beginning or the end of the word. We can get CV languages, (C)V language, CV(C) languages, (C)V(C) languages No re-ordering of the constraints would yield: V only VC only Some crosslinguistic preferences V(C) only CVC only (52) Stressability: Words have to be big enough to carry a stress. Universal Grammar in Optimality Theory: (Usually a long syllable [two moras] or a two-syllable foot). Constraints are universal (all languages have them—this is “UG”) Grammars differ only in the ranking of the constraints across languages. (53) Demarcation: Words should have stress near the edge. The child’s task during acquisition is to acquire the rankings. (Primary stress is usually first, last, or penultimate)

(54) Rhythm: Bounded. Avoiding lapses and clashes (alternating stress) Prosodic structure: PrWd Elements of prosodic structure This suggests that in Czech, one wh-word goes to SpecCP (like in English), and the rest 1 are analogous to elements of syntactic adjoin to IP like adverbs: Foot structure when describing the structure 1 of sentences. σ (56) CP qp 1 ′ µ Spec C kdo 3 1 CIP x (ho) rU Adv IP PrWd rychle rU Ft Ft co IP rU σ σ σ komu IP µµµ µ ¥ Parm Q: Question needs a wh-word in SpecCP p í s k %« lan ‘It rains so hard that it is dark or hard to see’ (Passamaquoddy) ¥ Parm W: All wh-words must be in front (near C—SpecCP or the edge of IP) ¥ Parm I: Whether wh-words can attach to IP Question formation English: [+Q ÐW ÐI] A (rough) typology of (overt) wh-movement Bulgarian: [+Q +W ÐI] Japanese: [ÐQ ÐW ?I] Move a single wh-word (English, French, …) Czech: [+Q +W +I] wh-movement Move all wh-words (Bulgarian, Polish, …) Commonly though that to write the meaning of a wh-questions, you must have an operator wh-in-situ Move no wh-words (Chinese, Japanese, …) (“For which x”) binding a variable (x). Movement yields an operator-variable structure.

In English, the one wh-word goes into SpecCP. (57) Whati did John buy ti ? (‘For what value of x is it true that John bought x ?’) In Bulgarian, the wh-words all scrunch into a single SpecCP (evidence: nothing can come between wh-words in the front). So, there’s a big question now: If ¥ If the meaning of a question requires movement (55) CP and ¥ Questions mean basically the same thing in all languages qp but • Some languages don’t move some/all of the question words ′ Spec C then ¥ How do those questions mean what a question means? 33 koj kogo CIP rU The answer to this which is commonly adopted is this: Adv IP pruv ... The structure of the grammar:DS  phrase structure rules 1 Another kind of multiple wh-fronting language: Czech, where all wh-words move to the 1 movement rules front, but the first one seems to be separate from the others (things can come between the surface structure (abstract)  SS first and the rest, e.g., parentheticals). 3 more movement rules “phonetic form”  PF LF  “logical form” (meaning) Irish English gives us evidence—in what all questions, all can be “left behind” in any of The underlying structure of a sentence (e.g., John bought what) is DS. the positions occupied by the wh-phrase. The meaning of a sentence is represented as LF. The pronunciation of a sentence is represented as PF. (60) a. What all did he say that he wanted to buy? West Ulster English The rule which takes a wh-word to the front of the sentence is a movement rule. b. What did he say all that he wanted to buy? c. What did he say that he wanted all to buy? In English, one wh-word moves between DS and SS, so we pronounce it first. d. What did he say that he wanted to buy all? and any other wh-words move between SS and LF— after we’ve committed to a pronunciation. It is “hidden” (covert) movement. US English also gives us evidence: If SpecCP is filled with whether, a moving wh-word The idea is that: can’t stop there and the result is bad: ¥ Since the meaning of a question (in any language) requires wh-words to move. ¥ In all languages, all wh-words move—just in some languages you can’t see it. (61) a. Which way did you say (that) Bill went? b. I wonder which way Bill went. So a revised typology is: c. I wonder whether Bill went East. ¥ Parm Q: Question needs a wh-word in SpecCP by SS d. * Which way do you wonder whether Bill went? ¥ Parm W: The point when all wh-words need to be in front: SS or LF. ¥ Parm I: Whether wh-words can attach to IP In Bulgarian, we know that several wh-words can pack into SpecCP at once—and sentences like (61d) are fine (since one wh-word in SpecCP doesn’t make it “full”).

English: [+Q W:LF ÐI] In Czech (like English, only one thing fits in SpecCP), the whether-type effect appears. Bulgarian: [+Q W:SS ÐI] Japanese: [ÐQ W:LF ?I] A language parameter: Czech: [+Q W:SS +I] ±MSC:Whether a language allows multiple elements in SpecCP. (This may not be independent of “Parm I” from before) Some languages appear to fall somewhere in the middle though—[W:– , Q: ±] Bulgarian: +MSC English: ÐMSC Czech: ÐMSC (58) a. Qu’a-t-il donné à qui ? French what has-he given to whom ‘What did he give to whom?’ A universal order of adverbs…

b. Il a donné quoi à qui ? A strict ordering restriction on adverbs—one has to come before another. The same He has given what to whom hierarchy appears in Italian, French, Norwegian, Serbo-Croatian, Hebrew, Chinese, ‘What did he give to whom?’ Albanian, … The English hierarchy is (partially) as given in the box below.

frankly > fortunately > evidently > probably > now > perhaps > wisely > Wh-islands and successive cyclicity usually > already > no longer > always > completely > well Synthesized hierarchy: Wh-words stop in each CP along the way from their underlying position to SpecCP: sincerely/truly/honestly intelligently/wisely almost (un)fortunately/unluckily usually completely (59) Which way did you say [CP — that Bill went — ] evidently often well probably already quickly now no longer again perhaps always necessarily continuously/just Italian:The verb can appear between any two adverbs. This means there needs to be a place for the verb to move to. (64) a. CP C This means we need a functional head for every adverb position. 3 Spec C′ Other languages seem to have overt morphemes that correspond to these positions. 3 CIP (62) ku pwun-i cap-hi-si-ess-ess-keyss-sup-ti-kka? Korean that person-NOM catch-PASS-HON-ANT-PAST-EPISTEM-HON-EVID-Q that # ‘Did you feel that he had been caught?’ Mary book bought

Languages seem to stick to this hierarchy, across languages... b. CP Universal Spec-head-complement order? qp IPi C′ #3 Linear Correspondence Axiom Mary book bought C ti (Essentially): X-bar structures can only be like this: Spec-Head-Complement that

(63) XP Provides an explanation for: SOV languages do not move wh-words to the front. 3 (In “head final” languages, C is final, so IP must have moved into SpecCP, SPEC X′ filling it up, meaning that wh-words can’t move to SpecCP). 3 X COMPLEMENT HEAD

(Denies the existence of the headedness parameter—even SOV languages are underlyingly SVO)

Provides an explanation for: Possible orders within NP of noun, , numeral, . Underlying order (universally): Dem Num Adj N Orders which come from N moving: Dem Num [N]i Adj ti Dem [N]i Num Adj ti [N]i Dem Num Adj ti Other options; moving larger constituents: Dem N Adj Num (comes from second movement: Dem [N Adj]i Num ti) N Adj Num Dem (comes from third movement: [N Adj Num]i Dem ti)

Point: No difference in the possible orders between prepositional and postpositional languages. If there were a headedness parameter, we would predict more orders than we actually see.

SOV languages Japanese John-ga [ Mary-ga hon-o katta to] omotteiru John-NOM Mary-NOM book-ACC bought that thinks ‘John thinks that Mary bought a book’