Ordered Rules

24.961 Ordered rules [1] Basque: Biscayan dialect PGG p. 22 [2] Kenstowicz, Michael. Phonology in Generative Grammar. Blackwell Publishing, 1994. © Blackwell Publishing. All rights reserved. ate ate bat* atie ‘door’ asto asto bat* astue ‘donkey’ [3] Baztan dialect A. B. noun definite gison gisona ‘man’ egun egune ‘day’ mendi mendie ‘mountain’ buru burue ‘head’ etʃe etʃia ‘house’ aʃto aʃtua ‘donkey’ noun definite alaba alaba ‘daughter’ neska neska ‘girl’ muge muge ‘limit’ fabrike fabrike ‘factory’ Basque: five vowels Biscayan i,u,e,o,a,u cf. bat ‘one’ 1 rules: umlaut: [+syll,+low] -> [-low,-back] / [+syll, +high] Co ____ glide epenthesis: 0 -> [+syll, +high, @back] / [+syll, +high, @back] ____ [+syll] raising: [+syll] -> [+high] / ___ [+syll] ordering raising precedes umlaut (feeding) /asto-a/ astu-a raising astu-e umlaut glide epenthesis precedes raising (counter-feeding) /buru-a/ /asto-a/ buruw-a -------- GE -------- astu-a Raising buruw-e astue umlaut [4] Baztan dialect • has umlaut rule • has raising rule • no glide epenthesis • umlaut precedes raising (counterfeeding) /buru-a/ /asto-a/ buru-e ---------- ---------• astu-a umlaut raising a-final noun; we expect two a’s in output but just one occurs: degemination [+syll, +low] -> 0 / ____ [syll, +low] • muge vs. ece: a-deletion precedes raising (bleeding) /muga/ /muga-a/ /ece/ /ece-a/ --------- mug-a ------- --------- a-deletion muge mug-e ------- -------- umlaut ------ --------- --------- eci-a raising 2 • dialects may differ by having the same rules and same underlying forms but different order of their rules (Halle 1962) • a new type of grammar/language change [5] Feeding and Bleeding (Kiparsky 1968, 1971) If Rule A creates potential inputs to rule B and rule B applies, we say A feeds B and this can be described by requiring A to precede B. If Rule A creates potential inputs to rule B and rule B does not apply, we say A counterfeeds B and this can be described by requiring B to precede A. If Rule A removes potential inputs to rule B and B does not apply we say A bleeds B and this can be described by requiring A to precede B. If rule A removes potential inputs to rule B and B does apply we say A counterbleeds B and this can be described by requiring B to precede A. e.g. flapping and shortening+raising before a voiceless consonant in English Canadian raising. Kiparsky (1968) proposed that rules may diachronically change their order towards feeding and bleeding relations. These are situations in which each rule is true of the surface form. Counterfeeding and counterbleeding create situations in which the earlier rule is not true of the surface form and hence “opaque”. One must undo the effects of the later rule to see the full effects of the earlier rule. It was suggested by Kiparsky (1971) that opacity was more difficult to learn. It would be interesting to revisit this question today with an artificial language learning experiment. [6] Summary The SPE model with ordered rewrite rules defined over sounds represented as distinctive feature matrixes showed that considerable analytic insight into the structure of a language could be obtained. Its concern with formal statements and explicit representations created a generative grammar: an input-output mechanism whose scope went well beyond mere summaries of the data in a corpus. Thus a scientific research program was created in which many new questions arise in extending the ordered rule format to more data both language-internally as well as crosslinguistically. We sample here a few of the questions that arose. [7] Multiple rule application 3 Vowel harmony is a challenge to the rewrite rule mechanism since the harmony may extend over an entire word, which can be very long in agglutinative languages such as Turkish ılımlı-laş-tır-dık-lar-ımız-dan # mı-sın ‘are you the ones who we made calm?’ sinirli-leş-tir-dik-ler-imiz-den # mi-sin? ‘are you the ones who we made angry?’ root-DER.Verb.-CAUS-NOM-PLU-1PLPOSS-ABL # Q-2SG A. Turkish vowel harmony front back high i y ɯ u mid/low ɛ œ a ɔ noun pl. his N. dal dal-lar dal-ɯ 'branch' kɔl kɔl-lar kɔl-u 'arm' kɯz kɯz-lar kɯz-ɯ 'daughter' kul kul-lar kul-u 'slave' yɛl yɛl-lɛr yɛl-i 'wind' gœl gœl-lɛr gœl-y 'sea' di∫ di∫-lɛr di∫-i 'tooth' gyl gyl-lɛr gyl-y 'rose' • roots contrast for eight possible vowels • most suffixes contrast for just [±high]; values for [back] and [round] determined by harmony [– cons] -> [α back] / [α back] Co ___ (palatal harmony) [– cons, +high -> [α round] / [α round] Co ___ (labial harmony) • what does grammar predict for ‘his slaves’? it could be [kul-lar-ɯ] or [kul-lar-u] depending on whether the [round] value of the possessive suffix is determined by the vowel of the preceding syllable or the first vowel of the root. • In fact it is [kul-lar-ɯ], suggesting that the harmony arises by successive applications of the rule over adjacent syllables, with one application creating the input to the next. [8] A couple of examples of problems with this view A. Istanbul Turkish (Kumbaraci 1966) [warning: I have not been able to confirm this data] 4 • raising and unrounding before palatals y, ʃ, dʒ infin. Imper. Standard written form ye-mek yi-yin ‘eat’ yiyin üʃü-mek üʃi-yin ‘be cold’ üs.üyün oku-mak okɯ-yɯn ‘read’ okuyun sakla-mak saklɯ-yɯn ‘hide’ saklayɪn [+syll] -> [+high, -round] / _____ [-syll, +high, -back, -round] • problems in obtaining correct output (I denotes the “archiphoneme” [+high, 0back, 0round] /okI-yIn/ oku-yun labial (and back) harmony okɯ-yun derounding /okI-yIn/ okɯ-yIn oku-yun • derounding labial (and back) harmony cycle /okI-yIn/ okI stem cycle oku labial (and back) harmony ------ derounding oku-yIn word cycle oku-yun labial (and back) harmony okɯ-yun derounding /okI-yIn/ okI stem cycle ------------ derounding oku labial (and back) harmony oku-yIn word cycle okɯ-yIn derounding oku-yun labial (and back) harmony 5 • the local unrounding change before the palatal must be insulated from labial harmony imposed by the preceding vowel but the output of derounding triggers harmony on what follows • how can we intercalate one rule inside another? • one approach (inspired by Chomsky’s (1979) strict cycle (cf. phase)) distinguished application within a stem and application across a boundary; once the stem application occurred, the rule could not return on a later cycle to undo the effects in prior cycle • stem application would normally yield a constant shape for the stem in a paradigm while suffixes could alternate; due to this side effect, it was sometimes called the Alternation Condition (Kiparsky 1973) /okI-yIn/ okI stem cycle ------------ derounding oku labial (and back) harmony oku-yIn word cycle okɯ-yIn derounding okɯ-yɯn labial (and back) harmony across boundary oku-yun blocked by strict cyclicity okɯ-yun blocked by locality another possible solution (based on Nevins 2010 Locality in Vowel Harmony) • vowel harmony is not the change of following vowels based on the triggering vowel but unspecified (noncontrastive) features becoming specified through local agreement • assimilation to consonant takes precedence since it is more local (adjacent) /okI-yIn/ okɯ-yIn derounding okɯ-yɯn labial (and back) harmony /okI-yIn/ /okɯ-yIn/ okɯ-yɯn high − + + --deround à −+ + --harmony à −++ back +0 0 +0 0 +++ round + 00 +− 0 + − − B. Chumash sibilant harmony (Poser 1982, 1993) [1] ha-s-xintila his gentile ha-ʃ-xintila-waʃ his former gentile 6 p-iʃ-al-nanʔ don't you two go s-is-sili-uluaqpey-us k-ʃapatu-tʃ I wear shoes cf. Spanish zapato ʃ-kamiʃa-tʃ he wears a shirt cf. kamisa they two want to follow it [2] anterior dissimilation cf. German1 /s-nanʔ/ > ʃnanʔ he goes /s-tepuʔ/ ʃtepuʔ he gambles [3] output of dissimilation does not undergo sibilant harmony /s-ti-yep-us/ > ʃtiyepus he tells him /s-ti-yep-us/ stiyepus sibilant harmony ʃtiyepus dissimilation [4] But the output of anterior dissimilation does trigger harmony to preceding sibilants /s-is-tiʔ/ > ʃiʃtiʔ he finds it [5] ordering paradox /s-is-tiʔ/ siʃtiʔ dissimilation ʃiʃtiʔ sibilant harmony /s-ti-yep-us/ stiyepus sibilant harmony ʃtiyepus dissimilation • unlike in Turkish, [±anterior] is contrastive in Chumash affixes • the harmony process might mask two changes: first neutralization of the [±anterior] contrast when followed by another sibilant in effect changing s and ʃ to S ([±anterior] > [0anterior]) followed by the anterior dissimilation and then the valuation of the 0’s by harmony 1 The underlying [+anterior] for the 3 sg. subject/possessive is based on Beeler (1970: 16) 7 C. Kikerewe (Odden 2000, Bantu, Tanzania) If rule iterates across a string, left-to-right vs. right-to-left application can maximize or minimize application of the rule (Kenstowicz & Kisseberth 1973, Howard 1973) ku-bal-a ‘to count’ ku-bóh-a ‘to tie’ ku-bóh-á Bulemo ku-bal-an-a ‘to count each other’ ku-bóh-án-a ‘to tie each other’ ku-bal-il-a ‘to count for’ ku-bóh-él-a ‘to tie for’ ku-bóh-él-an-a ‘to tie for each other’ ku-bóh-á Bulemo ‘to tie Bulemo’ a -> á / á Co _____ á -> a / ____ pause a = any vowel ku-twa:ng-il-a ‘to pound for’ ku-té:k-él-a ‘to cook for’ ku-tú-twá:ng-il-a ‘to pound for us’ ku-tú-té:k-el-a ‘to cook for us’ á -> a / á Co _____ Meeussen’s Rule /ku-tú-té:k-el-a/ ku-tú-te:k-el-a M’s Rule ku-tú-té:k-el-a H-doubling ku-yílúch-a ‘to chase’ ku-bá-yíluch-a ‘to chase them’ ku-bá-tú-yilukizya ‘to chase them for us’ /ku-bá-tú-yílukiza/ Right-to-left iteration leading to maximal application of rule ku-bá-tú-yílukiza ku-bá-tú-yílukiza ku-bá-tú-yílukiza ku-bá-tú-yílukiza ku-bá-tú-yilukiza M’s Rule applies ku-bá-tu-yilukiza M’s Rule applies ku-bá-tu-yilukiza • • Left-to-Right application would incorrectly give minimal application: *ku-bá-tu-yílukiza High tone doubling applies minimally: right-to-left 8 References Howard, Irwin. 1972. A Directional Theory of Rule Application in Phonology. MIT Ph.D. dissertation. Kenstowicz, Michael and Charles Kisseberth. 1973. The multiple application problem in phonology. C. Kisseberth, ed. Studies in Generative Phonology. Edmonton: Linguistic Research, Inc. 13-41. Kiparsky, Paul. 1968. Linguistic universals and language change. In Emmon Bach and Robert Harms, eds. Universals in Linguistic Theory. New York: Holt. Kiparsky, Paul. 1971. Historical linguistics. William Dingwall, ed. A Survey of Linguistic Science. College Part: University of Maryland. Kiparsky, Paul. 1973. Abstractness, opacity, and global rules. In Osamu Fujimura, ed. Three Dimensions of Linguistic Theory. Tokyo: TEC. Kumbaraci, Turkan. 1966. Consonantally conditioned alternation of vocalic morphophonemes in Turkish. Anthropological Linguistics 8, 11-24. Nevins, Andrew. 2010. Locality in Vowel Harmony. Linguistic Inquiry Monograph 55. Cambridge, MA. MIT Press. Odden, David. 2000. Opacity and ordering: H-deletion in Kikerewe. The Linguistic Review 17, 323-335. Poser, William. 1993. Are Strict Cycle effects derivable? Sharon Hargus and Ellen Kaisse, eds. Studies in Lexical Phonology. New York: Academic Press. Pp. 315-21. 9