LING 321 LECTURE #10 LEVELS OF REPRESENTATION Two-level model: UNDERLYING REPRESENTATION (UR) SURFACE REPRESENTATION (PR) UR -- too abstract? PR -- too detailed? Lexical Phonology: there is an intermediate level lexical representation Structuralists: BIUNIQUENESS PRINCIPLE Any sound in a given environment must be an allophone of one and only one phoneme. The biuniqueness requirement (= once a phoneme always a phoneme, see p. 118) is designed to prevent phonetic/phonological ambiguity. e.g. [p h ] is an allophone of /p/ However, consider the following problem: Many varieties in English have an optional rule: 1
Morpheme Form 1 Form 2 bat bæt bæ cap k æp k æ back bæk bæ butter b t b /p/ /t/ /k/ are optionally realized as [ ] word-finally, and /t/ as [ ] intervocalically. If [ ] is an allophone of /t/ (wider distribution!), then it cannot also be an allophone of /p/ and /k/: this would mean that [bæ ] is ambiguous between /bæt/ and /bæk/. Suggested solution by classical phonemics: Morpheme: cat /kæt/ optionally [k h æ ] back /bæk/ /bæt/ [bæ ] cap /k æp/ /k æt/ [k æ ] Thus the morpheme back has two allomorphs, one being homophonous to the morpheme bat; cap has two allomorphs, one being homophonous to the morpheme cat; 2
Further, If the /t/ allomorph is chosen, the /t/ [ ] is obligatory : cap never appears as [k æt] But the rule is optional if the /-t/ is the only allomorph: /kæt/ [k æt] or [k æ ] Conclusion: The biunique solution is counter-intuitive; the obvious solution is to allow OVERLAPPING: [p] [t] [k] /p/ /t/ /k/ \ / [ ] This means that phonetic strings can be ambiguous! Study the Dutch example on pp. 117-118. Regressive voicing: both neutralization and allophonic rule! Explain. English (pp.116-117): dictionary representations: phonemic transcriptions, e.g. pin [pˆn] not [p ˆn] but: looked [l kt] not l kd] -- the form incorporates the obstruent devoicing rule! It thus presents an intermediate level between the UR and the PR. 3
Structuralists: Phonology is independent of morphology and/or syntax; there is no information available for phonological analysis of an utterance except the phonetic segments in it. English plural allomorphs: [-z] [-s] [- z] morphophonemic: //-z// phonemic: /-z/ /-s/ /- z/ phonetic: [-z] [-s] [- z] Bloomfield (1933): morphophonemic representations were set up simply for convenience and, unlike phonemic representations, have a purely conventional status. as for the order of rules: The term before, after, first, then etc. in such statements tell the descriptive order. The actual sequence of constituents... is part of the language but the descriptive order of grammatical feature is a fiction and results simply from our methods of describing the forms. Generative phonology: UR: /-z/ (widest distribution!) 1. Schwa Epenthesis Rule 2. Devoicing Rule 4
Premises: the phoneme represent surface contrasts (minimal pairs!) a segment can only belong to one phoneme ( once a phoneme, always a phoneme ) Conforming to these premises leads to problems that obstruct generalization (consider the Dutch example). The theory of Lexical Phonology has provided an answer to the intermediate representation issue: Some of the phonological generalizations are stated in the lexicon -- phonological, morphological and semantic information of morphemes Other phonological generalizations are stated outside the lexicon There is a clear distinction between lexical and postlexical phonological rules Lexical rules/postlexical rules: Study (4) on p. 119. Reference to morphological labels (5) (6) on p. 120. Lexical rules have access to category labels (e.g., Noun Verb etc.) Exceptions A rule that has exceptions cannot be a postlexical rule. Example: English trisyllabic laxing (8) Lexical rules may have exceptions; they may also be exceptionless (9) 5
Structure preservation Lexical inventory: smaller than the inventory of the surface representation. their output contains segment that are present in their UR e.g.: English aspirated vs. unaspirated stop do not need to be contrasted (allophonic postlexical rule!); the new segments (aspirated stops) are not included in the lexicon. Structure preservation may have exceptions: (9), (10) Native-speaker intuitions Judgements by native speakers is based on the lexical representation regardless of surface differences. Example: flapping rule Application across word boundaries Only post-lexical rules apply across word boundaries. Lexical rules apply before postlexical rules First lexical rules apply; a rule that applies after a postlexical rule must be also postlexical. Phonological information in the lexicon may refer to units larger than segment, such as syllables or feet. Examples: English comparative and superlative suffixes (11) 6
German devoicing rule (12) Study the model of phonological grammar in (13) and (14)! Phonetic implementations: are they universal? the same phonological syllable is pronounced differently by speakers of different languages speakers of the same languages may pronounce the same sequences of segmenst differently fine tunings of articulation Models of implementation (optional) pp. 129-131. --------------------------------------------------------------------------------------- Lexical Phonology offers a solution to the problem of defining the level of representation between UR and PR. Instead of establishing a division between neutralization and allophonic rules, it proposes that PHONOLOGY EXISTS IN TWO SEPARATE COMPONENTS IN THE GRAMMAR: (i) lexicon morphological information + lexical phonological segments, (ii) phonological level, where no morphological information is available. Postlexical rules are different from lexical rules phonetic implementation rules Phonological rules work with features, moras etc. Phonetic implementation rules work with translating the UR into articulatons. 7
8