The behavior of phonological rules at "strata" in lexical phonology

(1)

The behavior of phonological rules at "strata"

in lexical phonology

著者（英） Akihito Kawasaki

journal or

publication title

Core

number 21

page range 65‑86

year 1992‑03‑15

URL http://doi.org/10.14988/pa.2017.0000014889

(2)

The Behavior of Phonological Rules品tStrata" 65

The Behavior o f Phonological Rules a t S t r a t a "

i n Lexical Phonology A k i h i t o K a w a s a k i

O . I n t r o d u c t i o n

Lexical Phonology，ぉ itsnam巳 shows，consists of two s巳parate but interrelated fields: phonology and

、

morphology. This theory deals in a different w呂ywith the lexicon which has traditionally be巴nmade light of; it has come to be considered that the lexicon tre品tsmorphological and phonological rules as well as individual words日 dmorph印 les.Accord‑

ingly，it is assumed that word‑formation rules of the morphology ar巴

directly paired with phonological rules grouped together at various levels."l In short， phonological and morphological rules are inputs to each other. In Lexical Phonology， the lexicon functions as something more than呂list;it has a hierarchical structure. This theory is bas巳don the idea that the lexicon has some ordered strata" (or levels)，昌ndso there has be巴nmuch discussion about how many strata constitute a language. In this paper 1 shall analyze the entity of strata"昌ndverify its applicability through the example of English.

(3)

66 The Behavior of Phonological Rules at Strata"

1 .

Lexical and pos

t 1

exical rule applications

Lexical phonology is well represent巳dby the idea that som巴phonolog‑ ical rules are applied along with th巴morphologicaloperations in the lex‑ lcon，品ndothers are applied postlexically. In reg丘rdto the distinction be‑ tween lexical and postlexical rule applications， therεhave been several vIews. Pesetsky (1979) and Kiparsky (1982a)2 explain this distinction in terms of cyclicitl: rules which apply cyclically are lexical; those which do not are postlexical. According to Mohanan (1982)， th巳determinantis the sensitivity to morphological inform旦tion. On the basis of the latter theory， the following rule can be made:⁴

(1) A rule application requiring morphological information must take place in the lexicon

Furthermore， Pullyblank (1986) presents the following summary of the contrast between lexical and postl巳xicalproperties・5

(2) LEXICAL POST ‑LEXICAL a. may refer to word‑internal a. cannot refer to word‑internal

structure structure

QU

‑A U

γムO W

QU

QM

o ri c a

y c

pドu

p c

a V J

gふC O ρ

巴し

n'b

V d V

ザ在a

m m

句h U

戸U

. may apply across words c. cannot be cyclic

d. if cyclicぅthensubject to strict d. non‑cyclic， hence across‑the‑ cycle board

e. structur巴‑preservmg e. need not be structure‑

f .

may have lexical exceptions

preservmg f

.

c丘nnothav巳lexicalexceptions

(4)

The Behavior of Phonological Rules at Strata" 67 g. must precede a11 post‑lexical g. must fo11ow a11 lexical rule ap‑

rule applications plications

2 . Stratum

According to Mohanan (1982)，the notion of the st日tumwぉ firstsug‑ gested by Siegel (1974)， and developed further by A11en (1978).，6 ， Strata are， so to sp回 k.ar巴aswhをreaffixation processes and phonologic丘1rules are concatenated. Goldsmith (1990) points out th巴昌dvantagぞ ofthe con‑ cept of str且taぉ fo11ows:

Through the formal use of strata， lexical phonology makes claim about (i) the r巳analysisof wh品tonce w巴rec品lledphonological boundaries; (ii) th巴 ordering of morphological elements， i.e

品ffixes;and (iii) th巳cycl巴inphonology?

Moreover， Mohanan (1982)， mentioning the similarity betwe叩 themon the one hand， distinguishes the notion of the stratum from that of the com‑

ponent in linguistic theory; there can be seen a unique丑ssociationwith a set of rules and r巳presentationsin the component whereas there cannot necessarily be a one‑to^司onecorrespondence between th巴stratumand rules and repr巳sentatlOns

The number of strat乱cannotbe det白mlll巳duniformly at present. Th巴re‑ fore different languages recognize differ叩 tnumbers of strata， but it c四 be said that every languag巴hasat le品sttwo strat且:the lexical stratum and the postlexical stratum. The domain in which phonological rules ar日昌pplied is determinεdぉ fo11ows

(5)

(3) Principles of Domain Assignment⁸

a. In the absence of counterevidence. assign the smallest number of strata as the domain of a rule

b. In the absence of counterevidence， assign the highest possible stratum as the domain of a rule (where lowest"

=

stratum 1) These assumptions imply that all unmark巳dphonological rules are appli‑ cabl巳solelyat the postlexical stratum. As for English， Halle and Moha‑

nan insist that this 1在nguagehas five strata: four lexical str呂t品 andon巴

postlexical stratum

The linear ordering is also a significant character of the concept of stratum. What composes the core of the word is an underived lexical item which contains a single morphem色 Suchwords c呂n日xistwithout the application of word‑formation rules; they appear in th巳lexiconwith the phonological， grammatical and semantic prop巴rtieswith which th巴ysur‑ face."g Stratum 1 consists of their counterparts‑bound morph巴meswhich cannot appear without品ttachingto some other forms. Most of the affixes which appear in stratum 1 corr巴spondto those divided by the

十

boundary in the SPE framework. In this regard， Halle and Mohanan describe all class 1 affixes (like ‑ic， ‑ion， ‑ity，丘ndin‑， which are丘ssociatedwith th巳十

boundary10 in SPE) attach at stratum 1^{， 1}^，^，1 and moreover Kiparsky says: This level (Stratum 1) includes derivational suffixes such且s‑al^ち

‑ous， ‑tか且ndinflectional suffixes such as those in kept^タmet，hid‑ den， children， a必初da，indices， foci，・.• and other stem‑changing morphology as in teethe， bleed， bathe， teeth， lice^町12

As r巳gardsstratum 2， Halle and Mohanan's巳xplanationis that class II affixes (like引ess，abjectival ‑ed， ‑hood， and un‑， which are associated with

(6)

The Behavior of Phonological Rules at Strata" 69 the :1* boundary¹³in SPE) att丘ch呂tstratum 2.，1，4 Kiparsky points out that

:1* boundary derivation and compounding take place at stratum 2

In addition to th巳abovetwo strata， English has thr凹 morestrata. Stra tum 3 is the pl品C巳compoundsare formed，在ndstratum 4 is the environ‑ ment where regular inflections occur. The fifth stratum is the postlexical stratum for words to constitute phrases and larger syntactic entities

3 .

Correlation between strata and phonological rules Some phonological rules are applic旦bleat mo陀 th且none stratum. Spe‑ cifically such rul巳S剖 thevowel shortening and the vowel lengthening have stratum 1品sth巳irenvironmen .tThose words which include class 1 affixes are subject to the Trisyllabic Short巴ningrules¹⁵which apply at stratum 1， although those containing class II affixes are not. Contrary to them， thεrule of Flapping¹⁶which app伺 rsat the postlexical stratum is applicabl巴rεgardlessof the position in the phonological phrase In other words， it applies not only inside words but also beyond word bound且n巴S

3.

1 .

Stem‑final Tensing

First of all， the巴xampleof Stem‑final Tensing which is comparativ巴ly prevalent in English dialects shall be taken. Halle and Mohanan describe that Stem‑final Tensing (SPE， 74) tenses nonlow vowels without simul‑ taneously diphthongizing and lengthening them.，1，7 This rule conspicuous‑ ly shows the influ印 cewhich the morphological structure exerts on phono‑

logical rul巴s;this rule tenses th巴st日m‑finalhigh vowel of a word. SPE r巳presentsthis rule as follows:¹⁸

(4) V →[+tense] / ̲

#

トlow]

(7)

70 The Beh昌viorof Phonological Rules at Strata"

If the t巴nsedstem‑fina1 vowe1 is a back vowe1， it is a1so subject to 1ength^司 ening. In the case where the stem‑fina1 vowe1 is a front vowe ，lon the con‑ trary， it is not 1engthened in most di旦1ectsof English. What matters here is that certain morpho1ogica1 e1ements cause the tensing of the stem‑fina1 vowel. Some examp1es are shown below:19

(5) a. city [S1ti]， happy [hεpi] ，vary [ve a ri]， remedy [remrdi] b. cities [S1tiz]ぅmarried[marid]

c. city hal1 [Srtih8 1]

d. happiness [hapina s]， 1adylike [leydilaik]， 1丘dyship [1巴ydis1P] e. various [ve a r1 a sD， radiance [r巴ydra ns]， remedia1 [rrmi:d1 a 1] These examp1巴sshow that 1 followed by class II affixes， inflectiona1 affixes， the word boundary or anoth巳rstem III品compoundis replaced by i， but the one before class 1 affixes is not. Therefore L巳xicalPhono1ogy r巴p resents Stem‑fina1 T巴nsingas follows:20

(6) Stem‑fina1 T印 sing(domain: stratum 2) V

→

[+tense] / ]

[‑low]

As th巳aboveillustration shows， the boundary marker of Lexica1 Phono1o‑

gy is signified by doub1e square brackets. This means that boundary mark‑

ers in Lexic品1Phono1ogy work as the mere indic旦tionof the initia1 and fina1 positions of constituents， wh巴reasthose of SPE且resequentia1 en‑ trtles

Here the word haPl^うinessshall be tr巴atedas an巴xamp1e.As the suffix ‑

ηess is added at stratum 2， two constituents of this word are represent巳dぉ

two separate units at stratum 1^噂 Interms of this， each case can be shown as follows:21

(8)

The Behavior of Phonological Rules at Strata" 71 (7) Stratum 1: [hap r] [n e s]22

Stratum 2: [[hap r] [n e s]'，]

Furthermore the brackets separating two morph巴mesof the noun are de‑ 1eted at th日endof stratum 2 wher巴twocons

t J

tu巳ntsare concatenated into

23

the noun. In consequence it gets the following representation: (8) [hap r nεs]

This a1ternation is due to the Bracket Erasure Convention stated by Mohanan:

(9) Bracket Er品sureConvention (BEC)24

After th色乱pplicationof all rul巴S品t丘stratum，the brackets betwe巳n th巴morphemes ar巴 de1et己d，so that reference to the constituent morphemes becomes impossible at subsequent strata

3 . 2 . N

asal Assirnilation and g‑Deletion

The alternation from [ 'J ] to ['J g] which can be seen in long [b 'J] and longest [1:) 'J g 8 st] also indicate that morphology and phonology are cor‑ relative in Lexical Phonology. The SPE appro呂ch，which regards [ 'J ] as the derived form of the under1ying nasa1s Iml and In/， shows the follow‑ ing altern呂tlOn:

(10) N asal Assimilation (domain: stratum 2)25

n→ っ /

(11) g‑Deletion (domain: st日tum216 g→のI[+nasa1) ]

(9)

7z The Beh呂viorof Phonological Rules at Str品句"

General1y there cannot appear inflectional suffixes within compounds， but they can follow compounds. According to this view， the distinction be‑ tween w巴l1‑formednessand ill‑formedness is shown as follows:²⁷

(12) W ell‑formed Il1‑formed railroad stations 率railsroadstation house‑hunted *railroads station house‑hunting 率houses‑hunt

This fact is well explained by assuming that inflections are suffixed at stratum 4 while compounding occurs at stratum 3. From this assumption two different ‑ing suffixes can be inferred: the inflectional， p且rticipal‑ing and th巳derivational，nominalizing ‑ing; the form巳roccurs且tstratum 4旦nd the latter at stratum 2. This can also be true of the comparative and super‑ lative suffixes ‑er and ‑est. Therefore， before stratum 4， long， longing (part.) and longest have the following represent品tions:²⁸

(13) [long] [long] [ing] [long] [est]

The extra spaces betwe巳n the morph巴mes indicate th且tthere has not occurred su日ixationyet at this stage. At stratum 2， th巴seforms are subject to N asal Assimilation and g‑Deletion， and they are represented as fol‑ loowwss:: ²⁹

(14) [10')] [109] [i9]

[10']] [ εst]

What matters here is that there is no rule that reinserts [g] after nasals in English. As for the form longest [l :) ' ]g a st ，]accordingly， the suffix ‑est must be added at stratum 1， because otherwise [g] is deletεd呂tstratum 2

(10)

The Behavior of Phonological Rules昌tStrata" 73 For the s丘mere且son，the past tense suffix Itl of such irregular verbs as kept， left and meant must appear at stratum 1. Eventually， at least， in irregular且djectiv巴sand verbs， the suffix is .added at stratum 1 whereas the regul且ron巴isadded at stratum 4

3 . 3 .

Syllabic Consonants

The phonological contrast seen betw田 n the words twinkling (N) and tωinkling (V) is evidenc巳ofthe necessity of stratal distinctions; the fol‑ lowing SPE rule cannot account for this contrast:³⁰

(15) sonorants become syll且bicI C :1:1:

This rule can tr巴at such c旦ses as burglel burglar， prismlp rismatic and cylinderl cylindrica .lIn the cas巳 of twinkling (N)呂nd twinkling (V)

howev巴r，the SPE theory is obliged to alter slightly as follows: the struc‑ ture of twinkling (N) is twinkle

十

ing(N); that of twinkling (V) is twinkle :l:l:iηg (V). This altεrnation contr旦dictsth巳SPEprinciple that ‑ing 'which is not subject to class 1 affixation must be added only :1:1:. Here it should be noticed that in all dialects of English， a syllabic consonant becomes non‑syllabic when followed by a vowel‑initial derivational suffix， whether it is class 1 or class 2.³¹Mohanan cites the following examples by Kenyon and Knott (1944) in which syllabic consonants are endowed with a verti‑ cal line below:³²

(16) a. simple (simpl]

b. couple (kA pl]

simplify (simplifai] simplicity [simplisiti] simply (simpli] coupler [k A pl d r] couplet [kA plit] coupling (N) [k A pli!J ]

(11)

74 The Beh且viorof Phonological Rules品tStrata"

c. double [d A bl]

d. burgle [b a rgl]

e. twinkle [twi 9 kl]

f .

wrinkle [ri 9 kl]

doubly [d A bli] doublet [dA blit] burglar [b a rgl a r]

twinkling (N) [twi 9 kli 9 ]

wrinkl y [ri 9 kli]

Thes巴examples are wel1 explained， if it is assumed that syl1able form品ー

tion takes place in forms derived呂tstr乱ta1 and 2 as w巳11as in forms listed as morphemes. Now two main principles to treat this m丘ttershall be presented: one is the Principles of H日adshipin English， and the other is the Principle of Maximal Syl1abification. Th巴se principl巴S 且re shown below:³³

(17) Principles of Headship in English

As巴gmentis syllabic if the following conditions are met: a. it is [+ sonorant]

b. it is stem final if it is [十cons]

c. it constitutes a SONORITY PEAK within its syllable In all other cases， it is nonsyllabic.

(18) Principle of Maximal Syllabification

Given alt巴rnatesyllabifications for且string，choose the one that minimises the total number of syllables

According to the Principles of H回 dshipin English， 1 can b日considered either syllabic or nonsyll丘bic;simplify， for instance， has eith巴rthr凹 or four syllables. Principle of Maximal Syllabificationラonthe other hand， re‑ quires for l to be nonsyllabic. In the case of simρle， however， 1 is necessari

(12)

The Behavior of Phonological Rules at Strata" 75 ly sy11abic by virtue of Principles of Headship in English. Aftεr all， 1 fol‑ lowed by vowel‑initial derivational suffixes is syllabic， whereas the one which is word‑final or precedes consonant‑initial derivational suffixes is nonsyllabic. In the light of the abovεanalysis， th巳applic品tionof the sy 11丘‑

ble formation rule at stratum 1 and the cy.clical旦pplicationof phonolog‑ ical rules at stratum 1 can be inferred. This inference is based on the op eration of stress rules on syllable tiers.

Such reapplication of Syllable F ormation to derived forms at stratum 1 can be found not only in sonorants but also in nonsonorants. Among the latter examples are logicllogi [kh]alityラ dialeqf']ologi⁴;their stem final / k/ and /t/ are not subject to aspir品tionin spite of the stressing of the sub‑ sequent vowe .lThese facts show the transition of /k/^昌nd/t/ from th巴

syllable‑final position to the syllable‑initial position， which means that English voiceless stops are subject to丘spirationonly when they are at the initial position of a stressed syllable

Furthermore this tendency can be seen in compounding and inflec‑

35 tlOns:

(19) a. doublεedged [d A bled 3 d]

b. bugle office [byuugl J fis]

c̲ bubble eye [b A blai]

doubling (V) [dA bli') ] or[dA bli')]

bugling (V) [byuugli ')

1

or [byuugli ' )

1

bubbling (V) [b

^

bli 'J

1

or [bA bli')

1

This distinction between the usu呂11 and that with a vertical lin巳atth巴

(13)

bottom justifies the stratal distinction between compounding and inflec‑ tion and class 2 derivation. The syllabicity retention of stem‑final con‑ sonants in compounds presuppos巴s the applicability of Principles of Headship in English at stratum 2 and its inapplicability at stratum 3. This is well reflected by the following pairs: logical的

J /

dialecωlogy and pick‑ axe/ hot oif6; the former aspirates and resy llabifies the voiceless stop， but the latter does not. The above Principles apply n巴itherat stratum 4 nor at the postlexical stratum. Eventually the optional syllabification of 1 in the previous examples of inflections can be explained by the following rule: (20) Sono日ntResyllabification (domain: stratum 4)37

V

→

C / ]

v

⁽ô^p^tⁱôⁿâ^l⁾

[ +cons]

This rule is inapplicable to compounds; its domain is not stratum 3 where compounding takes place， and] before V does not exist at stratum 4 by the application of Bracket Erasure Convention at stratum 3. Th巴reforethis rule can be one justification for a stratal distinction between compounding and inflectional affix抗ion. Similarly， Syllable Formation， which "re‑ applies obligatorily across derivational suffixes， but does not apply across the stems of且compound，"38reflects the distinction betw巴enstrata 2丘nd3.

3

. 4 .

Linking [r] in N onrhotic Accents

Syllable Formation affects a rule of r Deletion in in nonrhotic accents of English; the rule of r Deletion affects a rule of r Resyllabification. Those dialects in which the linking r" appears alternate as follows:

，，

39

(21) a. soar [so] vs. soaring [sori'J ]

b. bear [bea ] ，bearness [be a n a s] vs. bearest [be a r a st]

(14)

The Behavior of Phonological Rules at Strata" 77 c. star [sta]， st且rless[sta18 s] vs. starry [stari]

The star [star] is dirty. d. saw [so ，]sawing [SOi9 ]

e. idea [aidie ，]ideology [aidi:> 1 a d3 i]

(Cf. nonsense form bearology [be e r:> 1 a d 3 i] f

.

spa [spa]; The spa [spa] is dirty

In the c旦ses_{a~c ，}th巴品lternationcan b巳explainedby th巴followingrule: (22) r Deletion (domain: postlexical)4o

R

X r→ の / I

This rule shows that r品ppearssolely in the sy llabl巳‑initialposition. Addi‑ tion品lly，no segment in th巴onsetcan follow r by an indep巳ndentsyllable structur巳constr且int.The above rule and constraint imply that [r] app巴ars only before a syllabic segment in nonrhotic丘ccents Toapply r Deletion correctly， the distinction betw巳白1the left side of VS. and th巳rightside of it in the above examples of alternation is necessary. For example， [r] in star and sωrless 1巳mainsin the rim巴，whereas that in starηis resy llabified in the syllable‑initial position. The following rul巳 mustprecede r Dele‑ tlOn:

(23) r Resyllabification (domain: postlexical)41 R R R 0 R

X → X

r

(15)

The alternations between RP (Britain) and the NE dialect (U.S.A.) which are shown below arεexplicable by Syllable F ormation， r Deletion and r Resy llabification: 42

(24) a. th巳atre[8 i a ta ，]metre [miyta

1

b. theatric [8 iatrikJ， metric [metrikJ

c. met巴ring[mi yt a ri'J]， theatre expert [θi a t a rekspa t ，]th巳

meter 1S . [d a mi yt a riz

1

d. Homer [howm a ，]tartar [tat a

1

e. Homeric [howmerik]， tartaric [tatarik]

As shown above， the forms丘一ccontain an underlying structure /. . . Cr . . / wher回 sthe underlying structure of d and e is /. . . CVr. • ./. The forms d and巴 havea vowel which品ppearsb巴forer when it is stressed; a ‑c have no vowel but schwa carrying no stress before れ Th巳differencebe‑ tw巳ena‑c呂ndd and e is that schw品precedesr followed by an inflection‑ al suffix， the second word of a compound and other words， while schwa is not put before r which pr巴cedesa deriv品tionalsuffix. ThIs difference is due to the domain difference betw巴enSyllable Formation applicable at strata 1 and 2昌ndr R巴syllabificationappli巴dat the postlexical stratum. The former erases the nucleus; the latt巳rretains i .tAccordingly， Syllable F ormation changes the V slot preceding a derivational suffix to a C slot， while r R白 yllabificationseparates r with no deriv且tionalform b巴hindit from the V slot. Moreover the empty V slot is subject to [a ] Ins巳rtlOn

(16)

The Behavior of Phonological Rules且tStrat品 79

4. Some other remarks

4 . 1 . The l o o p

As is often pointεd out， stratum 2 affixation and compounding ar巴mter‑ depend巴nt;they work as inputs to each other. In reg呂rdto this matter， Mohanan insists that th巴rεlS呂 loop"between str品ta2 and 3・43

(25) Stratum 1: Class 1 derivation， irregul且rinfl日ctlOn

「 →Stratum2: Class II derivation

し‑Stratum3・Compounding Stratum 4: Regul呂rinflection

For the explanation of some dialects of English in which compounds ar巴

subject to class II旦ffixation，Mohananラsidea of loop" is necessary.

4 . 2 .

Brackets vs. Boundary Markers

( J

unctures) Originally there emerged Lexical Phonology as the solver" of the prob‑ lems with SPE. This circumstance is conspicuously reflected by SPE's three kinds of boundaries:

十， #

and

# # 4 4

As far as

+

and

# 且

redis

tinguished， no other special rule is needed. In re丘lity，the first equ呂Isstra‑ tum 1 ， and the other strata 2 and 4. But

#

and

# #

have two kinds. In some dial巴cts，for instance

、

th哩applicationof Stem‑final Tensing rule pre‑ cedes the ‑s plural suffixation and follows that of the ‑ed past t巴nsesuffix Then.お thedomain of St巳m‑final t巴nsingrul巳 stratum3 is呂dopted the‑5

plural suffix must be呂dd己d呂tstratum 4. On the other handV巴larizationof /1/， which "darkをns"or velarizes the lateral /11 in the rim巴posltlOn18品

(17)

80 The Behavior of Phonological Rules at Strat

イ '

case where the distinction between two kinds of

: # : #

is required. Som巴

examp1es are shown b巴1ow:⁴⁵ (26)

R R

1ull [1 A

+ ]

be1t [bε 守tJ belly[bε lI]

As the exp1呂nationof these examp1es， Halle且ndMohanan呂ssumethe fo1 10wing ru1e^A6

(27) 1‑V elarization R

X 11/→[+ back] /

̲ 1 ̲

In addition such a distinction can b巳a1sofound between compounds and

1. 47

pnrases:

(28) a wha1巳editionvs. th巳wha1e呂ndthe sh乱rk

l

l [

[士

l

the sea1 office vs the sea1 offered a doughnut

l

l [

[+]

As these examp1田 show，the word‑fina1 11/ in compounds hints the巴XlS

tence of a vowe1‑initia1 word after itう this is formerly c品lled 1‑ Resy llabification: 48

(18)

The Behavior of Phonological Rules at Strata" 81 (29) l‑R巴syllabification

σ1 σ2 σ1 σ2

R

ハ

人

R

→

R

X X X X X X

[1] [lJ

This rule appears呂tstratum 4 b巳causeV

十

ingforms，丘sw巴11as com pounds，品r巴subjectto it. Consequently 1‑Ve1ariz旦tion.which is乱pp1icab1e at the post1exical stratum is preceded and blocked by 1‑R己syllabific且tion; is never resyllabified across words in English phras巴s(which a閃 com‑

pos己dat th巴postlexicalstratum)ー

4 . 3 .

Lexical Representations

The following昌pplicationof Trisyllabic Shortening and Fl呂pplllgIII

the word diviniかisan example of th巴 correlationbetween phonological rules and morphological rul巳sin Lexical Phonology:49

(30) Lexical module

[diviinJ [iti) UNDERLYING REPRESENTA TION [diviin)

[[diviin) [itiJ) [[divin) [itiJ) [diviniti)

Postlexical module [diviniDi)

[ diviniDi)

Stress Affixation Trysyl. Short

LEXICAL REPRESENT A TION

Flapping

PHONETIC REPRESENT ATION

(19)

The lexical representation in th巴 abov喧 illustrationis the intermεdiate level between the underlying and the phonetic. In this r巴gard，Mohanan m呂intainsth且t lexicalrepresentation is the phonological representation of lexical entriesη50 Similarly， underlying representation can be regarded as the representation of morphemes， and phonetic represent且tionas the out‑ put of the phonological module as Monanan's illustration showS:51

(31 )

morph印 le

morphological and phonological rule

品pplic且tions

. underlying repr巳sentatlOn

words/lexical entries . . . ・・・6 lexical

，

syntactic and

I

phonological rule

i

pplications

representatlOn

output of phonology . . . phonetic repr巴sentatlOn

In addition Mohanan insists that phonological processes that require syn‑

(20)

The Beh旦viorof Phonologic丘1Rules at Strata" 83 tactic information take pl呂cein syntactic modul巴，，52As an example of the syntactic module， h巴takesnuclear stress in the following two sentences:⁵³ John has plans to LEAVE. (=John wants to leave.) and John has PLANS to leave. (= J ohn wants to leave the plans.) Th巳outputof the syntactic module corresponds to the phonological phrase， and those representations that appear in the syntactic module旦recalled syntactico‑phonological rep‑ resentations (SP‑r巴presentations)ー Whatassigns thes巴 representatlOnsto phonetic repres巴ntationsis the module of phonetic implem巴ntationwhich contains a process similar to an allophonic process of classical phonemics

5. Conclusion

In this paper 1 have analyzed the application of phonological rul巴S1ll

terms of the rel呂tlOnto strata". As a conclusion， 1 shall briefly mention som巴significanceof the organization of strata. First of all， the distinction between the lexical and postl巴xicalstrata makes it possible to tr巴atmor‑ phological and syntactical processes separately. N巳xt such a rule as Trisyl‑ labic Shortening applies only at stratum 1， whereas on巴likeg Deletion is applicable at stratum 2， not at stratum 1: this contrast well refl巳ctsthe dis tinction between strata 1 and 2. The ex且mplesof St巴m‑finalTensing and St巴m‑finalLength白lIngshow the need to require the division betw田 n strata 1 and 2. The former rule applies at stratum 2 in some dialects of English， whil，巴in others， it applies at stratum 3， not stratum 2; the latter is applicable not at stratum 2 but at stratum 3 in some dialects of English

In且ddition，Syllable F ormation， which takes place品tstrat昌 1and 2， could be further evidence for the division between strata 2 and 3. The necessity for th巳 separ呂tionof st日tum3 from stratum 4 is w巳11explain巳dby such rules as Sonorant Resyllabification呂nd1 Resyllabification; both occur丘t

(21)

stratum 4， not at stratum 3

Eventually these strata brought丘boutmore domains than the previous theory did. SPE

，

for instance

，

divided domains in terms of

， : 十 1 *

and

: 1 * : 1 *

boundaries. and the distinction between

: 1 *

and

: 1 * : ， * 1

in particular

，

was un‑ clear. In this sense， Lexical Phonology， which classifies domains of En‑

glish by means of strata'¥made a considerable contribution to facilitating the understanding of phonological and morphological processes

Notes

1. Francis Kat呂mb且

，

An lntroduction to Phonology (New York: Longman

，

1989)

，

p. 257

2. Paul Kiparsky，Lexical Morphology and Phonology"， in 1. S. Yang (ed.). Linguistics in the Morning Calm (1982)， pp. 3~4

3. With regard to cyclicity. Mohanan (1982) gives an explanation that "Phono‑ logical rules apply in the lexicon after every morphological operation to the output of which morphological operations may apply once again、，in K P. Mohanan

，

Lexical Phοnology

，

Ph. D. dissertation

，

MIT

，

1982

，

p.12. As an ex‑ ample of cyclicity the following品pplic品tionof English stress rules is shown in Morris Halle and K. P. Mohanan Segmental Phonology of Modern English"， Linguistic

! J

u{uiry 16 (1985)， p. 66

[theatre]

→

[theatre]

→

[theatr] [ic]

→

[theatr] [ic]

→

[theatric]

→

[theatric] [al]

→

4. K. P. Mohanan， The Theory of Lαical Phonology (Norwell: Kluwer Academic Publishers， 1986)， p. 9

5. Douglas Pulleyblank， Tone in Lexical Phonology (Dordrecht: D. Reidel， 1986)ラ

p .

7

6. K. P. Mohanan

，

Lexical Phonology

，

Ph. D. dissertation

，

MIT

，

1982

，

p. 16. For details see Dorothy Siege ，l Topics in English Morphology. Ph. D. dissertation，

MIT， 1974. and Marg品目tAllen. Moゆhologicallnvestigati仰 s，Ph. D. dissert昌一

tion， University of Connecticut， 1978

7. John A. Goldsmith， Autosegmental and Metrical Phonology (Cambridge: Basil Blackwell， 1990)， p. 237

8. Morris Halle and K. P. Mohanan， "Segmental Phonology of Modern Eng‑

(22)

The Beh呂viorof Phonological Rules at Strata" 85 lish"， Lin伊，istic1 nquirツ16(1985)， p. 58

9. Francis Katamba， An Introdution to Phonology， p. 259

10. In Noam Chomsky and Morris HaIIe， The Sound Pattern of English (New York: Harper and Row， 1968)，

+

boundary is regardedωthe specifier of tr丘n‑ sition between phonetic elements or as the boundary between form且tives (minimal syntactic functional units)

11. Morris HaIIe and K. P. Mohanan，Segmental Phonology of Modern En‑ glish'¥p. 58

12. Paul Kiparsky Lexical Morphology and Phonology¥p. 5

13. In Noam Chomsky and Morris HaIIe， The Sound Patter百 ofEnglish， # bound

旦ryis considered to indicate the boundary between lexical categories (i.e.N，

V，Adj)

14. Morris HaIIe and K. P. Mohanan， "Segmental Phonology of Modern Eng‑

lish"， p. 58

15. In John A. Goldsmith， Autosegmental and Metrical Phono，Z口'gy，p. 246， Trisyll呂

bic Shortening is shown昌sfoIIows V

→

[‑long]l Co

I

V

I

CoV

‑stress I

16. Flapped sounds are articulated by friction betwe巴nthe bl且deof the tongue and the p呂late.As the examples of Flapping there appear the ch且nges(diviniか

→

divinziρ]y，ωnter

→

ωrziρ]er) in Morris HaIIe and K. P. Mohanan， 'ζSegmen tal Phonology of Modern English"， p. 59

17. Locαt

18. This figure is cited from Noam Chomsky and Morris Halle， The Sound Pat‑ t計百ofEnglish

19. K. P. Mohanan， The Theory of Lexical Phonology， p. 26 20. Loc. cit

21. Morris HaIIe and K. P. Mohanan， "Segmental Phonology of Modern Eng‑

lishへp.60.

22. S巴e，however， K. P. Mohanan and T Mohanan， "Lexic^在1Phonology of the Consonant System in Malayalam¥Linguistic Inquiry 15 (1984)， pp. 575‑602 23. Morris HaIIe and K. P. Mohanan Segmental Phonology of Modern Eng

lishぺp.60

(23)

86 The Behavior of Phono1ogica1 Ru1es at Str且ta"

24. Ibi， d.p. 61 25. lbid.， p. 62

2 6 .

Loc.αι

27. lbid， pp. 62‑63 28. lbi .，d

p .

63 29. Loc. cit.

30. K. P. Moh且nan，The Theor

ツ

0/Lexical Phonology， p. 32 31. Loc. ciι

32. Loιαt

33. lbi ，.dp. 30. Some similar an呂1)ァsesc且nbe seen in Pau1 Kiparsky， "Metrical Structure Assignment is Cyclic"， Linguistic lnquiry 10 (1979)， pp. 421‑442 35. Loc. cit

3 6 .

Loc. cit. 37. lbi .，dp. 34 38. L

ョ

c.czt 39. lbi， d.

p .

36 40. Lo

ι

Clt.

4

1. Loc. cit 42. Ibi， d.

p .

37

43. Morris Halle and K. P. Mohan在日， Segmenta1 Phonology of Modern En.

glish"， p. 64

44. In N oam Chomsky and Morris Halle， The Sound Pattern 0/ English， :j:j: :j:j: boundary is used to indicate the boundary between units which are higher than the word like compounds and phrases， whereas :j:j: boundary is entirely the marker of the boundary between individua1 words

45. Morris Halle and K. P. Mohanan，Segmenta1 Phono1ogy of Modern En.

glish"， p. 65

4 6 .

Loc.ι α 47. Loc.

α

. t

4 8 .

Loc. cit.

49. K. P. Mohanan， The Theory 0/ Lexical Phonology， p. 10 50. Loc.

α

. t

51. Ibid.二p.11 52. Loc.

α

. t

53. Loc.α .tM在kereference to Joan W. Bresnan， "Sentence Stress and Syntactic Transformations"， Language 47 (1971)， pp. 373‑384