96. Metrical patterns · is [ra:t.le] rather than *[ra:.dle]. Maximisation of the onset is closely...

134796. Metrical patterns

Ruhlen, Merrit. 1987. A guide to the languages ofthe world. Stanford: Stanford University Press.

Schane, Sanford A. 1973. Generative phonology.Englewood Cliffs.

Skalicka, Vladimir. 1979. Typologische Studien.Braunschweig/Wiesbaden: Vieweg.

Sommerstein, Alan H. 1977. Modern phonology.London: Arnold.

Stampe, David. 1979. A dissertation on NaturalPhonology. New York & London: Garland.

Trubetzkoy, Nikolaj S. 31958. Grundzüge der Pho-nologie. Göttingen: Vandenhoek & Rupprecht.

Ultan, Russel. 1978. „A typological view of me-tathesis“. In: Greenberg 1978, 367�402.

van der Hulst, Harry & Smith, Norval. 1982. „ Anoverview of accent and metrical phonology“. In:van der Hulst & Smith (eds.) 1982, 1�45.

van der Hulst, Harry & Smith, Norval (eds.). 1982.The structure of phonological representation. Dor-drecht: Foris Publ.


1. Introduction2. The syllable as a constituent3. The foot4. Typology and metrical structures5. Universals, implications and correlations6. Conclusion7. References

1. Introduction

Metrical patterns in languages are obtainedby combining various elements of prosodicstructure: syllables and their constituents,feet, and other higher level organisationalunits like prosodic words, phrases and so on.Within a given metrical organisation, a par-ticular constituent may be the most promi-nent. This relative prominence is marked bystress, which is the central theme of this arti-cle. Stress, under this conception, is not mer-ely a phonetic feature, but is the means ofmarking relative prominence within variousorganisational groupings of metrical units(cf. Liberman 1975; see the articles by Ka-ger, and Halle & Idsardi in Goldsmith 1995for surveys of different metrical theories ofstress). In order to establish stress patterns,we first discuss how different metrical con-stituents are relevant for the phonologicalsystems as a whole. Since the covariation of

Vennemann, Theo. 1971. „The phonology of Gothicvowels“. Language 47: 90�132.

Vennemann, Theo. 1972. „Phonetic detail in assim-ilation: Problems in Germanic phonology“. Lan-guage 48: 863�892.

Vennemann, Theo. 1986. Neuere Entwicklungen inder Phonologie, Berlin etc.: de Gruyter.

Vennemann, Theo. 1988. „The rule dependence ofsyllable structure“. In: Duncan-Rose, Caroline &Fisiak, Jacek & Vennemann, Theo (eds.). Rheto-rica, Phonologica, Syntactica: A Festschrift forRobert P. Stockwell, London & New York: Rout-ledge, 257�283.

Vihman, Marilyn M. 1978. „Consonant harmony:its scope and function in child language“. In:Greenberg 1978, 281�335.

Vogel, Irene. 1982. La sillaba come unita fonolo-gica. Bologna: Zanichelli.

Thomas Krefeld, München (Deutschland)

linguistic variables is fundamental to lan-guage typology (cf. Plank 1997), our goal isnot merely to list the observed metrical pat-terns, but also to examine possible relation-ships between the different patterns.

To this end we will focus on ‘metrical co-herence’ from two perspectives and addressthe following questions. First, we ask whethera given metrical constituent varies in itsproperties within a single language. For in-stance, the metrical constituent ‘foot’ is gen-erally used to account for word stress. How-ever, there may be other processes which aresensitive to foot structure. If so, one wouldlike to know if foot types vary for differentprocesses within a given language, or whetherwith respect to a given metrical constituent,the system is coherent (Dresher & Lahiri1991). The second issue is whether the typeof stress a language has can predict the prop-erties of its metrical constituents. This par-ticular perspective has not been an issue inthe phonological descriptions of metrical pat-terns, but is extensively discussed in typologi-cal literature on the covariation of stress withthe nature of syllables, headedness of phrasalstress and such (cf. Donegan & Stampe 1983;Gil 1986). Thus, we begin by motivating syl-lables and feet as necessary metrical constitu-ents in the description of phonological sys-

1348 XII. Phonology-based typology

tems. For each of the constituents we provideevidence from segmental processes as well asfor stress, and then move on to issues on met-rical coherence and covariation of metricalunits, which are rather crucial for typologi-cal research.

2. The Syllable as a constituent

In this section, we first review the syllable’sstatus in phonology before discussing the roleof the syllable in the assignment of stress.The syllable has traditionally been assumedto consist of an onset followed by a rhymewhich is divided into a nucleus and a coda.The nucleus is the obligatory and most im-portant part of the syllable, while the onsetand coda are optional. The most frequent syl-lable inventory in natural language consistsof the following: V, CV, VC, CVC (see Blev-ins 1995 for a survey). The more complex syl-lable inventories arise from including moresegmental material in the onset and the coda,and even the nucleus can be branching. Com-plex onsets and codas are generally governedby the Sonority Scale which states that onsetconsonants increase in sonority and codasdecrease in their sonority (cf. Clements &Hume 1995). The accepted sonority scale interms of rising sonority is obstruents � na-sals � liquids � glides � vowels.

The notions ‘closed’ and ‘open’ syllablesplay an important role in phonology. Closed

(2) Resyllabification in Germanglaub [p] glaub-en [b] glaub-lich [p/b] ‘believe’ 2sg imp./inf./adj.Tag [k] Tag-e [g] täg-lich [k/g] ‘day’ sg./pl./adv.

syllables are those which are closed by acoda consonant, while open syllables end ina vowel (long or short) or a diphthong. Todecide whether medial consonants are partof onsets or codas, the principle of maximi-sation of the onset is often invoked. That is,when there is more than one intervocalicconsonant, whether all of them are part ofthe onset of the second syllable, depends onwhether the language permits ‘maximisingthe onset’ based on sonority principles.Phonological processes can help determinewhether consonants fall in the coda or not.This is illustrated with an example from Ger-man which has a process of syllable finaldevoicing. The data are from Vennemann(1972). German has a rule of syncopationwhich follows for the following types of al-ternations.

(1) Syncope and syllable final devoicingin German

Standard German‘flirt’ ‘sail’ ‘go by bicycle’

Infinitive li:bel�n ze:gel�n ra:del�n1sg.ind.pres li:bl�e ze:gl�e ra:dl�eStandard German, Northern pronunciation

li:bl�e ze:gl�e ra:tl�e

After syncopation, the consonant clustersthat are created are not equally accepted asonsets in the Standard German as comparedto the Northern pronunciation. In StandardGerman, the sequence [dl] is accepted as asyllable onset, and the maximisation of onsetprevents the [d] being in the coda. Hence,coda-devoicing does not apply. In contrast,the Northern pronunciation which allows[bl] and [gl] clusters, permits maximisation ofconsonants in these cases, but prevents [dl]from being part of an onset. As a result,coda-devoicing applies and the surface formis [ra:t.le] rather than *[ra:.dle].

Maximisation of the onset is closely re-lated to the notion of a core syllable, or a CVsyllable. There is a general tendency to avoidonsetless syllables such that in most if notall languages, a VCV string is syllabified as[V.CV]. Resyllabification to prevent onsetlesssyllables is central to the analysis of Germandevoicing as well (cf. Rubach 1990, Giegerich1992). The following alternations are rele-vant.

As we have seen before, coda devoicing makesthe word final consonants in the first columnvoiceless. A suffix vowel is added to thewords in the second column. Here the medialsequence VCV is syllabified as [V.CV] forcingthe medial consonant to be an onset, therebyblocking coda devoicing. Oddly enough, whenthe suffix begins with a sonorant consonant,and although the obstruent � liquid is a pos-sible onset (as we saw in the previous exam-ple), resyllabification can be blocked forcertain speakers and coda devoicing applies.Obviously, for those speakers who devoicethe obstruents, resyllabification is sensitive tocertain morphemes even if allowable onsetsmay arise. However, the crucial point is thatwhen a suffix with an initial vowel follows,resyllabification is obligatory since Germanalways requires a syllable with an onset.


Words or syllables without a surface conso-nant are always preceded by a glottal stop:cf. Atmen [?atmen] ‘breathing’, abteilen [?ap-tai len] ‘to separate’, mitarbeiten [mıt?arbai -ten] ‘to cooperate’ etc. For some speakers theglottal stop insertion is restricted to stressedsyllables; hence, Theater [the?a6tB]) ‘theatre’,but Bebauung [bebaw wn] ‘building develop-ment’, and not [bebaw ?wn].

2.1. Preferred syllable structureOnce we accept the fact that languages havepreferred syllable structures, any deviationfrom these preferences are repaired. Strate-gies for repairing them can differ. For in-stance, if affixation leads to unacceptablesyllables, either epenthesis or syncope are in-voked to maintain the preferred structures.In a language like Koryak (a Paleosiberianlanguage spoken in Kamchatka; Spencer 1996:63�64), the most complex syllable structurepermitted is CVC. Hence any affixation whichleads to complex structures is resolved byschwa epenthesis.

(3) Koryak schwa epenthesisVerb root /pnlo/ ‘ask’Prefixes: t- 1sg.subj. mt- 1pl.subj.,na- 3pl.subj.

(a) t-pnlo-n tep.ne.lon ‘I asked him’(b) mt-pnlo-n met.pen.lon ‘we asked him’(c) na-pnlo-n nap.ne.lon ‘they asked him’

If the segments are syllabified from left toright obeying the preferred CVC syllabic tem-plate, then the introduction of the schwa isentirely predictable. If we did not assume thatepenthesis was syllable based, it would notbe possible to account for the difference be-tween the schwa insertions in the verb rootin (3a) and (3b): pnel vs. penl.

Epenthesis is one of the most frequentways to resolve unwanted clusters and toobtain a preferred syllable template. Relatedlanguages often exhibit a difference in theacceptance of initial and final clusters. Astriking example comes from certain final li-quid � obstruent clusters in Germanic lan-guages. English and German allow [l � ob-struent] clusters in words like milk or Milch,but Dutch disallows such clusters and intro-duces a schwa as in melek.

Along with epenthesis, deletion is anothermeans for cluster simplification. In Bengali,the present indicative ending begins with ageminate affricate -tstsh which is degeminatedwhen added to a verb root ending in a con-sonant (Fitzpatrick-Cole 1994, 1996; Lahiri2000).

(4) Bengali degemination as cluster sim-plification

(a) su-lam su-tstshi ‘sleep 1past/1present’

(b) bos-lam bos-tshi *bos-tstshi ‘sit 1past/1present’

Bengali does not allow coda clusters. Sincea geminate consonant belongs to the coda ofone syllable and the onset of the followingsyllable, if the preceding syllable ends in aconsonant, the geminate introduces a codacluster and is degeminated to fit the syllabletemplate of the language.

Thus, both deletions and insertions arefrequently found in languages, and almost al-ways in the context of repairing an unaccept-able syllable. Preference for syllable types,and hence repairs, is usually restricted to thelexical level. In the postlexical level, there ismore variation. The last example of degemi-nation can also be viewed as shortening, andas we will see in § 2.3., lengthening and short-ening phenomena are also linked to syllablestructure. However, in these cases it is theweight of the syllable which plays a crucialrole.

2.2. Syllable quantity and weightOne view of representing syllable weight is byusing moras. The moraic theory of represen-tation views moras as phonological positionswhich come between prosody and segments(rooted in the feature tree). Long and shortvowels, and long and short consonants (i. e.geminate and single consonants) are dif-ferentiated by their moraic representation.Moraic representations in (5) are based onHayes (1989).

(5) Moraic representationsm m m mt � � t t[a] [a:] [p] [p:]

Short vowels have one mora, long vowelshave two moras, a single consonant has nomoras, and a geminate consonant comes withone mora. A single consonant is not assigneda mora in the lexical representation. It mayor may not be assigned a mora depending onwhether it is in the coda and whether the lan-guage treats closed syllables as heavy. If thecoda is counted as heavy, then weight-by-po-sition assigns a mora to the coda consonant.Geminates, on the other hand, are part of theonset of a syllable, but must close the pre-ceding syllable as well, automatically adding


weight to this syllable. (A problem arises inlanguages where geminates do not contributeto weight but long vowels do; see Lahiri &Koreman (1988), Hayes (1989), Kager (1989)for further discussion.) Hypothetical syllabi-fications are given below.

(6) Syllable structure assignment

Just as languages often try to preserve pre-ferred syllable structures, we often find pro-cesses which attempt to maintain the weightof a syllable. Bimoraic syllables are heavy, ir-respective of whether they are closed syllables(the coda consonant adding weight to thesyllable), or whether they have a long vowel.However, not all languages necessarily con-sider closed syllables to be heavy. Languagestend to avoid trimoraic syllables althoughthey do exist. Further consequences of sylla-ble weight will be discussed when we con-sider stress.

2.3. Compensatory lengthening

Similar to deletions and insertions, shorten-ing and lengthening processes are closely re-lated to the syllable. A frequent process oflengthening is compensatory lengthening,where the loss of a segment is compensatedby lengthening an adjacent segment. This canbe accomplished by total assimilation or byvowel lengthening. For instance, in Bengalian [r] followed by a coronal consonant is op-tionally deleted and the consonant becomesa geminate (Hayes & Lahiri 1991). The as-similation can apply within words, acrossmorphemes, as well as across words, the con-straint being that the [rC] sequence must be-long to a single phonological phrase. Someexamples are given in (7).

(7) Bengali total assimilationpcrda pcd6a ‘curtain’por-tam pot6am ‘wear-1sg.past

habitual’ghcr dzamai ghcdz6amai ‘house son-in-law;

son-in-law wholives in the houseof his in-laws’

Other common instances of compensatorylengthening involve the loss of a coda conso-nant which leads to the lengthening of thepreceding vowel. We find this in Old Englishwith the loss of a coda nasal. If we comparethe words for five and tooth in Old HighGerman, Old English and their modern de-scendants, we find the pattern in (8). Sinceungrammatical forms are marked elsewherewith an asterisk, the Proto-Germanic recon-structed forms will be indicated with thesign †.

(8) Compensatory lengthening in Ger-manic

German OHG English OE Proto-Germanic

fünf fimf, five fıf †fimfifumf

Gans gans goose gos †gans

The Proto-Germanic words had a short vowelfollowed by a nasal consonant. The nasal hasbeen retained in German and the vowels arestill short. The loss of the nasal in English,however, has led to long vowels (whichwere later sometimes diphthongised) � aninstance of compensatory lengthening whichwe can represent in a nonlinear fashion. In(9a), V and N represent any vowel or a nasal.Since long vowels are bimoraic, delinkingafter the loss of the nasal and reassociation,gives us the desired result. In (9b), the sameeffect is realised for the [r] deletion and con-comitant gemination in Bengali, except thatonly a single mora is involved. Here the mora,which was originally linked to the [r] in thecoda, is then linked to onset consonant (rep-resented by C) in the next syllable, thus creat-ing a geminate.

(9) Compensatory Lengthening as spread-ing

(a) Germanicm m m mt t J � �V N V


(b) Bengali

It is worth noting that when the loss of theconsonant in such circumstances is closelylinked with vowel lengthening, it is invariablyconfined to a particular syllable position, anda similar loss elsewhere in the phonology ofa language will not show any concomitantlengthening. For instance, in Old English the[n] sometimes disappeared between conso-nants: OE elboga beside elnboga ‘elbow’; OEsAterdAg beside sAterndAg ‘saturday’.

(10) Loss of nasal in Old English not lead-ing to compensatory lengtheningCompensatory lengthening as spread-ing onto a free mora

In (10) the deleted [n] is not immediately pre-ceded by a vowel. It is in a branching coda,sharing the mora with another consonant.The loss of the nasal does not free the moraof the coda and therefore there is no spread-ing and no lengthening. Thus, compensatorylengthening can be viewed as maintaining theweight of a syllable.

2.4. AmbisyllabicityThe notion of ambisyllabicity has been usedin two ways: as an environment for syllable-based processes and as a means of providinga coda to add weight to a syllable. The mostfrequently discussed phenomena where am-bisyllabicity plays a role are aspiration andflapping in English (cf. Kahn 1976; Gussen-hoven 1986). Both processes are governedby surface syllable structure, and hence arestress-sensitive. The ambisyllabicity resultsfrom the attraction of the first consonantalonset of an unstressed syllable to form a codaof the preceding syllable. This consonant thenbecomes ambisyllabic, since it belongs bothto the onset and the coda of two syllables.

The usual onset and coda constraints of thelanguage apply. This procedure leading toambisyllabicity is labelled as Extended RightCapture in Gussenhoven (1986: 130), the for-mulation of which is based on two differentprocesses in Kahn (1976).

(11) AmbisyllabicityExtended Right Capture

Ambisyllabicity accounts for a number ofpostlexical phonological rules of AmericanEnglish like flapping, aspiration, glottaliza-tion etc. Flapping weakens coronal stops[t, d] to a flap [J] when they are ambisyllabic.This accounts for why the coronal stops inlater, shouting, matter are subject to flapping,while those in latex, bait, tail are not. In thelatter set of words, the stops are either fol-lowed by a stressed syllable (cf. latex), areonly in the coda (cf. bait), or only in the onset(cf. tail), and hence none of them are ambi-syllabic. Similarly, aspiration is also subjectto ambisyllabicity. Aspiration of voicelessstops in American English occurs when inabsolute syllable onset position, and ambi-syllabic consonants cannot be aspirated. Thisis different in British English where absoluteonset position is not required for aspiration.Thus, words like happy, where the medialconsonant is ambisyllabic, may be aspiratedin British English, but never in AmericanEnglish. However, British English also re-quires ambisyllabicity as a structural possi-bility, since rules like weakening (which‘weaken the oral closure of obstruents’ in fastinformal speech, Gussenhoven 1986: 125�6)can operate on the output of aspirated conso-nants. However, weakening only operates onambisyllabic aspirated consonants, and thosethat are in absolute onset position are ex-empt.

As we mentioned above, ambisyllabicityhas also been argued to play a role in assign-ing syllable weight (cf. van der Hulst 1985and Lahiri & Koreman 1988 for Dutch;Ghini 2001 for Miogliola, a northern Italiandialect). Under these analyses, ambisyllabi-city not only allows a consonant to be partof an onset in one syllable and a coda in theother, the coda consonant also projects a


mora. The representation would look as fol-lows:

(12) Ambisyllabicity and weight

After ambisyllabicity, the phonological repre-sentation is identical to that of a geminate.However, in languages in which ambisyllabi-city is invoked for syllable weight, there areno contrastive geminates (cf. also Borowsky,Ito & Mester 1984). Whether one could thenassume that all such ambisyllabic consonantscould be treated as geminates is a much de-bated topic.

The notion of ambisyllabicity has notfound favour with many researchers particu-larly because of the dual linking of a singleconsonant to two syllables (Kiparsky 1979;see also Blevins 1995 for a discussion). How-ever, the arguments in Gussenhoven (1986)are very persuasive and since dual linking hasto be permitted for geminate consonants thatin itself is not a sufficient argument againstambisyllabicity.

So far we have focused on two differentaspects of syllable structure: syllable as acontext for phonological rules and strategiesto maintain preferred syllable structure andsyllable weight. We now move on to discussthe role of syllable weight and its interactionwith metrical stress.

3. The foot

A fundamental insight in metrical theory isthat syllable weight plays a crucial role instress assignment. As we mentioned above,the weight of a syllable usually depends onwhether it has a long bimoraic vowel orwhether the coda of a closed syllable con-tributes a mora to the syllable. Vowel qualityis never taken into account where syllableweight is concerned. It would be very odd in-deed, if for instance, all front vowels weretreated as heavy while other vowels includinglong vowels were light. However, syllables arenot sufficient to account for stress assign-ment in languages of the world. Recent theo-ries of metrical stress argue that the foot,which is a constituent built on groups of syl-lables, accounts for stress. In general, sylla-

bles are considered to be grouped into metri-cal feet consisting of strong and weak sylla-bles. The feet differ in terms of whether thehead of the foot, i. e. the stressed position,occurs at the left or right edge. A left headedfoot is known as a trochee, and a rightheaded foot is an iamb. We will not however,begin with the assumption that syllableweight and feet are crucial elements in stressassignment. Instead, with examples from twolanguages, we will trace step-by-step themotivations for assuming (a) syllables are anecessary constituent for assigning stress, (b)syllables are not enough to account for uni-versal stress systems, (c) a fixed inventory offoot types built on syllables can delimit allstress patterns and (d) the weight of syllablesplay a role in building feet.

3.1. Are syllables necessary for stress?Chomsky & Halle (1968) accounted for Eng-lish stress using only a linear sequence ofconsonants and vowels, ignoring any hier-archical constituent like the syllable. How-ever, referring only to a linear sequence is notenough. Let us consider the facts of the wellknown Latin stress rule, which has been dis-cussed in metrical terms for a long time. Thelength of a vowel is indicated with a macron.

(13) Latin stress3rd vowel 2nd vowelfrom end from endmınimus refectus incudismurmuris voluptas relatusexıstimo delectat inimıcusadsimıliter excerpsit refecit

If asked which vowels are stressed withouttaking recourse to syllables, one would comeup with the rule in (14):

(14) Latin stress rule based on vowels(i) If the penultimate vowel is long, it is

stressed.(ii) If the penultimate vowel is followed

by two consonants, it is stressed.(iii) Else, the antepenultimate vowel is

stressed.

Such a rule, however, would cause problemsfor the following words.

(15) Problem cases3rd from left Expected 2nd from lefttenebras *tenebrasvolucres *volucresmaniplis *manıplislatebras *latebras


If we take the second part of the stress rulewhich says that if the second to last vowelis followed by two consonants it should bestressed, we have incorrect results. The solu-tion lies in the evidence from syllabificationfollowing the sonority hierarchy. The abovewords are syllabified as [te.ne.bras], [vo.lu.cres], [ma.ni.plis] and [la.te.bras], as against[re.fec.tus], [vo.lup.tas] etc. Thus, the stressrule can be simplified as follows: If the pe-nultimate syllable has a short vowel with nocoda then the antepenultimate syllable bearsthe main stress; otherwise the penultimatesyllable is stressed.

This description is very characteristic ofstress rules. Syllables with long vowels orwith a coda consonant pattern together. Andnow we come to syllable weight. As we haveseen before, this division is descriptivelycharacterised as heavy syllables versus lightsyllables. The Latin stress rule can then bestated as:

(16) Latin stress rule in terms of syllableweight

(i) If the penultimate syllable is heavy, itis stressed.

(ii) Else, the antepenultimate syllable isstressed.

3.2. Are syllables enough for stress?So far we have seen that a linear string ofconsonants and vowels is not enough toaccount for stress. Instead we require thenotion of syllables, and particularly syllableweight. But is syllable weight enough to cap-ture stress facts of various languages? Thatis, is it always the case that stress assignmentcan be characterised in terms of heavy andlight syllables? The answer is no. Let us lookat a more complicated case � Creek, aMuskogean language. The data in this papercomes from Haas (1977). Haas describesCreek as having tonal accent, which falls ona ‘key syllable’ (p. 195). There can be morethan one ‘key’ syllable, each one being ‘onestep lower than the preceding’ one (p. 196).The tones themselves can be level, falling orrising. Assuming that the key syllables are theprominent syllables indicating main and sec-ondary stresses, Creek provides us with a richsource of data. Further data is given in Hayes(1995: 64�65). Consider the following facts:

(17) Creek data (length is indicated withthe diacritic [:])pocoswa ‘axe’cofı ‘rabbit’

acolakı ‘old timers, elders’osahwa ‘crow’ahicita ‘one to look after’ca:lo ‘trout’sokca ‘sack’famı:ca ‘canteloupe’

If we assume that all consonant clusters arebroken up into onset and coda, and thatclosed syllables are heavy, Creek stress can bedescribed as follows:

(18) Creek stress � first approximation(i) If the penultimate syllable is heavy, it

is stressed.(ii) Else, the last syllable is stressed.

So far, the Creek data look quite similar toLatin, the only difference being that in Creekif the penult is not heavy the last syllable isstressed. Now let us consider a few morewords.

(19) Further data in Creekifoci ‘puppy’imahicıta ‘one to look after for

(someone)’itiwanayipıta ‘to tie each other’acahankatıta ‘one to count me’

In the above words, the penultimate syllableis light, but nevertheless it is stressed. Ac-cording to our preceding assumption, the lastsyllable should have borne stress. Perhapswe can salvage the analysis by the followingstatement:

(20) Creek stress � second approximation(i) If the penultimate syllable is heavy, it

is stressed.(ii) If the penultimate syllable is light,

stress the final or penultimate which-ever is even-numbered, counting leftto right.

Unfortunately this does not solve the prob-lem. Consider the following words:

(21) More data from Creekaktopa ‘bridge’wa:kocı ‘calf’hoktakı ‘women’inkosapita ‘one to implore’

Clearly, our previous rules will not suffice. Inall the words, the penult is light, but the syl-lable that is stressed is not even-numberedcounting left-to-right. What we need to do isnot to start counting from the beginning ofthe word, but from the rightmost heavy sylla-


ble. The stress rule could then be describedas follows:

(22) Creek stress � third approximation(i) If the penultimate syllable is heavy, it

is stressed.(ii) If not, examine the maximum string

of light syllables at the end of theword.

(iii) Within this string, stress the right-most even-numbered syllable count-ing left-to-right.

We go through two examples following thesteps elaborated above.

(23) Deriving Creek stress � third ap-proximationinkosapita acahankatıta

(i) � �(ii) in (ko sa pi ta) acahan (ka ti ta)

(iii) * *1 2 3 4 1 2 3

in (ko sa pi ta) acahan (ka tı ta)

This description is not particularly illuminat-ing. Clearly we are missing a generalisation.If the penult is not heavy, it is not syllablecounting that gives us the right answer, butsome constituent which groups syllables to-gether. Such a constituent in poetic meter isknown as a foot, and in the next section wediscuss the universal inventory of feet thathave been suggested for natural language.

3.3. Inventory of feetHayes (1995) argues that there are three basicfoot types used in linguistic systems univer-sally: a syllabic trochee, a moraic trochee, andan iamb. The syllabic trochee groups any twosyllables together regardless of their weight.A moraic trochee and an iamb are weightsensitive. These three foot types are givenbelow. The foot is demarcated between pa-rentheses and the strong and weak branchesare indicated by a [x] and a dot [.] respec-tively.

(26) Syllabic trochee (Pintupi, a Pama-Nyungan language of Australia)Foot construction: Left to RightMain stress: End Rule Left (indicated with X)(X ) (X ) (X ) (X )(x .)(x .) ( x .)(x .) (x .)(x .)(x .) (x .)(x .)(x .)s s s s s s s s s s s s s s s s s s s s s s

malawana pulinkalatju tjamulımpatjunku tılirınulampatju‘through from behind’ ‘we (sat) on the hill’ ‘our relation’ ‘the fire for our

benefit flared up’

(24) Trochees and Iambs(a) Syllabic trochee (weight insensitive)

(x .)s s

(b) Moraic trochee: left headed (con-structed over two light syllables orone heavy syllable)(x .) (x)s s st t � �m m mm

(c) Iamb: right headed (construed overtwo light syllables, a light plus aheavy syllable, or one heavy syllable)(. x) (. x) (x)s s s s st t t � � � �m m m mm mm

Although the moraic trochee and the iambare both weight sensitive (i. e. the weakbranch cannot be heavier than the strongbranch), under Hayes’ analysis these two feetare asymmetric. Under this system, an iambmay have a [L(ight) H(eavy)] sequence, but atrochee is not permitted to have a branchinghead. The way stress assignment works is asfollows. A string of syllables are parsed intofeet going from left-to-right or right-to-left.The last foot on the left or the right is assignedmain stress: End Rule (left/right). Thus, mainstress is always at an edge of a word, edgebeing defined by foot structure and not by syl-lables or vowels. To assign stress, we thereforerequire the following parameters:

(25) Stress assignment(a) Foot type(b) Direction of parsing(c) End Rule

We illustrate this first with the most straight-forward foot type, namely the syllabic tro-chee, which is weight insensitive. The syllabictrochee groups syllables together regardlessof their internal structure. The analysis isfrom Hayes (1995: 62�63).


Let us now turn back to Creek and investi-gate which of the foot types would be appro-priate to account for the entire set of data.Clearly syllable weight plays a role since thepenultimate syllable is stressed only when itis heavy. The End Rule appears to be on theright, since stress falls always towards theright edge of the word. Now we need to de-termine the direction of parsing and whetherthe foot type is a moraic trochee or an iamb.The syllabic trochee cannot be consideredsince it is quantity insensitive. The decision isnot a difficult one since in our third approx-imation we saw that when a sequence oflight syllables occur at the end of a word, therightmost even numbered syllable can getstressed. As a result final light syllables maybear stress and this is not possible for a tro-chee. Thus, if the foot inventory is indeed suf-ficient, then the foot type must be an iamb.

The final decision regarding stress assign-ment must be the direction of parsing. Againin the last approximation, if the penult wasnot stressed, the grouping of syllables into alarger constituent began after the last heavysyllable. Hence, the parsing must be fromleft-to-right. Following Hayes (1995), apply-ing these parameters to Creek we obtain thefollowing structures:

(27) Stress assignment in Creek: final versionSyllable weight: Long vowels and closed syllables are heavyFoot type IambFoot construction: Left to RightMain stress: End Rule Right( X) ( X ) ( X) ( X ) ( X )(. x) (. x) (. x)(. x) (. x)(. x) (. x)m m m m m m m m m m m m m m m mm mco fı i f o c i a co la kı i mahi cı ta fa mı: ca( X) ( X) ( X )(x)(. x) (x)(. x)(. x) (. x)(x) (. x)mm m m mm m m m m m m mm m m mwa: ko cı in ko sa pi ta a ca han katı ta

Thus, although syllables may provide an ade-quate description for stress patterns in somelanguages, they are not sufficient to accountfor the complicated systems like Creek. Oncewe introduce a larger constituent groupingsyllables together into feet, the analysis of thestress pattern becomes very simple.

In addition to the three basic foot tem-plates, it is necessary to invoke the notionof extrametricality to understand some otherstress patterns. Syllables or segments (usuallyconsonants) at right edges are often extra-metrical; that is, they behave as if they do

not exist for footing and, therefore, not forstress. Thus, there are four parameters to betaken into account: foot type, extrametri-cality, direction of foot parsing and the endrule. In the two following examples takenfrom Hayes (1995) we see instantiations ofthe moraic trochee with and without extra-metricality (cf. (28)).

This brings us back to Latin stress. Recallthat in Latin, stress fell on the penultimatesyllable if it was heavy. Otherwise the ante-penultimate syllable bore stress regardless ofweight. We can now analyse Latin in thefollowing way (cf. (29)).

The moraic trochee along with the extra-metricality does away with the oddity of thesyllable based description which required thatsyllable weight was responsible for attractingstress on to the penultimate syllable but notfor the antepenultimate syllable. The antepe-nult could be stressed regardless of syllableweight; it depended on the lack of weight ofthe penult. In the foot based analysis, the ex-planation rests on the fact that the antepenultand the penult together can make up a singlefoot if both are light.

The inventory given above excludes thepossibility of asymmetric moraic trocheeswhich are the mirror images of iambs. How-

ever, others like Dresher & Lahiri (1991),Lahiri & Dresher (1999, for Germanic), Ja-cobs (1989, 2000, for Latin), Kager (1989, forEnglish) have claimed that asymmetrictrochees incorporating [H L] sequences arenecessary as well. For instance in Latin, usingan asymmetric trochee would mean thatwords like murmuris would be parsed intofeet as ([mur. mu] *ris+). The traditional tro-chee is, in fact, asymmetric (cf. Hayes 1981).Although we are not in a position here to ex-haustively compare these various proposals,while discussing the foot based phonological


(28) Moraic trochees with and without extrametricalityMoraic trochee (with extrametricality)Cairene Arabic (No classical words are considered)Syllable weight: Long vowels and closed syllables are heavyExtrametricality: Final consonant of a word (indicated by * +)Direction of Parsing: Left to RightMain stress: End Rule Right(X ) ( X) ( X ) ( X ) ( X ) ( X )(x) (x) (x) (x) (x) (x.) (x .)(x .)mm m m mm m mm m mm m m mm m m m m m mbe:ta*k+ gato: katab*t+ mudarri*s+ mudarrısi*t+ katabıtu‘your (m.sg.) house cake’ ‘I wrote’ ‘teacher’ ‘teacher (f. construct)’ ‘she wrote it (m.)’

Moraic trochee (Wargamay: Pama-Nyungan language of Australia)Syllable weight: Long vowels are heavyDirection of parsing: Right to LeftMain stress: End Rule Left(X ) (X ) ( X ) ( X )(x) (x .)(x .) (x .) (x .) (x .)mm m m m m m m m m m m m m mmu:ba gıJ-awulu gagara J-uragaymıri‘stone fish’ ‘freshwater jewfish’ ‘dilly bag’ ‘Niagara-Vale-from’

(29) Latin stress revisitedSyllable weight: Long vowels and closed syllables are heavyExtrametricality: Final syllableDirection of Parsing: Right to LeftMain stress: End Rule Right( X ) ( X ) (X ) (X )

(x) (x) (x) (x .) (x)m mm mm mm mm mm m m mm mm m mmre fec *tus+ de lec *tat+ vo lu *cres+ mur mu *ris+

(30) Constraining gemination in OEGemination blocked:(x .) (x .)([mm] m) [mmm] m ([m m] m) [m mm] mH L L L Lwı tje � *wıtte æ pe lje � *æ pel le‘punishment dat. sg.’ ‘noble dat. sg.’

Gemination permitted(x .) (x) (x .) (x ) (x .)([mm] m) m ([mm]) ([mm] m) ([m m]) ([mm] m)H L L H H L L L H Lwe ste nje � we sten ne cy nje � cyn ne‘desert dat. sg.’ ‘race dat. sg.’

processes, we will draw on evidence from seg-mental rules and stress in Germanic to pro-vide support for asymmetric trochees.

3.5. Foot-based phonological processesPhonological processes can be sensitive to thefoot. The foot relevant for Germanic was aresolved moraic trochee, which is essentiallyan asymmetric trochee, where the head must

branch. An example of a foot based processis West Germanic gemination, which is sim-ply a process by which all consonants aredoubled when followed by a front glide /j/. Itis constrained only when the head becomestrimoraic. The following Old English nounsillustrate gemination, where the head of thefoot is circumscribed by square brackets (cf.(30)).


The form *wıtte is impossible because thehead cannot be trimoraic. Similarly, *ATelleis disallowed because the weak branch of thehead is strengthened and again the head be-comes trimoraic. Not just strengthening pro-cesses, but deletions can also be sensitive tofoot structure. For instance, high vowels inOld English were deleted in the weak branchof a foot (Dresher & Lahiri 1991). In thefollowing examples the underlined vowels aredeleted (cf. (31)).

The [u] in lofu is not deleted because it iswithin the head. In contrast, in wordu andfAreldu, the [u] is in the weak branch of thefoot, and hence is deleted. Note that a tri-moraic head in fAreld is permitted becausethere was no choice to begin with. However,a process like gemination is prohibited fromcreating one as we saw in the case of *ATelle.The foot also accounts for stress: the head ofthe foot in each word bears main stress.

3.6. Minimal word and the footThe phonological word is the next constitu-ent above the foot. Just like the syllable andthe foot, most languages adhere to con-straints which try to maintain a ‘minimalword’. Most languages have a minimal wordrequirement which is closely related to a foot.The minimal word must be at least a foot,or two syllables, or bimoraic, or some otherprosodic constraint. Our interest here is pri-marily on the correlation between minimalword requirements and metrical coherence.We will, therefore, briefly illustrate the roleof the minimal word in prosodic phonologyand morphology.

(31) Foot based syncope in OE(X ) (X ) X ) (X )(x .) (x .) (x .) (x .)([mm] m) m ([mm] m) ([m mm] m) ([mm] m) mH L L H L L H L H L Lhea fuß de wor duß fæ rel duß clı we nuheafde word færeld clıwenu‘head’ ‘word’ ‘journey’ ‘ball of thread,

clew’dat. sg. nom.pl. nom.pl. nom. pl.

(X ) (X ) (X ) (X )(x .) (x) (x .) (x ) (x )([mm] m) [mm] ([m m] m) ([m m]) ([m mm])H L H L L L L L L Hhea fuß des we ru duß lo fu su numheafdes werud lofu sunum‘head’ ‘troop’ ‘praise’ ‘son’gen.sg. nom.pl. nom.pl. dat.pl.

Vowels are often lengthened to meet aminimal word requirement. In Bengali, forinstance, a vowel in a monosyllabic word isalways lengthened unless the vowel nucleushas a diphthong (Fitzpatrick-Cole 1994).

(32) Vowel lengthening in Bengalia. tsa: ‘want, ask for 2p.familiar

imperative / tea’b. tsa�i � tsai« ‘want, ask for �1p. present’c. tsa:i ‘teaonly’d. na:k ‘nose’e. nak�i ‘nose�adjectival suffix /

nasal’f. na:ki ‘noseonly’

The morpheme /tsa/ can be both a verb root‘to want’ or the noun ‘tea’. The ‘�’ boundaryindicates a suffix while the ‘’ sign marks aclitic. The different suffixation and cliticisedforms show the vowel length alternation. Itshould be noted that Bengali does not havecontrastive vowel length. At first glance thelengthening of the vowel in (32 c, f) seemsto be a counterexample to the minimal wordrequirement. In fact, (32e) shows that a de-rived word which is disyllabic does notlengthen a vowel similar to the monosyllabicword in (32b). However, the final vowel in(32f) is not a suffix but a clitic, just as in(32c). Clitics are added to a word and hence,the minimal word requirement must be metbefore the clitic is added. We see a differencein (32b) and (32c) where the former is a suf-fixed word and the resulting diphthong satis-fies the minimal word requirement. In (32c),however, the final vowel is a clitic and again,the initial vowel is lengthened.


These facts are not unusual. Many lan-guages, including Germanic languages likeEnglish, Dutch and German, also have min-imal word requirements. No content wordof these languages can end with a single laxvowel: *[si] or *[bi] would be completelyimpossible words in these languages. Theyshould either have a long vowel as in sea/see[si:], or should be closed as in sit [sıt]. Theresearch of McCarthy & Prince (1990) de-monstrates that the minimal word plays animportant role in the prosodic morphologyof languages. In their discussion of this phe-nomenon in Arabic, they give examples of asmall number of nouns (usually related tobody parts and kinship terms) which disobeythe bimoraic, minimal word requirement (fi-nal consonants are extrametrical and hencedo not count for weight): [?ab] ‘father’, (?ax]‘brother’ etc. However, when these nounsserve as the basis of regular word formationprocesses, they acquire an extra consonant,thus fulfilling the minimality requirement:[?ab] ‘father’, but [?abaw-iy] ‘paternal’; cf.[masør] ‘Egypt’, [masør-iy] ‘Egyptian’.

Minimal word requirements are also oftenreflected in blocking the application of rulesthat may shorten a word beyond the mini-mum. For instance, Lardil has a disyllabicword minimum. Apocope applies freely totrisyllabic or longer stems, but it is blockedin disyllables since it would shorten a wordbeyond the acceptable minimal word require-ment (Kenstowicz 1994). In the following ex-amples we see that the final stem vowel is al-ways deleted in the uninflected form, exceptin the last two words which are disyllabic.

(33) Lardil apocopeuninflected inflected glossyalul yalulu-n ‘flame’mayar mayara-n ‘rainbow’karikar karikari-n ‘butterfish’mela mela-n ‘sea’witøe witøe-n ‘inferior’

Minimality constraints can also add a moraor a syllable when the base has less than theweight required to satisfy the minimum wordrequirements. Such a process is also evidentin Lardil (Kenstowicz 1994).

(34) Addition of a mora in Lardiluninflected inflected glosskentapal kentapal-in ‘dugong’yaraman yaraman-in ‘horse’yaka yak-in ‘fish’tøera tøer-in ‘thigh’

In the last two examples, the base forms arenot disyllabic since the suffixation shows thatthey are consonant final stems (cf. the exam-ples above). However, to meet the minimalword requirement, a final vowel is added tothe base stem or the uninflected form to en-sure that it surfaces as disyllabic.

3.7. The foot and typological premisesHow do the above analyses fit into the usualtypological premises made when referring tostress? We have argued that stress is not afeature on a vowel, but rather is the linguisticmanifestation of rhythmic structure. As such,although one could state that a given syllablein a word bears the main stress, this is notthe best way to account for stress rules. Analternative and better way is to construestress placement as the parsing of a word intometrical feet. This does not preclude the pos-sibility that there are languages with fixedmain stress either on the initial syllable or onthe final syllable. Out of 300 languages, Hy-man (1977) noted that 114 languages haveinitial stress, 97 final stress, 77 penultimatestress and only 12 have stress on the secondsyllable. Such a statement, however, saysnothing about how secondary stress couldwork. Under a metrical foot analysis, a lan-guage with final syllable stress could easilyhave either a moraic trochee or an iamb, ifall final syllables happened to be heavy andparsing was from right to left. Predictions forsecondary stress, however, would be dif-ferent. Consider the following hypotheticalexample.

(35) Final syllable stress: iamb or trochee?Parsing Right to Left, Moraic tro-chee, End Rule Right( X)

(x) (x .) (x) (x)s s s s s s sParsing Right to Left, Iamb, EndRule Right( X)(. x) (. x)(. x)s s s s s s s

The final syllable obtains stress in both in-stances. But parsing into metrical feet pre-dicts that the third syllable from the begin-ning could bear secondary stress only if thefoot is a moraic trochee, and not if it is aniamb. Thus, broad typological statementssuch as main stress is final, can be mislead-ing. This does not mean that stress cannot befixed regardless of the type of syllable. Usu-


ally this happens if stress, or rather the metri-cal parameters, are in some way morpholog-ically governed. For instance, with Englishproductive affixation like beautiful, stress isinsensitive to rhythmic patterns. Stress fallson the antepenultimate syllable because ithappens to be the initial syllable of the stem.Another way of looking at it is that in de-rived adjectives the final syllable is extra-metrical. Morphological effects of stress canbe also found when certain suffixes alwaysbear stress as in German -ier, (e. g. Juwelıer),or English -ee, (e. g. devotee). However,even with morphologically stress, the windowwithin which stress falls is usually constrained.Hence, when typological correlations aredrawn with respect to stress, it is worthwhileto be more precise about the rhythmic organ-isation and the type of foot. We will discussthis in more detail in § 4.

In this section, we have covered a widerange of facts involving metrical structures.We have briefly discussed various phonologi-cal processes sensitive to metrical structure,including shortening and lengthening pro-cesses, segmental alternations, repair strate-gies for preferred syllables, etc. The centralgoal was to show that along with stress as-signment, phonological rules do not onlyoperate in local segmental contexts, but thathierarchical structures like syllables and feetalso constrain representations and processes.

4. Typology and Metrical Structures

Having established the necessity for metricalconstituents like syllables and feet, we arenow in a position to address issues of typo-logical implications. In the preceding sec-tions, we outlined different types of syllablestructures and feet which languages appearto have. We have not discussed, however, anyparticular correlations between such struc-tures nor any possible relationship betweenthe existence of different types of metricalstructures within a given language. For in-stance, is it possible that all types of feet co-exist in a single language? Would it be pos-sible to infer the preferred syllable structureof a language if its type of foot is known?Are there any implications to be drawn frompreferred metrical structures of language andthe types of rules that are permitted? Suchquestions are rarely if at all addressed withinphonology. We will draw attention to threepossible typologically interesting issues: (a)

the relevance of metrical constituents outsidephonology, (b) the coherence of metrical unitswithin a given language, and (c) correlationsdrawn between metrical constituents and otherphonological and morphological patterns.

4.1. Metrical constituents outsidephonology

Research in prosodic morphology (cf. Mc-Carthy & Prince 1986, 1993) indicates thatmetrical categories required in phonology arethe same that are necessary for morphol-ogical processes like reduplication. Typologi-cally, therefore, the implication is that if anew category is found to be necessary to de-scribe either stress or any other phonologicalprocess, it ought to be found relevant for amorphological process as well. The followingtwo examples support the view that both syl-lables and feet are relevant for morphology.

In Mokilese, the progressive is expressedby a form of reduplication (Harrison & Al-bert 1976; McCarthy & Prince 1986). Sam-ples of the data are given below.

(36) Mokilese progressivea. pc.dok pcd-pc.dok ‘plant’b. pa paa-pa ‘weave’

di.ar dii-di.ar ‘find’c. scc.rck scc-scc.rck ‘tear’

caak caa-caak ‘bend’d. an.dip an.d-an.dip ‘spit’

o.nop on.n-o.nop ‘prepare’

At first glance, it seems as if the progressiveis formed by some sort of a prefix which isequivalent to a syllable. However, the natureof the syllable differs for each word type. Thereduplication is essentially the prefixation ofa bimoraic syllable [smm]. In (36a) the initialsyllable of the stem is monomoraic and hencethe following onset is included to form theprefix. In (36b), where the stem is monosyl-labic or where the second syllable has noconsonantal onset, the vowel is lengthened tosatisfy the bimoraic requirement. In (36c) thebimoraic requirement is met by simply takingthe initial long vowel. The most interestingcase is (36d) where [an] or [on] would surelyhave met the bimoraic requirement. How-ever, the output form would then have been*[on-onop], which would then be syllabifiedas *[o.no.nop] and the initial prefix would bemonomoraic. Thus, the constraint is that themorphological prefix must be an entire bi-moraic syllable, and simultaneously the onsetof the next syllable must be maximised. Forthe latter, if the stem begins with a vowel,


either the onset of the second syllable is usedeven if it has to be doubled.

In the next example we see that a morpho-logical process can be sensitive to a foot. Thedata is from Ulwa (McCarthy & Prince1990).

(37) Ulwa construct state (3 sg. possessed)a. kii (kii)-ka ‘stone’

bas (bas)-ka ‘hair’sana (sana)-ka ‘deer’sapaa (sapaa)-ka ‘forehead’amak (amak)-ka ‘bee’

b. suulu (suu)-ka-lu ‘dog’baskarna (bas)-ka-karna ‘comb’siwanak (siwa)-ka-nak ’root’anaalaaka (anaa)-ka-laaka ‘chin’karasmak (karas)-ka-mak ‘knee’

The data in (a) is straightforward � a suffix[ka] is added to the stem. However, the datain (b) shows that the [ka] behaves like an in-fix and the way in which the stem is dividedup appears to be different in each case. In thefirst two examples, the [ka] is added to thefirst syllable, while in the others it is addedafter two syllables. If we compare the twosets, we can see that [ka] is added after twosyllables only when the first syllable is lightand the second is heavy or light � [a.naa] �but not, when the first syllable is heavy andthe second is light as in [suu.lu]. Thus, it isnot a syllable which is the relevant constitu-ent but a foot and a typical iamb [smsmm]. Themorphological process is thus the following:add the suffix [ka] to the leftmost iambicfoot.

Thus, the metrical constituents includingsyllables, feet and minimal word that are usedfor the description of stress and are sensitiveto other phonological processes are argued tobe the same for morphological processes. Theobvious question that now comes to mind iswhether within a given language, the samemetrical constituent is used both for stressand other phonological processes, and in ad-dition whether morphology and phonologyshare precisely the same type of constituentas well. We now turn to these issues.

4.2. Metrical coherenceTypically languages adhere to the same footfor stress as well as for other phonologicalprocesses. A language following such a prin-ciple would be judged as being metricallycoherent and arguably such a language wouldalso be easier to learn (cf. Dresher & Lahiri1991). One could also extend the principle of

metrical coherence to morphology as well. AsHayes (1995) points out, usually in any givenlanguage, the kind of foot used for stress isthe same as that used in morphology. We willaddress this issue first from a purely phono-logical perspective and then briefly addressthe notion of coherence within prosodic pho-nology and morphology.

4.2.1. Coherence in stress andphonological processes

There is nothing discussed in the earlier sec-tions that suggests that metrical coherence isa must or even preferred. One could imaginethat metrical structures are counting devicessuch that one type of foot is used for stressand another could be used as the context forother phonological rules. For instance, Key-ser and O’Neill (1976) suggested that in OldEnglish, stress required an initial quantityinsensitive left-headed foot while a quantitysensitive right-headed iamb was necessary toaccount for the rule of high vowel deletion,where R rhyme

(38) Feet in Old English following Keyserand O’Neill

a. Word stress: left-headed unboundedfoot

b. High vowel deletion: following aright-headed quantity sensitive footin an open syllable

Under such an analysis, headedness is com-pletely arbitrary. For the purposes of stress,the left head of a foot is the strongest, whilefor syncope, the right head of the foot isstrong. However, such a system would beincredibly difficult for the language learner toacquire. As shown in § 3.5., this analysis isnot the only one that could account for thedata. Germanic is essentially metrically co-herent, and both stress and deletion of highvowels can be accounted for by a single foottype. Other languages have similar proper-ties. One example is Unami, an Eastern Al-


gonquian language described in Hayes (1995:211�213) based on Goddard (1979). Hayesargues that stress in Unami is accounted forby forming iambs left to right with foot ex-trametricality and End Rule Right. More-over, voiceless consonants other than [h] aregeminated after a strong vowel which wouldbe the head of a foot.

(39) Unami stress and gemination( X)(. x) *(. x)+ne me t eme 6 ‘I follow a trail’J( X)(. x) *(. x)+ne me t t eme 6Compare/meteme6w/ J [meteme6(w)]‘he follows a trail’

In general, phonological rules sensitive tofoot structure are usually deletion, weaken-ing or strengthening processes as we haveseen in earlier sections. In most instances,they apply to either repair metrical structuresor are invoked to build preferred structures.Therefore it is not surprising that the foot re-quired for stress and other phonological pro-cesses would be the same. What is more inter-esting is the notion of metrical coherence out-side phonology to which we turn next.

4.2.2. Metrical coherence and morphologyAs we mentioned earlier, the set of metricalconstituents that are relevant for phonologyare the same that are used in prosodic mor-phology. For instance, often, suffix allomor-phy is governed by a metrical unit. In Dutch,the nominalising suffix {-aar} adds only to bi-syllabic verbs, while {-er} is added to mono-syllabic verbs: luister ‘to listen’ � luister-aar‘listener’; bel ‘to ring’ � bell-er ‘ringer’. Booij(1997) argues that the underlying reason forthis allomorphy is to maintain a propertrochee on the surface. The form *luistererwould not form a proper trochee whileluisteraar ends up as two acceptable feet.

Our interest here is whether the type offoot necessary for a language’s morphologicalsystem is the same as that required for stress.That is, can we extend the notion of metricalcoherence to morphology as well. It appearsthat almost always the foot used for stress isthe same as that used for prosodic morphol-ogy. One example is Manam (Austronesianlanguage, spoken in New Guinea). The de-scription of stress is given in Kenstowicz(1994: 614, 659) and the morphological pro-

cess of reduplication is described in McCar-thy & Prince (1986: 39).

(40) Manam Stress and Reduplicationa. mo.tu ‘island’

ma.nam ‘Manam island’wa.rı.ge ‘rope’ma.la.bon ‘flying fox’?i-po.a.sa.ge.na ‘we are tired’

b. la?o la?o- la?o ‘go’moıta mo-ita-ıta ‘knife’malabon mala-bom-bon ‘flying fox’

Stress is assigned by a moraic trochee parsedfrom right to left. The reduplication alsoclearly refers to the same foot. However, thereis nothing inherent about metrical structuresthat requires the same foot to be used forphonology and morphology. One glaringcounterexample in the literature is AxinincaCampa. While stress is iambic (Payne 1981;Spring 1990a; McCarthy & Prince 1993),there is disagreement in the literature aboutthe foot type(s) needed for morphology(Spring 1990a, 1990b; McCarthy & Prince1993). For instance, Spring (1990b) claimsthat while the foot necessary for genitive allo-morphy is the moraic trochee, the foot forverb reduplication is the iamb.

However, other analyses of AxinincaCampa suggest that the situation is not ascomplicated as suggested by Spring. Whilethe distinctive base of both genitive allomor-phy and reduplication is characterised by abimoraic foot, the reduplicant is a differentconstituent with a strong tendency towardsdisyllabicity. Metrical coherence may be main-tained for Axininca Campa if the bimoraicfoot is regarded as ‘minimal iamb’, while thedisyllabic foot is considered to be a ‘maximaliamb’. Only a small sample of the crucialdata is presented to show the analysis (cf.(41)).

The descriptive generalisation for the geni-tive allomorphy is that if the stem containsonly two moras, the suffix /ni/ is taken, other-wise the suffix is /ti/. This led Spring (1990b)to suggest that the genitive allomorphy wassensitive to a trochee. However, a bimoraicfoot could be both a trochee as well as aniamb, the difference being the headedness.Since there is no evidence of headedness inthis context, the analysis could just as wellbe that if the foot is a minimal iamb, thesuffix is /ni/. The verb reduplication is morestraightforwardly an iamb. If the non-pre-fixed stem is a single syllable, the reduplica-tion includes the prefix. That the reduplicant


(41) Axininca Campaa. Stress (following McCarthy & Prince 1993, Spring 1990a, 1990b): Iamb, Left to Right,

Final light syllable extrametrical.(ki.mı).ta.*ka+ ‘maybe’(o.cı).(to.mo).*ko+ ‘monkey’(i.raa).(wa.na).*ti+ ‘su caoba’

b. Genitive allomorphy (data Spring 1990a)/sima/ no-sima-ni ‘my fish’/mii/ no-mii-ni ‘my otter’/sawoo/ no-sawoo-ti ‘my cane’/maini/ no-maini-ti ‘my bear’/cokori/ no-cokori-ti ‘my armadillo’

c. Verb reduplication-prefixed pattern (Payne 1981)/kiNtha/ non - kintha - kintha ‘tell’/kawosi/ non - kawosi - kawosi ‘bathe’/naa/ no - naa - nonaa ‘chew’/na/ no - na - nona ‘carry‘/osaNpi/ n - osampi - sampi ‘ask’/apii/ n - apii - napii ‘repeat’

is a canonical iamb is seen in the last exam-ple. Normally, the stem initial vowel is notreduplicated as in [n-osampi-sampi]. How-ever, when the stem itself is disyllabic as in/apii/, the reduplication does not ignore theinitial vowel. The reduplicated form is [n-apii-napii] and not *[n-apii-pii] showing thatthere is strong preference to have a canonicaliamb if possible (Black 1991; Loewe 1996).

Thus, even in a complex set of interactionsas in Axininca Campa, there is no clear evi-dence that entirely different feet are requiredfor morphology and phonology. Since in mostinstances, morphology requires either a disyl-labic or bimoraic foot with no clear headed-ness preferences, or just a minimal word (cf.§ 3.6.), the stress facts do not clash with mor-phological processes (cf. McCarthy & Prince1990 for various examples). It seems, there-fore, that metrical coherence can be extendedto morphology as well.

5. Universals, implications andcorrelations

In the typological literature, little has beensaid concerning the details of metrical con-stituents and implications thereof. The focusof attention has been either to associatemorphological types like agglutination vs.flection with stress-timing or syllable-timing,or with general rhythmic patterns like iam-bic and trochaic with possible syllable typesand types of clusters. For example, Lehmann(1973, 1978) claims that agglutinating lan-guages correlate with simple syllable struc-

ture, pitch accent and mora-counting, whileflective languages have complex syllable struc-tures, stress accent with reduction of un-stressed syllables, and syllable-counting. Plank(1998) provides a detailed account of the var-ious attempts in the literature to draw corre-lations within phonological constituents aswell as between phonological and morpho-logical categories. The results are highly un-satisfactory, and as Plank points out, oftencontradictory. Our interest here is primarilyon metrical constituents and we will discussbriefly some of the proposals put forward inthe literature.

Donegan & Stampe (1983) argue thatrhythmic properties determine different mor-phological types and syntactic word order.The central notion is that since accent is theonly factor which is pervasive through alllevels of language, it is the only meaningfuldeterminant in connecting the different levelsof language. For instance, there is a direct re-lationship between word order and phrasalaccent. In a sentence, the operand or headis given information, while the operator ormodifier is asserted and hence bears the mainaccent, regardless of the relative order of thetwo parts. They make a very strong claimthat rising (final) vs. falling (initial) phraseaccent is the primary variable and that oper-and/operator order follows from it. In fact,primary phrase accent determines more thanjust word order: syllable and word canons,phonological segments, as well as timing areclosely related. For instance, initial phrase ac-cent correlates with trochaic word accent,


syllable-timing or mora-timing, a preferencefor (C)V(C) syllables and geminate clusters.Final phrase accent and iambic word accentgoes hand in hand with stress-timed lan-guages, with (C)V or (C)(C)V(Glide)(C) syl-lables and non-geminate clusters. Thetrochaic pattern goes with agglutinative mor-phology while the iambic pattern goes withmore flective morphology.

Gil’s approach to prosodic typology isalso based on rhythm (Gil 1986), but it seemsto make somewhat opposite predictions. Al-though Gil agrees with Donegan & Stampethat trochaic rhythm patterns with syllabletimed languages and iambic rhythm withstress timed languages, he supports the viewthat agglutinative languages prefer iambicrhythm, are stress timed, have a high conso-nant-vowel ratio, and have a simple syllablestructure. In contrast, flective languages pre-fer trochaic rhythm, are syllable-timed, havecomplex syllable structure, and have a lowconsonant-vowel ratio. Thus, the correlationbetween agglutinative/flective morphologywith trochaic/iambic rhythm is the opposite.Gil (1987) however, lays less emphasis on thetypological prominence of phrasal rhythm,since, he states, most languages tend to haveiambic patterns on the phrase and clauselevel. What he had in mind is perhaps thatan intonation phrase is invariably divided upinto head+nucleus where the nucleus con-tains the most important information (cf.Hayes & Lahiri 1991). However, it is not thecase that within the nucleus, languages alwaysprefer iambic patterns. Banking on the differ-ence between iambic and trochaic rhythmwithin phonological phrases, Nespor, Guasti& Christophe (1996) argue that languagelearners correlate this difference with thebranching nature of syntax. Thus, a trochaicrhythm correlates with left-branching struc-tures and an iambic rhythm with right-branching structures. The syntactic branch-ingness correlates with heaviness. Crucially,the authors make no attempt to link wordstress with phrasal stress, the former being anindependent variable.

In sum, in the attempts to draw corre-lations between rhythm (which is of interestto us since it is related to metrical structures)and other phonological, morphological, andsyntactic structures, there are two majorapproaches. First, a correlation is establishedbetween phrasal rhythm and syntactic struc-ture (Donegan & Stampe 1983, Nespor, Gu-asti & Christophe 1996), but Gil (1987) views

phrasal rhythm as not being a dependablevariable to support typology. The differencebetween Donegan & Stampe and Nespor etal. is that the former connect phrasal rhythmwith word rhythm, while for the latter wordstress is an independent variable.

Second, a correlation is made between wordrhythm and other phonological constituentslike syllable structure, stress/syllable timing,and morphological structure. Here, as wementioned earlier, Donegan & Stampe, andGil sometimes make opposite predictions.For instance, Donegan & Stampe associateiambic rhythm with complex syllable struc-ture while Gil has it the other way around.Other authors who have attempted thesecorrelations do not always agree either (seePlank 1998: 216 for details). What then dothese typological correlations mean? We willconcentrate on the correlations drawn on thebasis of word rhythm since this has been theprimary focus in this paper.

The authors (Donegan & Stampe, and Gil)are basing their hypotheses and analyses onsamples of data from which they have ob-served certain patterns. But the conclusionsdo not appear to be based on detailed prop-erties of metrical constituents as we have dis-cussed so far. It is true that if a language pre-fers an iambic foot, it is almost certain thatit would have long vowels since an iamb isby nature asymmetric in quantity. However,there is nothing to prevent a language frompreferring an iambic foot where the quantityis determined solely on the basis of closed vs.open syllables. But more important, prefer-ring an iambic word rhythm, does not neces-sarily mean that the language prefers stressto be at the right edge of a word, which iswhat the correlations seem to imply. Whetherthe right or the left edge of a word bearsmain stress depends on the End Rule and notonly on the foot type. Consider once again ahypothetical example.

(42) Iambs: End Rule Right/LeftLanguage AParsing Right to Left, Iamb, EndRule Right( X)(. x) (. x) (. x)s s s s s s sLanguage BParsing Right to Left, Iamb, EndRule Left( X )(. x) (. x) (. x)s s s s s s s


If each parsing represented a language, thenfor the given sequences of light and heavysyllables, in Language A the main stresswould fall on the word final syllable, while inLanguage B, main stress would fall on thesecond syllable of the word. Although theseare hypothetical examples, they are by nomeans exceptional. Consider the examplesgiven in (28). Both Cairene Arabic and War-gamay have a moraic trochee, but the direc-tion of parsing as well as the End Rule aredifferent. As a result, although two and threesyllable words look very similar for stress as-signment, four syllable words are different.We repeat the crucial examples here.

(43) Moraic trochees in Cairene and War-gamayCairene Arabic: Parsing Left-to-Right, End Rule Right(X ) ( X )(x) (x .)(x .)mm m m m m mbe:ta*k+ katabıtuyour (m.sg.) house she wrote it (m.)Wargamay: Parsing Right to Left,End Rule Left(X ) (X ) ( X )(x) (x .)(x .) (x .)mm m m m m m m mmmu:ba gıJ-awulu gagarastone fish freshwater dilly bag

jewfish

Cairene tends to have main stress towardsthe end of a word (katabıtu), while Wargamayhas stress at the beginning of the word (gıJa-wulu).

Thus, before we attempt to establish cor-relations between word rhythm and otherphonological constituents, we have to firstestablish whether we are dealing with iam-bic/trochaic feet or whether we are referringto merely word edges. Since we are dealingwith foot type, direction of parsing, End Rule(plus extrametricality) as parameters for as-signing stress, any correlation concerningword rhythm could refer to all of them collec-tively or any one of them individually.

What types of correlations and implica-tions can we then draw given what we knowabout metrical patterns? One sort of correla-tion could be connected with metrical coher-ence. If languages prefer to stick to a giventype of foot for stress, phonological processesas well as morphology, one might think thatvarious types of rules would conspire toachieve a preferred foot. Thus, if the lan-

guage requires iambs for stress, then vowellengthening or gemination rules would applyto convert a minimal iamb [smsm] to a canon-ical one [smsmm]. Such iambic lengtheningexamples are frequently mentioned in Hayes(1995). Similarly, as we saw in § 2., epenthesisor deletion processes are often invoked to ob-tain preferred syllable structures. One couldconstruct a set of strategies that correlatewith preferred metrical structures as in (44).

(44) Correlations of preferred metricalstructures and repair strategiesPreferred Strategiesstructures

(a) Languages prefer coda deletion,open syllables vowel epenthesis

(b) Languages prefer apocope or syn-closed syllables cope; no coda

deletion, vowelepenthesis

(c) Iamb Iambic length-ening

(d) Moraic trochee Trochaic short-ening

(e) Minimal Word Vowel lengthen-ing, epenthesis

Unfortunately, a constant problem in pho-nology is that surface output forms can easilybe opaque with respect to the metrical struc-tures of the language. Thus, just as iambiclengthening is frequent, deletion of un-stressed syllables are equally frequent, reduc-ing a disyllabic iamb [smsmm] to a monosyl-labic one [smm]. This type of opacity has beenaddressed repeatedly throughout the historyof generative phonology.

In recent research, opacity is elegantlycaptured in terms of constraint interactionwithin the framework of Optimality Theory(OT, cf. Prince & Smolensky 1993; McCar-thy & Prince 1995; for a recent survey, seeKager 1999). In OT, the explanatory burdenis shifted from processes to output candi-dates. The central claim is that universalgrammar is made up of a set of constraints,all of which are available to a given language.The grammar generates a potentially infiniteset of output candidates for each input,which are then evaluated based on the con-straint system of the language. The candidatewhich best fits the constraint system is thevictor. Languages differ in the way the con-straints are ranked. For our purposes, sinceconstraints are violable and can easily conflict,one of the most interesting consequences isthat an output form need not conform to all


the constraints of the language. Let us con-sider some of the preferences stated in (44)and examine how they can conflict. A lan-guage could have the following preferencestranslated into constraints:

(45) Possible ranked constraintsa. Last syllable cannot be stressed.b. Foot is not right-headed; i. e. it is

iambic.c. Main stressed foot aligns with the

right edge of a word.

The constraints (a) and (c) are in direct con-flict. Constraint (c) says that the head of thefoot should coincide with the right edge ofthe word. Since the foot is an iamb, the lastsyllable stress seems to be preferred. How-ever, there is a direct conflict with (a) whichsays right syllables must not be stressed. In(46) we provide two possible candidates giventhe above ranking.

(46) Possible candidatesi. (. x) (. x) (. x)

s s s s s sii. (. x) (. x)

� s s s s s s

Since (43a) is ranked above (43c), the lastconstraint will lose out and (46 ii) would bethe preferred candidate. The output is thenopaque to the fact that the head of the footshould preferably line up with the right edgeof the word.

What would this mean for the possibletypological correlations suggested in (44)?Given that preferences and constraints canbe conflicting which predicts that the outputmay be opaque, we can only state the corre-lations we have as preferences and not abso-lute. Obviously, the preferences remain asopen questions.

6. Conclusion

In this article, we have established that thereare systematic organising principles whichconstruct metrical patterns in natural lan-guages. These organising principles combinevarious units of prosodic structure like syl-lables and feet. There is a limited inventoryof syllable and foot types for all languages.Stress, which marks relative prominencewithin metrical constituents, is a result ofvarious independent parameters includingthe type of foot and the edge of a word anddirection of parsing.

Metrical constituents are however, notonly relevant for stress but for other phono-logical and morphological phenomena. Wehave argued that typologically, languagesadhere to metrical coherence within both thephonological and morphological systems ofa given language. However, when it comesto drawing correlations between metricalstructure and other aspects of phonology andmorphology, the typological literature issomewhat uncertain in drawing any conclu-sions. Part of the reason is that only fixedtemplatic structures have been taken into ac-count and metrical patterns are not viewed asorganising principles of grammatical systems.However, since we do have a well understoodset of metrical patterns and are aware ofpossible processes and constraints, future re-search will undoubtedly try to lay out mean-ingful typological implications in this area.

7. References

Black, Andrew. 1991. “The optimal iambic footand reduplication in Axininca Campa”. In: Phonol-ogy at Santa Cruz 2: 1�18.

Blevins, Juliette. 1995. “The Syllable in Phonologi-cal Theory”. In: Goldsmith, John (ed.), 206�244.

Booij, Geert. 1979. “Allomorphy and autonomy ofmorphology”. Folia Linguistica 31: 25�56.

Borowsky, Toni & Ito, Junko & Mester, Armin.1984. “The formal representation of ambisyllabi-city: evidence from Danish”. North Eastern Lin-guistic Society 14: 34�48.

Chomsky, Noam & Halle, Morris. 1968. The soundpattern of English. New York: Harper and Row.

Clements, G. Nick & Hume, Elizabeth V. 1995.“The internal organisation of speech sounds”. In:Goldsmith, John (ed.), 245�306.

Donegan, Patricia J. & Stampe, David. 1983.“Rhythm and the holistic organisation of languagestructure”. In: Richardson, John F. & Marks,Mitchell & Chuckerman, Amy (eds.), Papers fromthe Parasession on the Interplay of Phonology, Mor-phology and Syntax, 337�353. Chicago: ChicagoLinguistic Society.

Dresher, B. Elan & Lahiri, Aditi. 1991. “The Ger-manic Foot: Metrical coherence in Old English”.Linguistic Inquiry 22: 251�286.

Eisenberg, Peter & Ramers, Karl Heinz & Vater,Heinz (eds.). 1992. “Silbenphonologie des Deut-schen”. Studien zur deutschen Grammatik 42. Tü-bingen: Gunter Narr Verlag.

Fitzpatrick-Cole, Jennifer. 1994. The prosodic do-main hierarchy in reduplication. PhD dissertation,Stanford University.


Fitzpatrick-Cole, Jennifer. 1996. “Reduplicationmeets the phonological phrase in Bengali”. TheLinguistic Review 13: 305�356.

Ghini, Mirco. 2001. Asymmetries in the phonologyof Miogliola. (Studies in Generative Grammar.)Berlin: Mouton de Gruyter.

Giegerich, Hans. 1992. “Onset maximisation inGerman: the case against resyllabification rules”.In: Eisenberg, Peter & Ramers, Karl Heinz &Vater, Heinz (eds.), Silbenphonologie des Deut-schen, 134�171.

Gil, David. 1986. “A prosodic typology of lan-guage”. Folia Linguistica 20: 165�231.

Gil, David. 1987. “On the scope of grammatic the-ory”. In: Modgil, Sohan & Modgil, Cecilia (eds.),Noam Chomsky: Consensus and Controversy, 119�141. Barcombe: Falmer Press.

Goddard, Ives. 1979. Delaware Verbal Morphology.New York: Garland publishing.

Goldsmith, John (ed.). 1995. The Handbook ofPhonological Theory. Oxford: Blackwell.

Greenberg, Joseph H. & Kashube, Dorothea. 1976.“Word prosodic systems: A preliminary report”.Working Papers on Language Universals 20: 1�18.

Gussenhoven, Carlos. 1986. “English plosive allo-phones and ambisyllabicity”. Gramma. 10.2: 119�141.

Haas, Mary. 1977. “Tonal accent in Creek”. In:Hyman, Larry (ed.) Studies in stress and accent.Southern California Occasional papers in Linguis-tics 4, 195�208. Los Angeles: University of South-ern California, Department of Linguistics.

Halle, Morris & Idsardi, William. 1995. “Generalproperties of stress and metrical structure”. InGoldsmith, John (ed.), 403�443.

Harrison, Sheldon P. & Albert, Salich Y. 1976.Mokilese Reference Grammar. Honolulu: Univer-sity of Hawaii Press.

Hayes, Bruce. 1981. A metrical theory of stressrules. Cambridge, Mass.: MIT PhD dissertation.Distributed by Indiana University Linguistics Cluband published by Garland Press in 1985.

Hayes, Bruce. 1989. “Compensatory lengthening inmoraic phonology”. Linguistic Inquiry, 20: 253�306.

Hayes, Bruce. 1995. Metrical Stress Theory: Prin-ciples and Case Studies. Chicago: University ofChicago Press.

Hayes, Bruce & Lahiri, Aditi. 1991. “Bengali into-national phonology”. Natural Language and Lin-guistic Theory 9: 47�96.

Hulst, van der, Harry. 1984. Syllable structure andstress in Dutch. Dordrecht: Foris.

Hulst, van der, Harry. 1985. “Ambisyllabicity inDutch”. In: Bennis, Hans & Beukma, Frits (eds.)Linguistics in the Netherlands. Dordrecht: Foris.

Hyman, Larry M. 1997. “On the nature of linguis-tic stress”. In: Hyman, Larry M. (ed.) Studies inStress and Accent. Southern California OccasionalPapers in Linguistics. 4: 37�82.

Jacobs, Haike. 1989. Historical studies in the non-linear phonology of French. PhD dissertation, Uni-versity of Nijmegen.

Jacobs, Haike. 2000. “The revenge of the uneventrochee: Latin main stress, metrical constituency,stress-related phenomena and OT”. In Lahiri,Aditi (ed.). 333�352.

Kager, Rene. 1989. A metrical theory of stress anddestressing in English and Dutch. Linguistic models14, Dordrecht: Foris.

Kager, Rene. 1995. “The metrical theory of wordstress”. In: Goldsmith, John (ed.), p. 367�402.

Kager, Rene. 1999. Optimality Theory. Cambridge:Cambridge University Press.

Kahn, David. 1976. Syllable-based generalizationsin English phonology. Cambridge, Mass.: MIT PhDdissertation. Distributed by Indiana UniversityLinguistics Club.

Kenstowicz, Michael. 1994. Phonology and genera-tive grammar. Oxford: Blackwell.

Keyser, Samual J. & O’Neill, Wayne. 1976. Rulegeneralization and optionality in language change.Dordrecht: Foris.

Kiparsky, Paul. 1979. “Metrical structure is cyclic”.Linguistic Inquiry 10: 421�441.

Lahiri, Aditi (ed.). 2000. Analogy, Levelling, Mark-edness: Principles of change in phonology and mor-phology. Berlin: Mouton de Gruyter.

Lahiri, Aditi. 2000. “Hierarchical restructuring inthe creation of verbal morphology in Bengali andGermanic: evidence from phonology”. In: Lahiri,Aditi (ed.), 71�123.

Lahiri, Aditi & Dresher, B. Elan. 1999. “Open syl-lable lengthening in West Germanic”. Language75: 678�719.

Lahiri, Aditi & Koreman, Jacques. 1988. “Syllableweight and quantity in Dutch”. West Coast Confer-ence on Formal Linguistics 7: 217�228.

Lahiri, Aditi & Riad, Tomas & Jacobs, Haike.1999. “Diachrony”. In: van der Hulst, Harry (ed.)Word Prosodic Systems in the Languages of Europe,335�422. Berlin: Mouton de Gruyter.

Lehmann, Winfred. 1973. “A structural principleof language and its implications”. Language 49:47�66.

Lehmann, Winfred P. 1978. “Conclusion: Towardsan understanding of the profound unity underlyinglanguages”. In: Lehmann, Winfred P. (ed.) Syntac-tic Typology: Studies in the Phenomenology of Lan-guage, 395�432. Austin: University of Texas Press.

Liberman, Mark. 1975. The intonational system ofEnglish. Cambridge, Mass.: MIT PhD dissertation.Distributed by Indiana University LinguisticsClub.

136797. Tone systems

Loewe, Martha. 1996. Prosodic structures. MA the-sis, University of Konstanz.McCarthy, John & Prince, Alan. 1986. “ProsodicMorphology”. Manuscript, University of Mas-sachusetts, Amherst and Brandeis University.McCarthy, John & Prince, Alan. 1990. “Foot andWord in Prosodic Morphology: the Arabic brokenplurals”. Natural Language and Linguistic Theory8: 209�282.McCarthy, John J. & Prince, Alan. 1993. “General-ized alignment”. Yearbook of Morphology, 79�153.McCarthy, John J. & Prince, Alan. 1995. “ProsodicMorphology”. In: Goldsmith, John (ed.), 318�366.Nespor, Marina & Guasti, Therese & Christophe,Anne. 1996. “Selecting word order: the rhythmicactivation principle”. In: Kleinhenz, Ursula (ed.)Interfaces in Phonology. Studia Grammatica 41,1�26. Berlin: Akademie-Verlag.Payne, David. 1981. The phonology and morphol-ogy of Axininca Campa. Publication of the SummerInstitute of Linguistics and the University of Texasat Arlington.Plank, Frans. 1997. “Linguistic Typology: intro-ducing a new journal”. Linguistic Typology, 1, 1�3.

97. Tone systems

1. Defining “tone”2. Typologizing tone systems by phonological

contrasts3. Typologizing tone systems by function4. Typologizing tone systems by tone rules5. Tone and/or accent6. Tone-marking conventions7. References

1. Defining “tone”

Within the phonological realm, few typologi-cal issues have generated as much discussion(and controversy) as the issue of tone. It isgenerally assumed that as many as half ofthe world’s languages are “tonal”. Whilemost language families in the world have oneor more tonal offsprings, including those inNorth and South America, Europe, and Oce-ania, languages with fully developed tone sys-tems are highly concentrated in SubsaharanAfrica, Southeast Asia, and Mexico. Beyondthese generalities, the typological study oftone systems has at times faltered on the verybasic question of what constitutes a “tone”and hence a “tone system”. Welmers’ (1959,

Plank, Frans. 1998. “The covariation of phonologywith morphology and syntax”. Linguistic Typology,2, 195�230.

Prince, Alan & Smolensky, Paul. 1993. “Optimalitytheory: Constraint interaction in generative gram-mar”. Manuscript, Rutgers University and Univer-sity of Colorado at Boulder.

Rubach, Jerzy. 1990. “Final devoicing and cyclicsyllabification in German”. Linguistic Inquiry 21:79�94.

Spencer, Andrew. 1996. Phonology. Oxford: Black-well.

Spring, Cari. 1990a. Implications of AxinincaCampa for Prosodic Morphology and Reduplication.PhD dissertation, University of Arizona, Tucson.

Spring, Cari. 1990b. “How many Feet per Lan-guage?” In: The Proceedings of the Ninth WestCoast Conference on Formal Linguistics, 493�508.

Vennemann, Theo. 1972. “On the theory of syllabicphonology”. Linguistische Berichte 18: 1�19.

Aditi Lahiri, University of Konstanz(Germany)

1973) definition is as good as most: “A tonelanguage is a language in which both pitchphonemes and segmental phonemes enterinto the composition of at least some mor-phemes.” Thus, tone is clearly indicated inthe case of such pairs as Pawaian [Oceanic]su ‘tooth’ and su ‘road’ and Mende [SierraLeone] pili ‘house’ and bili ‘trousers’ (see§ 6. for tone-marking conventions). While Pike(1948) had suggested that a tone languagehas “contrastive, but relative pitch on eachsyllable”, Welmers improves on this definitionby recognizing the existence of toneless mor-phemes, especially grammatical morphemeswhich take their tone from the surroundingcontext. If we reinterpret Welmers in modernterms to mean that the pitch phonemes mustbe presented in “the underlying representa-tions of at least some morphemes”, this willallow for the Mende toneless postpositions-hu ‘in’ and -ma ‘on’, which copy their tonefrom the preceding nominal, e. g. pili-hu,pili-ma; bili-hu, bili-ma. It is also possiblefor lexical morphemes to be underlyinglytoneless. Hyman (1981) has analyzed tone inSomali as being largely predictable on the

96. Metrical patterns · is [ra:t.le] rather than *[ra:.dle]. Maximisation of the onset is closely...

Documents

Transcript of 96. Metrical patterns · is [ra:t.le] rather than *[ra:.dle]. Maximisation of the onset is closely...