To appear in GLOT International
Reduplication as Alliteration and Rhyme
University of California, Irvine
University College London
The prevailing view of reduplication (Marantz 1982, McCarthy and Prince 1986, 1995) analyses it as an abstract affix whose segmental content is copied from the base in accordance with certain principles or constraints. Among these are stipulative constraints (such as Anchor in OT), that ensure that prefixal reduplicants take their segmental material from the beginning of the word, and suffixal reduplicants take their material from the end of the word. Cross-linguistically this is a robust generalization: we find data such as these simplified samples: (Data from McCarthy and Prince 1986, 1993)
Tagalog: l kad (pag) la-l kad ‘walking’
takboh (um) t -takboh ‘will run’
Diyari: ku ku a ku ku-ku ku a ‘to jump’
tjilparku tjilpa-tjilparku ‘bird sp.’
Kamaiura: o-huka o-huka-huka ‘he kept laughing’
je-umirik je-umiri-miri-k ‘I tie up repeatedly’
Chamorro: dankolo dankolo-lo ‘really big’
bunita bunita-ta ‘very pretty’
The standard analysis misses one obvious generalization: the reduplicant and the copy are always adjacent, resulting in rhyming sequences. I suggest that the real core of reduplication is an attempt to produce sequences that rhyme and alliterate, and that rather than involving an abstract affix, reduplication is caused
: Output must contain at least one pair of adjacent syllables with identical rhymes
Input segments may have two output correspondents (violating Integrity
); there is no base or affix in the familiar sense. For an input /pati/ there will be two outputs which fully satisfy this: [pa-pati] and [pati-ti], analogous to Tagalog and Chamorro respectively.Note crucially that this does not
mean that reduplication necessarily copies syllables, something known to be untrue since Moravcsik 1978, Marantz 1982. Firstly, markedness considerations such as *ComplexOnset
can result in “undercopying”, as in Tagalog [ta-trabaho] or Ponapean [ke-kens] ‘ulcerate’. Secondly, if the rhyming portion precedes the alliterating portion, we get VC copy as in Tzeltal
, Oyakangand, or Mangarayi (see below). Below I elaborate on this and other consequences of this approach.
It is well known that some languages show VC copy, and this has posed a problem for prosodic morphology, since VC is not a prosodic unit. I show that my proposal predicts systems of this type. Typical data are given below:
an ‘feel with palm’
ol ololan ‘make rows’
algal alg-algal ‘straight’
igu ig- igu ‘go’
Just like CV reduplication, these outputs satisfy Alliterate
, but in the reverse order. In the examples like Tagalog la-l kad
in (1-2), the reduplicant segment that alliterates precedes
the segment that rhymes, but in (5-6) the reverse is true. Subscripts A, R show segments that alliterate or rhyme respectively. The reduplicants in both cases contain an alliterating onset and a rhyming rhyme, but only in the first case do they compose a syllable of their own.
38 38 3g8 g 38 38
lA aR - lA R k a d iR gA - iR gA u
Often there is no clear way to tell if reduplication is prefixing or suffixing, posing problems for an affixation account. Consider these Mangarayi facts:
-an ‘knowledgeable person’
Conveniently, there is no need in the approach proposed here to commit to which is base
, which is reduplicant; each relevant input segment simply has two output correspondents. Note also that like the VC-reduplicants above, the rhyming segment precedes the alliterating one.
Some languages use unmarked segments in reduplication (Alderete et al 1998). In the approach advocated here, if Markedness >> Alliterate
, onsets will become unmarked, and if Markedness >> Rhyme
, rhymes will become unmarked. Chaoyang (Yip 1998, to appear) shows a whole collection of patterns illustrating these interactions: in some patterns
, unmarked oral onsets become [l], while in others unmarked rhymes become [i] ([i ] for stop-final inputs). Again, no commitment to a base/affix distinction is needed. A hypothetical syllable /hop/ is used to illustrate the patterns.
Pattern 1 hop hop
Alliterate, Rhyme >> Markedness
Pattern 2 hop lop (+ suffix)
Rhyme >> Markedness >> Alliterate
Pattern 3 hi hop (kio)
Alliterate >> Markedness >> Rhyme
Pattern 4 hi hop lop (kio)
Alliterate, Rhyme >> Markedness
The units which rhyme or alliterate may need additional specification, since they can be feet as well as syllables. Foot reduplication was seen in (1) Diyari and (2) Kamaiura. A simple extension of the Rhyme and Alliterate constraints will deal with this. Foot-Rhyme /Alliterate: Output must contain two adjacent feet that rhyme/alliterate. They may need even further refinement: the relevant unit may be not just any syllable, but the head or stressed syllable. Samoan can be viewed in this way: a-lo-lófa ‘love’, fa-na-náu ‘be born, give birth’. Again, a simple extension zeroing in on stressed syllable rhymes and onsets will work here:Stress-Rhyme/Alliterate: Output must contain a syllable that rhymes/alliterates with the stressed syllable.Interestingly, the syllable/foot/head are also the units that may rhyme in poetry, a point to which I return.
The alert reader will no doubt have realized that one quite common type of reduplication poses a problem for this approach. In some languages the reduplicant must be a heavy syllable, even when a light syllable would be a better rhyme:
The non-existent *[pu
pusa] would better satisfy Rhyme.
One possibility is that this is word compounding
, with a minimal word effect requiring a bimoraic reduplicant. The actual output [puspusa] still rhymes featurally, given the following constraint: FeatRhyme: Any structurally identical positions (i.e. nuclei, off-glides, codas) in the two rhymes must match in features.
The nuclei are both /u/, and satisfy this, and the /s/ has no matching position, so it satisfies the constraint vacuously.
If this approach is right, reduplication and devices familiar from poetry are not so dissimilar, something noted by Kiparsky as early as 1973, and more recently by Holtman in her 1996 dissertation. Fabb (1997) gives an excellent summary of poetic rhyme and alliteration, and the parallels to reduplication are indeed striking. Both pay attention to syllable structure. Both may single out prosodically prominent syllables as the targets to copy (alliteration in Old Icelandic dróttkvaett/ Samoan reduplication). Both usually copy elements at the edges of words or other constituents (Kiparsky points out that in poetry languages with final stress tend to show end-rhyme, while ones with initial stress often use alliteration). Both show matching of non-constituent CV sequences (Finnish rhyme in the Kalevala/ Ponapean light-syllable reduplication). Both may allow matching of “equivalence classes” of sounds instead of full identity (Tuareg rhyme / Chaoyang reduplication, where all codas turn into their less marked glottal or velar equivalents, so that [ ] rhymes with [m] and [ ] rhymes with [k]). However, Fabb notes one difference between alliteration and rhyme in poetry: alliteration requires greater proximity than rhyme. There is no obvious correlate of this observation in reduplication.
I am led then to the conclusion that it may be useful to view reduplication as the most grammaticalized end of a continuum that can be diagrammed as follows:
reduplication with an iconic residue (plurality, intensification, iteration)
I do not suggest that a language can only have reduplication if it has all the phemomena above it in this continuum. The claim is more limited: humans have both an aptitude and a taste for creating repetitive sequences, and they may use this skill in a variety of ways that are more or less part of the core grammar of the language. (Of course, au fond grammatical knowledge is involved in all linguistic repetition, as can be seen by the implicit knowledge of syllable structure betrayed by these phenomena.) The mechanisms are the same, no matter what the function of the repetition. More specifically, I have claimed that the mechanisms distinguish rhyme and alliteration, and it is this that allows for repetition of either the whole or the part. These mechanisms have their grammatical incarnation as universal OT constraints, motivated here through reduplicative data, but presumably also involved in poetry if one accepts Holtman’s central claim that poetic rhyme is controlled by an OT grammar.2
Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy, and Suzanne Urbanczyk. 1996. Reduplication and segmental unmarkedness. Ms., University of Massachusetts, Amherst. ROA-134.
Fabb, Nigel 1997 Linguistics and Literature Basil Blackwell, Oxford.
Holtman, Astrid 1996 A generative theory of rhyme: an Optimality approach. Ph.D.Dissertation. OTS. Utrecht.
Inkelas, S. and C. Zoll 1999 Reduplication as double stem selection. Talk given at Phonology 2000 Symposium, Harvard/MIT April 1999.
Kiparsky, Paul. 1973. The role of linguistics in a theory of poetry. In Language as a human problem. Special issue of Daedalus. Reprinted in D.C. Freeman (ed.) 1981. Essays in modern stylistics. Methuen, London, p9-23.
Marantz, Alec. 1982. Re Reduplication. Linguistic Inquiry 13:483–545.
McCarthy, John, and Alan Prince. 1986. Prosodic Morphology. Ms., University of Massachusetts, Amherst and Brandeis University.
McCarthy, John & Alan Prince. 1993a. Prosodic Morphology I: Constraint Interaction and Satisfaction. To appear, MIT Press. Technical Report #3, Rutgers University Center for Cognitive Science. Pp. 184.
McCarthy, John and Alan Prince. 1995. Faithfulness and reduplicative identity. UMass Occasional Papers in Linguistics 18: Papers in Optimality Theory. ed. by Jill Beckman, Laura Walsh Dickey, and Suzanne Urbanczyk. 249-384.
Moravcsik, Edith (1978). Reduplicative constructions. In J. Greenberg (ed.), Universals of Human Language, vol 3: Word Structure (pp. 297_334). Stanford: Stanford University Press.
Raimy, E. And W. Idsardi 1997. A minimalist approach to reduplication in Optimality Theory. In K. Kusomoto, ed., NELS 27.
Sherrard, Nicholas 1997. Blending theory - Correspondence Theory with multiple inputs. Handout for talk at Linguistics Association of Great Britain, September 1997. U. of Essex.
Yip, M. (To appear) The role of markedness in onset change. Proceedings of “On the formal way to Chinese languages”. Ed. By C.S. Liu and S.W. Tang. Stanford, CA: CSLI Publications.
Yip, M. 1998 Multiple systems of reduplication in a single language: a markedness account Talk given at Oxford University.
c:\moira\chaoyang\glot.wpd June 4, 1999
39 Halsey Street
London SW3 2PT