Phrasal reduplication and dual description
Elinor Keane
University of Oxford
The patterns of reduplication that have dominated recent theoretical discussions involve either complete lexical items or subparts thereof, and the reduplicant is accordingly characterised as either an affix or part of a stem-stem compound (e.g. McCarthy & Prince 1995). Some types of reduplication, however, may affect not only single words but also sequences of lexical items. Such ‘phrasal reduplication’ challenges theories in which reduplication is confined to the lexicon, and has important implications for where it belongs in the grammar, and even how the grammar itself is organized. This paper discusses examples of this type involving echo reduplication, surveying previous work on Bengali and Kannada and presenting the results of fieldwork on the Dravidian language, Tamil. A new proposal, set within the framework of declarative phonology and termed the ‘dual description’ approach, offers a unified description of these data, accommodating both their fixed segmentism and the variation in their constituency.
Echo reduplication belongs to a range of reduplicated structures shared by the Indian languages, and is distinguished both by its fixed segmentism (the introduction of one or more fixed segments into the reduplicant – see Alderete et al. 1999) and the fact that its base is independently meaningful. The nature of the fixed segments varies between (and to some extent within) languages: example (1) illustrates the most productive pattern in Hindi, whilst (2) and (3) demonstrate how fixed segmentism interacts with the transfer of distinctive vowel quantity in Tamil.
(1) kele ‘bananas’ kele vele ‘bananas and such fruit’
(2) paam ‘money’ paam kiam ‘money and so forth’
(3) maau ‘cattle’ maau kiiu ‘cattle in general’
The exact meaning is context-dependent, although it usually involves generalisation, and such expressions are restricted to colloquial speech.
Typically echo reduplication applies to individual lexical items, and research on Hindi (Reynolds 1998) suggests that this may be the only possibility in some languages. In others, however, it may affect a range of constituents, from subparts of words to whole phrases. This is claimed to be the case in Bengali by Fitzpatrick-Cole (1996) and in Kannada by Lidz (2000): in both languages nominal roots can be echoed independently of associated inflectional material, and a few instances of phrasal reduplication are cited. The situation in Tamil was investigated through a questionnaire given to twelve dialectally homogenous native speakers. Their answers were relatively consistent, although by no means uniform, and disfavoured echoing subparts of words. However, the range of reduplicated phrases permitted was broader than in either Bengali or Kannada: adjectival, postpositional and some verbal phrases, including example (4), were all judged acceptable.
(4) [kumaarukku kuutteen] [kimaarukku kuutteen]-u poy collaatee
Kumar.dat give.past.1s echo quote lie say.neg.imp
‘Don’t lie that you gave it to Kumar or some such nonsense.’
A single pattern of reduplication thus allows variation in the make-up of its base, but how can the different possibilities be characterised? Fitzpatrick-Cole and Lidz draw very different conclusions: Lidz claims that the base corresponds to a morphological or syntactic constituent in Kannada, whereas Fitzpatrick-Cole states the restrictions on the base for echoing in Bengali in prosodic terms. The Tamil data are amenable to a syntactic analysis: all licit phrases are headed by a lexical category. However, not enough is known about Tamil prosody to determine whether they correspond consistently to prosodic constituents, and are only indirectly dependent on syntactic structure.
In either case, reduplication is clearly able to manipulate both sublexical and supralexical structure, which poses considerable problems for lexicalism, and indeed any theory in which reduplication is seen solely as a morphological process of word formation akin to affixation or compounding. Solutions involving various different morphological frameworks are discussed, including splitting reduplication into two processes (the assignment of a morphosyntactic [+reduplication] feature and provision of appropriate phonological substance at a later stage of the derivation) or moving it into a post-syntactic morphology component. Either is compatible with the new ‘dual description’ model, which involves no commitment to reduplication occurring in the lexicon.
Dual description is based on the principles of declarative phonology (e.g. Bird 1995), which distinguishes between objects and their descriptions, with the descriptions acting as constraints upon the form of the object. The special properties of reduplication are explained by having two descriptions jointly acting as constraints on a single object, the reduplicant – both the description of the base, which is fully specified structurally and segmentally, and a partial description of the reduplicant. The two descriptions will not coincide in every respect and may well conflict, especially if fixed segments are involved. In correspondence theory (McCarthy & Prince 1995) this conflict is modelled by competing constraints, on the nature of the reduplicant versus the relationship between base and reduplicant, and mediated by relative ranking of the relevant constraints. In dual description there is no such competition: the description of the reduplicant always takes priority over the description of the base. This could be formalized in various ways, such as a default inheritance system, in which the description of the reduplicant overrides the default description of the base.
Various examples are provided in the paper to show how this conjunction of descriptions works out in practice. For instance, if a single lexical item is echoed, the description of the reduplicant may contain a prosodic word node (w) to indicate its constituency, as well as the fixed segments. This would be represented as in (5a), and combined with a fully specified base, such as that shown in (5b), to produce the reduplicant in (5c).
(5a) w (b) w (c) w
s s s s s
m m m m m m m
k i m a u k i u
Instances of phrasal reduplication are accommodated by allowing the base to contain more than one lexical item, and the constituency of the reduplicant can be stated in terms of prosodic, morphological or syntactic structure. Hence, for the phrasal example in (4), the description of the reduplicant would include the syntactic characterisation VP, and also specify the fixed segments of the first syllable. Dual description thus offers an integrated account of echoed phrases, using the same basic mechanism, prioritised conjunction of descriptions, to implement both fixed segmentism and also variations in the constituency of the reduplicant.
References
Alderete, John, Beckman, Jill, Benua, Laura, Gnanadesikan, Amalia, McCarthy, John J. & Urbanczyk, Suzanne. (1999). Reduplication with fixed segmentism. Linguistic Inquiry 30. 327–364.
Bird, Steven. (1995). Computational phonology. Cambridge: Cambridge University Press.
Fitzpatrick-Cole, Jennifer. (1996). Reduplication meets the phonological phrase in Bengali. Linguistic Review 13. 305–356.
Reynolds, Elinor. (1998). Echo words, with particular reference to Hindi. M.Phil. dissertation, University of Oxford.
Lidz, Jeffrey. (2000). Echo reduplication in Kannada: implications for a theory of word-formation. University of Pennsylvania Working Papers in Linguistics 6(3). 145–166.
McCarthy, John & Prince, Alan. 1995. Faithfulness and reduplicative identity. Papers in Optimality Theory: University of Massachusetts Occasional Papers 18. 249–384.
2