THE RELATION BETWEEN PROTOTYPICAL AND MARGINAL MORPHOLOGY: THE CASE OF REDUPLICATIVE CONSTRUCTIONS

THE RELATION BETWEEN PROTOTYPICAL AND MARGINAL MORPHOLOGY: THE CASE OF REDUPLICATIVE CONSTRUCTIONS The article investigates a paradigm case of a borderline phenomenon in linguistic analysis: constructions in which linguistic material is meaningfully iterated (or repeated) and their relation to prototypical and marginal areas of morphology. The fuzzy state of affairs prevailing in this research area is described and a survey of relevant and irrelevant iteration phenomena is undertaken. The discussion finally narrowing in on morphological reduplication, the data dealt with (mostly) come from the typologically oriented Graz Database on Reduplication (gdr). In light of certain data encountered there (i.e. reduplicative imperatives and a pragmatically emphatic vowel copying construction), the morphological process of reduplication is further differentiated, the general term reduplicative construction (Moravcsik 1978) thereby being endowed with a special meaning subsuming both prototypical and marginal instances of the process under scrutiny, couching the investigation into a morphological framework along lines similar to those proposed by Zwicky/Pullum (1987) and Dressler (2000). The study concludes with a tentative general picture of repetition phenomena in language in which grammaticalization theory can comprehensively account for such constructions that linguistically range from discourse to morphology. Finally, the potential benefit of the present approach for any typological undertaking in the realm of the reduplication process is highlighted. Lingvistika-2011-01-93 1/5/12 1:32 PM Page 133

To this Moravcsik (1978: 300-301) adds the one a priori restriction that, within one language, the same particular meaning distinction paired off with a certain iterative form difference in utterances of the above kind has to recur in at least one other utterance pair in order to be counted as a reduplicative construction, a sensible restriction which in the present context is understood as a kind of minimal productivity requirement (essentially saying that an utterance pair thus defined should not constitute an isolated, hapax-like phenomenon in a language).Furthermore, it is important to note that in this conception reduplication results in a new, non-repetitive meaning, i.e. by repeating linguistic elements the process is not just saying the same all over again (cf.Stolz et al. 2011: 27).
As Gil (2005: 61, endnote 1; emphasis original) points out, Moravcsik's "use of the term reduplication is actually broader than that of most other scholars, subsuming also many constructions which would generally be considered as involving repeti-tion," thereby perfectly demonstrating the demarcational uncertainty that usually prevails in this area of research even in an oft-cited study which (in my opinion nonetheless justly) is regarded as a seminal treatment of reduplication up to this day.Moreover, additional obscurity is due to the fact that even in languages which lack reduplication in the narrower morphological sense of a productive process of inflection and/or word formation (i.e. the sense that the other scholars alluded to in Gil's quotation above adhere to and which is also adopted in the present study) the systematic and meaningful repetition of lexical material can be considered as a widespread (I would suggest: universal; see section 4) phenomenon (cf. Stefanowitsch 2007: 29).
It is exactly this fuzzy state of affairs which the ensuing discussion tries to endow with some more clarity.Drawing on a body of earlier work as well as applying and modifying some of the more generally proposed theoretical frameworks by confrontation with a typological dataset,1 the following sections tackle the questions of how to best distinguish reduplication proper (i.e.morphological reduplication) from other iteration/repetition phenomena and how to classify a newly defined concept of reduplicative constructions within a linguistic interface model differentiating between prototypical and marginal morphology.

THE RANGE OF (IR)RELEVANT ITERATION PHENOMENA
It seems more than warranted to approach the task of classifying and demarcating different iterative constructions from both a formal and a semantic angle.I deem the (interestingly in reduplication research) very frequently encountered bias towards one or the other pole of the linguistic sign to necessarily render any analysis incomplete. 2 The term reduplicative construction popularized by Moravcsik (1978) -but rather infelicitously used interchangeably with reduplication (a term reserved for the morphological use of repetition in this article; see section 1) by her -comes in very handy, in so far as it will be exploited here as an umbrella term for all kinds of repetitive constructions which at first sight share certain formal and functional characteristics but eventually may be further differentiated provided the right morphological model and criteria are applied (see section 3).But the starting point of inquiry must be an even more general one.Formally, phenomena ranging from the potentially infinite repetition of linguistic material on the discourse or text level right down to the duplication of single segments or sound features need to be considered at the outset.Semantically, it is indispensable to try and correlate formal properties with the functions or meanings they seem to serve or express, reduplicative constructions, as will become evident, being able to convey such different senses as (pragmatic) emphasis or (grammatical) plural agreement, arguably with some principle of iconicity at work as the overarching driving force (as I note only in passing with no intention to dwell upon this in itself intriguing subject any further here).

Excluded phenomena
The data collected and analyzed for the gdr within the Graz reduplication project on which this study mainly relies (see footnote 1) are a priorily restricted in that certain constructions and phenomena have been excluded right from the start, the database trying to focus on morphological reduplication as far as possible (but see footnote 5).Thus, by virtue of their lack of meaningfulness, definitely (also according to Moravcsik's broader definition and use; see section 1) out of the picture are -harmonic phenomena like assimilation and vowel harmony -phonological/prosodic doubling merely fulfilling templatic functions (cf. the definition and scope stated on the project's website and Hurch et al. 2008: 8-9) while somewhat more arbitrarily also not included are reduplication in sign language, child language and language games. 3The exclusion of sign language reduplication is only due to its incompatibility with the database structure (which was designed for integrating reduplication in oral languages) and should not be misunderstood as a neglect of its grammatical (i.e.morphological) status. 4 Apart from the fact that a contribution to meaning exclusively through reduplication seems to be missing in them, language games on the other hand have figured prominently in arguments for socalled expressive or extragrammatical morphology (e.g.Zwicky/Pullum 1987: 332-334), so their discussion in the present context most probably would not add anything useful to the general picture headed for (see especially sections 3.3 and 4).In essence, the last points also hold true for child language reduplication (e.g.Dressler et al. 2005) and although the question of meaning carried by reduplication in this domain seems to be somewhat less clear in many of the attested instances, some of them readily are conceivable as fulfilling -like phonological/prosodic doubling (see above)mere templatic functions, for example.

Problematic phenomena
Considerably less uncontroversially, the gdr additionally refrains from analyzing indepth the following phenomena, as is also explicitly stated on the Graz reduplication project's website: -repetitive syntactic operations -recursive morphological operations -contrastive reduplication These arguably constitute less negligible cases in the present context.But although they will consequently be discussed in more detail right below, the following subsections will also show that these phenomena nevertheless are relatively easy to delimitate from reduplication proper in most cases.5

Repetitive syntactic operations
The repetition of elements applying across words is a discourse or syntactic phenomenon (cf.Gil 2005: 31) of either a reinforcing and/or iconic nature when looking at its content side (see the examples in Gil 2005: 39-46).Example (1) below, illustrating a form of iconic repetition (namely durativity), comes from Arapesh (Torricelli, Papua New Guinea):6 3SG.M SUBJ-R-with dogs '... he went hunting with dogs and went and went and went a long way with the dogs.' (Conrad/Wogiga 1991: 53) Even more clear-cut are constructions that (optionally or obligatorily) use a linker explicitly conjoining repeated elements (e.g. the conjunction and in the English translation of the Arapesh example above) or that offer prosodic cues pointing at the relative independence of their component parts (e.g.pauses between the elements which are repeated, clearly separate intonational curves for each repeated unit, etc.).In most cases the size of the repeated constituent and/or the number of its (potential) repetitions (i.e.essentially formal characteristics) make clear whether one is dealing with repetition or reduplication.These and other diagnostics are especially fruitful for the distinction in languages that abundantly have both construction types. 7Of course, on the one hand, the extremely rare cases of grammatical triplication -e.g.Mokilese roar 'to give a shudder' -roar∼roar 'to be shuddering' -roar∼roar∼roar 'to continue to shudder' (Harrison 1973: 426) -could then be mistaken for syntactic repetition judging superficially from the number of repetitions alone.But for the Mokilese example this can easily be rebutted by the triplication's systematic meaning opposition vis-à-vis its simplex and properly reduplicated counterparts as well as the inability to add more than two copies to the base for the same result.A language like English, on the other hand, can pose problems if isolated examples like he is very very bright (Moravcsik 1978: 301) are investigated.It is maintained here that this does not instantiate a case of reduplication (English most certainly being a non-reduplicating language) but rather repetition, a view which is supported by Moravcsik (1978: 312) herself when she states that "it is perhaps true in all languages that an emphatic modifier [...] can be open-endedly reduplicated [i.e.repeated; see above; TS] for additional degrees of emphasis."This alleged open-endedness of the process runs counter to the usual restrictions found with morphological reduplication.

Recursive morphological operations
Having thrown a glance at the far outskirts of morphology, paying special attention to repetition phenomena in discourse and syntax in the preceding section, this one touches upon a genuinely morphological operation sometimes confusable with reduplication.A German word like Ur-ur-großvater 'great-great-grandfather' illustrates what could well be mistaken for an instance of affix reduplication exemplified here by Fijian (Austronesian, Fiji): vanua 'country' -vēī-vanua 'various countries' -vēī~vēī-vanua 'larger number of countries' (Schütz 1985: 367).While in Fijian the collective or distributive prefix vēī-is reduplicated for the expression of greater number in the noun, the multiple prefixation of Ur-in German is a recursive operation in which the (potentially infinite) repetition of the prefix is accompanied by simultaneous addition of the same meaning with each single occurrence.It follows that the meaning of Ur-ur-ur-ur-ur-großvater 'great-great-great-great-great grandfather', for example, can be arrived at by knowledge of the basic meanings of Ur-and Großvater and simply adding them up, thus almost mechanically calculating the intended overall linguistic content of the resulting form.Every instance of Ur-exerts its own independent meaning contribution upon the respective base words (i.e.Großvater, Ur-Großvater, Ur-ur-großvater, etc.) while the reduplication of vēī -above is a non-repetitive (see section 1), grammaticalized expression for a special plural meaning (plus it is restricted to the duplication of the affix so that, say, *vēī~vēī~vēī-vanua would be ungrammatical).Furthermore, this stacking of affixes in recursive morphological operations most probably only works with certain derivational meanings lending themselves easily to multiplication, a point which in contrast is not met by the rather inflectional affix reduplication example from Fijian above.

Contrastive reduplication
The last of the problematic cases excluded by the gdr treated in this section is at first sight perhaps also the hardest to keep apart from reduplication proper.Contrastive reduplication (or contrastive focus reduplication) can be illustrated by (colloquial) English SALAD-salad (as opposed to, e.g., tuna salad) which yields a sort of prototypical reading of the reduplicated item in question (see Ghomeshi et al. 2004).Formally, it looks like reduplication in that it is restricted to one repetition of the base, but it is quite different from it in that also whole phrases as in i don't LiKe-HiM-like-him (Ghomeshi et al. 2004: 321) may be duplicated.Functionally, it seems to be restricted to very special contexts in which a contrast to a (explicitly or implicitly) present counterpart wants to be achieved.Nevertheless, a certain affinity to genuine reduplication meanings like specificity ('specific type of X'; see Kiyomi 1995Kiyomi : 1154) cannot be denied.But given the construction's special information structural and formal status (Ghomeshi et al. 2004: 342 sum up that "the base is of variable size, in both syntax and phonology"), all in all the exclusion of contrastive focus reduplication from the gdr is not at odds with the overall morphological conception of the Graz database.

Residual phenomena
What has been done in this section was mainly to recapitulate and at points adapt existing viewpoints of rather well investigated areas related to the adequate delimitation of linguistic iteration phenomena.Looking through the eyes of morphology, phenomena from right across its external boundaries (repetitive syntactic operations; see section 2.2.1), other morphological operations (namely recursive ones; see section 2.2.2) and constructions with an unclear status as to their appropriate linguistic module of handling (contrastive reduplication; see section 2.2.3), all showing considerable similarities with morphological reduplication (the main focus of this article), have been examined.
The progress that this study aims to make lies in concentrating on the residue that is left (i.e.reduplication in the narrow sense) and demonstrating -by looking at specific typological data from the gdr -that also inside this domain more finegrained, essentially morphology internal distinctions can and have to be made, as not everything which looks like reduplication actually is located on the same level, a lesson which in their own way the preceding subsections have already taught from far less surprising perspectives.

REDUPLICATIVE CONSTRUCTIONS
In the beginning of section 2 it was already announced that the term reduplicative construction will be furnished with a new reading for the purposes of this article.Generally speaking, it is here understood as an umbrella term for constructions exemplified by ( 2) and (3) as well as by ( 4) to (6) below, each representative being taken from the gdr: (2) Eastern Oromo (Afro-Asiatic, Ethiopia) isáan did∼díkk'-oo8 they PL∼small-ADJ.F 'they are small' (Owens 1985: 93) (3) Chamorro (Austronesian, Guam) Hu li´e´i há∼hatsa.I see the

Reduplicative imperatives
In a recent typological survey of imperatives in language one comes across the statement that "reduplication is never the sole marker of an imperative" (Aikhenvald 2010: 33).This falls in line with an even more recently compiled list of inflectional categories not expressed by total reduplication which inter alia displays the imperative category 10 The fact that the Swahili simplex verb in (5) loses its final vowel in the course of reduplication (see Novotna 2000: 60) should not distract from the fact that the resulting structure is one of total reduplication. 11The occurrence of the glottal stop in ( 6) is non-phonemic: It is inserted between two vowels that occur in adjacent syllables as a result of morphological, syntactic or pragmatic factors in Tukang Besi (cf.Donohue 1999: 25).
(see Stolz et al. 2011: 194).But both of these assertions from the pertinent typological and reduplication literature are obviously in conflict with examples (4) and ( 5) above. 10 Admittedly, neither for Lavukaleve nor for Swahili do the respective authors explicitly speak of reduplication as expressing the imperative, the reader can merely infer this from the translations of the examples given.Still it seems to be the most natural reading as the Lavukaleve example (4) appears in the subsection on reduplication which is part of the higher-level discussion of morphophonemic processes in the language in general (see Terrill 2003: 35-36) while the Swahili example ( 5) is adduced by Novotna (2000: 60) when she speaks about the reduplication of verbs which lose their final vowel after having explicitly stated before that reduplication is understood as a fully grammaticalized process serving either morphosyntactic or word-formation purposes (cf.Novotna 2000: 58).
Nevertheless, this section does not intend to falsify Aikhenvald's and Stolz et al.'s typological claims in their entirety.Neither does it want to establish the imperative as a legitimate reduplicative category on a par with much more frequent meanings like (nominal or verbal) plurality, diminution or intensification (see Kiyomi 1995; for an example of intensifying reduplication see section 3.3).Rather it tries to pave the way for a motivated view in which the latter, prototypical reduplication meanings are adequately distinguished from marginal meanings like the imperative which in turn need to be kept apart from the repetition phenomena encountered in section 2.2.This calls for a gradient theoretical device appropriately capturing the facts at hand which with the highest probability seems to be hidden in a differentiated view of morphology itself.But first a closer look at a further construction type from the above examples is required.

Pragmatic emphasis
What remains to be discussed with example (6) is another special case of reduplication found in the gdr.Again it is the author of the primary source himself who discusses and classifies the phenomenon in question jointly with other much more clearly grammatical cases of the reduplication process (see Donohue 1999: 42).At the same time, though, its special status is quite vividly uncovered by describing (stressed) vowel reduplication 11 in Tukang Besi as being "used for pragmatic effect to emphasise the truth values of one word sentences" (Donohue 1999: 42).These quite severe formal and functional restrictions cast serious doubt upon the process in question as constituting an instance of what is commonly called morphological reduplication.Next to imperatives (see section 3.1) this is a second argument for a differentiated view of morphology in general and morphological reduplication in particular, finally outlined in the next subsection.

Reduplicative constructions vis-à-vis prototypical and marginal morphology
We are now able to pinpoint what actually is understood by the concept of a reduplicative construction in this article: The term basically refers to conventional morphological reduplication but at the same time is supplemented by components of theoretical frameworks stressing the non-uniformity of morphology and morphological processes as envisaged in the programmatic outlines of Zwicky/Pullum (1987) and Dressler (2000).Essentially conflating these two approaches, the basic division drawn here is one of prototypical (or core) morphology versus marginal (or peripheral or -and this is crucial -expressive) morphology.The main deviation from tradition that I introduce thus is a consequence of granting the marginal areas of morphology (i.e. less productive phenomena transgressing external and internal boundaries of morphology) a certain amount of expressivity while when looking at the basic works mentioned above Dressler's extragrammaticality (i.e. that which lies outside of grammar) and Zwicky/Pullum's expressive morphology more or less can be equated (cf. Dressler 2000: 1).
Applied to reduplication, such a view predicts that prototypical and marginal instances -which collectively will be referred to as reduplicative constructions -are possible for this specific morphological process as well, a prediction borne out by the gdr data, recall the inflectional and derivational examples ( 2) and (3) as well as the special formations (4), ( 5) and ( 6).It is argued that the latter three are located at the expressive margin of morphology, slowly passing over to non-morphological types of repetition.This needs some elaboration.
Sections 3.1 and 3.2 have to some extent dwelled upon reduplicative imperatives and Tukang Besi vowel reduplication, respectively.It has been shown that the pragmatic nature of the Tukang Besi case is first-handedly described in Donohue (1999).Prototypically performing directive speech acts in attempting to make the addressee do something (cf.Aikhenvald 2010: 14, endnote 1), imperatives, too, are a very pragmatically oriented category of language.And as Dressler (2000: 7) points out in his plea for the recognition of marginal morphology, a "close relation to pragmatics, i.e., morphopragmatics, is non-prototypical for morphology".12Furthermore, one can test the two constructions under scrutiny against the seven criteria for expressive morphology assembled by Zwicky/Pullum (1987: 335-338): The pragmatic effect of both construction types has already repeatedly been pointed out, the concrete imperative examples (4) and ( 5) even seeming to contain additional emphasis arguably due to some kind of time pressure present in situations in which they are normally uttered.Concerning the promiscuity with regard to input category it is interesting to note that not only verbs (i.e. the expected locus of imperative marking) are attested as bases, see (4), while when judging from the description and examples of Tukang Besi vowel reduplication the input category of this latter process should be potentially open.As demonstrated in (4), alternative outputs definitely exist in Lavukaleve, the Tukang Besi case in contrast is quite restricted formally.Special syntax obviously is present in Tukang Besi vowel reduplication due to its exclusive occurrence in one word sentences, but while syntactic make-up also plays an important role in the typology of imperatives (see Aikhenvald 2010: 92-97), no information for the Lavukaleve and Swahili examples is available on this topic.In addition, no useful information for either construction type can be extracted from the relevant sources for the criteria promiscuity with regard to input basehood, imperfect control and interspeaker variation.From a quick glance, all in all this may not leave a very satisfying impression in the first instance.But as the last three criteria do not explicitly fail (the relevant information is simply lacking) and three others are definitely fulfilled for each construction, it does not seem too daring to ascribe them some expressivity and thus a special status within reduplicative morphology after all.This gains additional support from the requirement formulated by Zwicky/Pullum (1987: 338;emphasis original) that "for a phenomenon to be classified as expressive morphology, it must have a significant number of the above criterial properties" which in itself is very vague and open to interpretation.
Remember that especially discourse or syntactic repetition has been characterized by the function of reinforcement (see section 2.2.1).In the conception put forward here it is not regarded as a mere coincidence that morphology external repetition and morphology internal intensifying reduplication (e.g.Arapesh rɨpok 'cut ' -rɨ~rɨpok 'hack up';Dobrin 2001: 36) should eventually meet and yield boundary phenomena like the ones described in detail above.Some time has been devoted there precisely to show the expressive (or reinforcing) character such formations convey.On the other hand one has to acknowledge their formal restrictions which can go as far as constituting a single vowel copying pattern as the one found in Tukang Besi.Considering the constructions' form and meaning properties they consequently neither are repetition nor prototypical reduplication but rather a boundary phenomenon.Contra Zwicky/Pullum (1987: 338) this surely is a continuum approach to the morphological component perfectly in line with the conclusion drawn in Dressler (2000).For I am convinced that if adopted and steadily tested against an ever-growing amount of relevant data, the proposed model of prototypical and marginal morphology in the realm of reduplication can eventually "be elevated to the rank of a systematic study of basic questions which are likely to illuminate research in morphology at large" (Dressler 2000: 8).

CONCLUSION: THE EMERGING GENERAL PICTURE OF REPETITION PHENOMENA IN LANGUAGE
Returning to the broad starting point of this study outlined in section 2, one can now try and see how the discussion at hand in general bears on the classification and possible development of repetition phenomena in language.What emerges (for details see below) is almost the exact mirror-image of the summarized prototype account (the details of which need not concern us here) by Stolz et al. (2011: 69).This is hardly surprising since these authors concentrate exclusively on formal aspects in their work when they differentiate between exact and non-exact partial and total reduplication in syntax and morphology from the point of view of prototypicality, with exact syntactic total reduplication (note the again much wider use of the term reduplication as devised here) constituting the prototype.This work, in contrast, concentrates on formmeaning pairings having the special property of displaying repetitive elements (in a fairly wide sense at first) and their gradual narrowing down from unrestricted reinforcing repetition to constrained reduplication (including partial reduplication but by no means confined to it) for grammatical purposes.I already tried to show how marginal morphological reduplicative constructions fit into this picture.
A promising way to comprehensively capture meaningful repetition phenomena ranging from discourse to morphology is provided by grammaticalization theory, the reduction of multiple repetition to simple binary reduplication structures being an essential feature of the grammaticalization process (cf.Dressler et al. 2005: 456).Gradual transitions and hence boundary, marginal and prototypical grammatical phenomena fall out naturally from such a conception and enable a more adequate take on the reduplication process and its related constructions from a typological perspective.
A little gedankenexperiment to conclude this study: In the colloquial style of my native language (Austrian) German it seems perfectly fine to say something like schnell schnell 'quick(ly) quick(ly)' (but also more repetitions like schnell schnell schnell, schnell schnell schnell schnell, etc. are possible) approximately meaning 'quickly!, do it quickly!, hurry up!'.This is very reminiscent of (one is tempted to say: parallel to) the Lavukaleve example in (4).Now, German is by all means as non-reduplicating as English (cf.Section 2.2.1), still it has a reduplicative construction very similar to one found in a quite extensively reduplicating language.Of course I do not claim that German is special in this regard.Although not a native speaker, I can very well imagine a similar construction being possible in English, for example.I would even go as far as conjecturing that it is a universally available construction type that we are dealing with here.But recall two facts: First, in the German case multiple repetitions are possible.Second, Lavukaleve shows multiple patterns for the same meaning ranging from total reduplication to partial CV-reduplication.In accordance with any garden-variety theory of grammaticalization such formal differences in restriction and reduction are indicative correlates assigning German a less grammaticalized status concerning the construction under scrutiny when compared to its Lavukaleve correspondent.All this is expected under the view purported here.The universal availability of repetitive imperatives suggested above is linked to the universality of general reinforcing repetition in discourse, the difference being that in German the restricting systematization has not progressed as far as in Lavukaleve, making the repetitive imperative a morphology external phenomenon in the former and a marginal morphological (i.e. a truly reduplicative) phenomenon in the latter language.It is not entrenched deeper in the reduplicative system of Lavukaleve as has been shown in section 3.3.The question is of course if further entrenchment is a possibility at all without showing any additional formal markers (e.g.imperative particles, simultane-ous imperative affixes, etc.).Aikhenvald's and Stolz et al.'s typological generalizations (see section 3.1) suggest a negative answer to this question.But although it will not be possible to settle the issue here once and for all, it needs to be stressed that the findings made in this article at least render a total exclusion of reduplicated-only imperatives from the domain of morphological reduplication untenable.The fact that they nonetheless are to a certain extent different from constructions normally called reduplication merely highlights the necessity of a more fine-grained conception of morphology and its relation to neighboring linguistic modules.