Discontinuous reduplication: a typological sketch

1 Introduction

Reduplication is undoubtedly one of the linguistic phenomena that has attracted linguists’ attention in the last decades. This is probably due to its fascinating duplicative, non-concatenative and polyfunctional nature. At the same time, there is as yet a shortage of adequate descriptions and, therefore, our understanding of the phenomenon is still largely limited (cf. Stolz 2008: 122 on full reduplication), for exactly the same reason: its formal and functional variability, both inter- and intra-linguistically, makes it particularly hard to catch its inner nature. In this paper, we aim at filling one of the many gaps still remaining in the domain of reduplication, by giving a first cross-linguistic description of one of the reduplicative patterns that have been described less, namely, discontinuous reduplication.

To begin with, we define the main types of reduplication described in the literature (Section 2). Then, we narrow our object of analysis to discontinuous reduplication (Section 3) and present our methodology, i.e., the language sample and the parameters used for the analysis (Section 4). In Section 5, we present the formal and functional properties that discontinuous reduplication displays in the languages of our sample, while Section 6 focuses on the typological generalizations that can be identified from the observation of the available data. Finally, before concluding with some considerations and suggestions for future research (Section 8), we discuss a case study on discontinuous reduplicative patterns in Italian in the light of our cross-linguistic investigation (Section 7).

2 Reduplicative patterns: definitions and types

The strong (and still increasing) interest in reduplication has produced a considerable amount of knowledge about this phenomenon in all its manifestations, from both a theoretical and comparative point of view (e.g., Hurch 2005; Inkelas 2014; Inkelas and Downing 2015; Inkelas and Zoll 2005; Key 1965; Moravcsik 1978; Stolz et al. 2011; Urdze 2018; among many others) and a language-specific or areal point of view (e.g., Abbi 1992; Fabricius 1998; Mattes 2014; Mous 2013; Kallergi 2015; among others).

In the literature, several different definitions have been proposed. However, the most widely recognized definitions are the following ones:

The systematic repetition of phonological material within a word for semantic or grammatical purposes. (Rubino 2005: 11)

[T]he repetition of part or all of one linguistic constituent to form a new constituent with a different function (Inkelas 2014: 169)

In other words, reduplication represents the doubling of (part of) a linguistic element in order to express some kind of (grammatical or lexical) meaning. The most widely known classification is probably the one distinguishing between partial and full reduplication. While partial reduplication consists in the doubling of a sub-segment of a linguistic element (be it a single phoneme or a syllable or even more than one syllable) and it may involve its initial, final or an internal part (cf. 1), full (or total) reduplication consists in an exact copy of the linguistic element itself (cf. 2).^[1]

However, some other special types such as echo reduplication/formation or automatic reduplication have also been recognized in the world’s languages (cf. e.g., Inkelas 2014; Rubino 2005). The former is defined by Inkelas (2014: 170) as the “reduplication of a word, with replacement of the onset or, sometimes, vocalism or internal material in one copy”; in other words, it consists in the repetition of an entire linguistic element, but the reduplicant^[2] displays one segment that differs from the reduplicand (cf. 3a). The latter is a kind of “reduplication that is obligatory in combination with another affix, and which does not add meaning by itself to the overall construction”, as can be seen in the Ilocano example (3b), in which partial reduplication is mandatorily combined with the simulative prefix agin- (Rubino 2005: 18).

Another type of reduplication that has been less widely investigated is discontinuous reduplication, the focus of Section 3.

From a functional point of view, reduplication is often described as a highly iconic grammatical device, due to its “duplicative nature” (Inkelas and Downing 2015: 503). Hence, the functions that reduplication expresses are generally traced back to the notions of plurality/augmentation/more/increase, such as plurality, distributivity, pluractionality, and intensification. Despite this, the literature also acknowledges some other functions that have been described as non-iconic (e.g., diminution, approximation, focus, and so on). The debate on the alleged iconic nature of reduplication is extensive and, since it is not one of the topics we are going to discuss here, we will not go into it.^[3]

3 Discontinuous reduplication: our object of analysis

This paper focuses on one of the special types of reduplication that still lack systematic description, namely discontinuous reduplication (henceforth DR). DR is generally defined as a kind of reduplication “where other morphological material may appear between the reduplicant and the base” (Velupillai 2012: 101; cf. also Rubino 2005: 17). Therefore, DR qualifies as a type of both “non-prototypical reduplication” in the sense of Urdze (2018) and “non-canonical reduplication” in the sense of Stolz (2018), since one of the basic features of reduplicative structures – namely adjacency – is breached.

We also take into account strictly morphological structures like those in (4) and (5), and more syntax-like strings like those in (6) and (7).

Our choice of considering cases generally described as (morphological) reduplication and cases generally described as (syntactic) repetition is due to a difficulty in discerning them from a typological perspective (see Barotto and Mattiola 2020; Gil 2005; Stolz and Levkovych 2018).^[4] In other words, we avoid positing a clear-cut boundary between morphology and syntax and between reduplication and repetition which could ultimately prevent us from unveiling interesting correspondences in cross-linguistic perspective. This also means that DR as we intend it includes what Inkelas (2014: 172) calls “syntactic doubling” (cf. the Ewe example in 7), i.e., a case where “the same word is deployed twice in a grammatical construction, sometimes separated by linking material or other syntactic elements” and what Stolz (2009: 101) calls “syndetic reduplication”, i.e., reduplicative constructions where two identical chains of segments are connected by additional phonological material, primarily a coordinative conjunction (8).

The latter example is reminiscent of Jackendoff’s (2008) NPN construction – exemplified by English expressions like day by day, face to face, volume after volume – where the linking element is a preposition instead of a conjunction.

Here, we focus on discontinuous patterns involving full exact reduplication (as in 4–7), leaving in the background similar strategies such as full but non-exact reduplication (cf. 9, where the repetition is not fully symmetric, due to the diminutive marker on the copy) or echo-reduplications that involve replacement of the beginning of the copy with a fixed substring (as in Yiddish-derived shm-expressions like English breakfast schmeakfast, cf. Inkelas 2014: 171).

Our choice is driven by the fact that we are not sure that these phenomena can be interpreted as actual cases of DR. They seem to be special cases of DR (like 9) or different phenomena that at a first glance resemble DR (viz. echo reduplication). Since this is a first exploratory study of DR, we took into account only cases in which the repetition is exact, in order not to alter our description.

4 Methodology: language sample and parameters of analysis

As mentioned in Section 1, our aim is to offer a preliminary description of the formal and functional properties of what we define as DR from a cross-linguistic perspective. To do this, we need a sample of languages. Our sample counts 99 languages (see Appendix A) and was designed starting from the 76-language sample proposed by Bybee et al. (1994). We adapted the latter sample for our convenience, that is, we included Bybee et al.’s (1994) languages for which we had grammatical descriptions. We replaced languages for which we did not have a grammar with the genealogically most closely related language for which we found a grammatical description. We then added 23 new languages to this balanced sample, trying insofar as possible to maintain the overall balance. The addition is for convenience (all additional languages display DR patterns) and for maximizing linguistic variety. Consequently, our sample should be considered as something closer to a convenience sample than a variety sample, but, at the same time, it was not fully designed for convenience and therefore still retains some degree of representativeness. The geographic distribution of the languages within the sample is represented in Map 1.

Map 1:

Geographic distribution of the languages within our sample.

In our sample, we found 24 (out of 99) languages displaying at least one case of DR. The geographic distribution of languages with at least one occurrence of DR is illustrated in Map 2.

Map 2:

Geographic distribution of the languages with at least one occurrence of DR within our sample.

Even though we cannot draw any areal conclusions because of the nature of our sample, it is nevertheless remarkable that almost all the languages we found displaying DR are mainly spoken in West Africa and South-East Asia, but also Oceania and South America. These are areas in which reduplicative patterns in general are particularly widespread. Interestingly, we did not find occurrences in other areas in which reduplication is largely present, in particular North America (more specifically, Pacific Northwest languages, such as Salishan, Tsimshian, Wakashan, etc., cf. Mithun 1999: 42), but also East and South Africa (e.g., Cushitic – cf. Mous 2013 – and Nilotic, Khoisan, and Bantu languages, cf. Rubino 2005: 22).

It is also noteworthy that no European language in our sample was found to display DR, despite Stolz’s (2009) findings (cf., e.g., the Latvian case in 8). However, this is not so unexpected if we consider that we based our cross-linguistic investigation on grammatical descriptions (and not on previous literature): it is well-known that reduplication in grammars of European languages is generally under-recognized, let alone discontinuous reduplication.

From a methodological point of view, we collected all the occurrences we could find within grammatical descriptions in a dataset and then we annotated each of them for five parameters, four of which pertain to form (i–iv) and one (v) to function:

1. Morphological reduplication versus syntactic repetition

2. Nature of the interposing element (linker, adposition, etc.)

3. Nature of the base (noun, verb, adjective, etc.)

4. Presence versus absence of other (contiguous) reduplicative patterns in the same language

5. Function of the DR pattern

Our final dataset counts 214 examples corresponding to 61 different types of constructions^[5] (see Appendix B) in 24 languages. The findings we are going to discuss in Section 5 are based on the number of types/constructions (and not on the number of examples), since the ratio between the two is quite uneven in our dataset. Consider the data in Figure 1.

Figure 1:

Number of types and examples for each language within our dataset.

Figure 1 shows the number of types and examples we found for each language displaying at least one case of DR. We can see that the overall picture changes if we consider the types or the single examples. The two languages with by far the highest number of examples are Uduk (Koman, Central Koman) and Kuku-Yalanji (Pama-Nyungan, Yimidhirr-Yalanji-Yidinic), with 40 and 31 respectively, but if we consider the number of types, the same languages have a much smaller number of constructions, both displaying only 2 DR types. This is obviously due to the fact that in Uduk we found 2 types of DR but, for one of them, we retrieved 39 examples (1 for the other type); in Kuku-Yalanji we found 2 types but, again, for one of them we retrieved 30 examples (1 for the other). In terms of types, the languages with the highest incidence and variety of DR are Obolo and Mungbam (both Atlantic-Congo, Volta-Congo), with 8 and 7 types respectively, followed by Chácobo (Pano-Tacanan, Panoan) and Chinese Mandarin (Sino-Tibetan, Sinitic), both displaying 6 different types. Most languages with DR, however, exhibit only 1 or 2 different types.

Since our main aim in this paper is to give a first typological sketch of DR, we want to avoid any over-representation of construction types that could blur our results, thus we center the subsequent discussion on types rather than examples.

5 Formal and functional properties of discontinuous reduplication

In this section, we discuss the formal and functional characteristics of DR based on what we found in our language sample. In Section 5.1, we present the formal properties of DR, while in Section 5.2 we focus on the functions that DR patterns express.

5.1 Formal properties of DR

As explained in Section 4, we annotated each construction type in our dataset according to the five parameters listed in Section 4. We will now discuss each of these parameters in more detail and illustrate the possible values with examples.

5.1.1 (Morphological) reduplication versus (syntactic) repetition

In our sample, we found occurrences of both (morphological) reduplication (10) and (syntactic) repetition (11).

(10)

Standard Malay (Austronesian, Malayo-Polynesian)			[Hudson 1995: 94]
saya	kirim-men-girim	surat	dengan-nya
I	send-meN-send	letter	with-him/her
‘He/she and I send letters to each other.’

(11)

Ilocano (Austronesian, Malayo-Polynesian)		[Rubino 1997: 149]
sao	a	sao
talk	lig	talk
‘always talking, loquacious’

In the dataset we found 31 types of (morphological) reduplication and 30 types of (syntactic) repetition. If we look at the distribution of (morphological) reduplication and (syntactic) repetition for each language, we see that only 2 out of 24 languages show both strategies: Chácobo (Pano-Tacanan, Panoan) and Khasi (Austroasiatic, Khasi-Palaung). All other languages display either only reduplication or only repetition (Figure 2).

Figure 2:

Number of types of (morphological) reduplication or (syntactic) repetition per language.

5.1.2 Nature of the interposing element

In our sample, we found 13 different types of interposing elements: adpositions, case markers, empathic markers, epenthetic elements, exclusive particles, intensive markers, linkers (e.g., ‘and’), negative elements, numerals (i.e., ‘one’), quantifiers, reciprocal markers, stative markers, and, finally, unknown elements. Figure 3 summarizes the number of types we collected for each kind of interposing element.

Figure 3:

Number of types per interposing element.

The types of adpositions we found in the dataset are elements that usually express comparative, instrumental, and spatial relations. In (12), we give an example of an adposition expressing a spatial relation.

(12)

Hindi (Indo-European, Indo-Iranian)					[Kachru 2006: 101]
tum	din	pər	din	ɖhīʈh	hote	ja
you	day	upon	day	impudent	become.ipfv.pl	go
rəhe	ho.
prog.pl	prs.pl
‘You are becoming more and more impudent day by day.’

The second type of interposing element are case markers. The most common are genitive and dative markers. The latter is exemplified in (13).

(13)

Mparntwe Arrernte (Pama-Nyungan, Arandic-Thura-Yura)
				[Wilkins 1989: 348]
Re	gotta	lhe-rle	urrkape-tyeke	arlte
3sg.sbj	have_to	go-gen.evt	work-purp	day
arrpenhe-k-arrpenhe .
other- dat -other
‘He has to work each day.’

Markers that encode some kind of emphasis can be found in DR patterns, as in the Majhi (Indo-European, Indo-Iranian) example in (14).

(14)

Majhi (Indo-European, Indo-Iranian)		[Dhakal 2014: 20]
kapal	→	kapal-e kapal
head		head-em head
‘head’		‘nothing but heads, all heads’

Some languages of the world display an interposing element that does not have (or no longer has) any kind of semantic value – that is, an epenthetic element (15).

(15)

Chácobo (Pano-Tacanan, Panoan)				[Tallman 2018: 879]
náa	bari	hawɨ	mi	a-ʔ-á =baʔina=ʔá
dem1	day	thing	2sg	do∼ ep -do=all/each_day=inter:p
‘Just recently, today you had been killing the things (the monkeys)?’

In our sample, we found a single language displaying an exclusive particle in its DR pattern, i.e., Punjabi (Indo-European, Indo-Iranian):

(16)

Punjabi (Indo-European, Indo-Iranian)			[Bhatia 1993: 98]
dil	vicc	hii	vicc
heart	in	exc.part	in
‘right inside the heart’

In Mongsen Ao (Sino-Tibetan, Kuki-Chin-Naga), the interposing element is represented by the intensive marker -tsáʔ.

(17)

Mongsen Ao (Sino-Tibetan, Kuki-Chin-Naga)			[Coupe 2007: 366]
tsafùʔ	pi	la	atsaŋ-tsáʔ-atsaŋ -ə̀ɹ-ùʔ
cooking_pot	prox	top	be_heavy-intens3- red-prs-dec
‘This cooking pot is really heavy.’

One of the most frequent interposing elements are linkers. By linker we mean any element whose main function is to connect the two copies. In the majority of cases, the linker is a coordinating element corresponding to ‘and’.

(18)

Kuku-Yalanji (Pama-Nyungan, Yimidhirr-Yalanji-Yidinic)
		[Patz 2002: 108, 229]
dunga-y	→	dunga-n-dunga -ri-y
go-npst		go- lnk -go-pl.sbj-npst
‘go’		‘keep going’

In two languages of the sample (Punjabi and Maithili), we also found negative elements interposed in DR patterns. See for instance (19).

(19)

Punjabi (Indo-European, Indo-Iranian)					[Bhatia 1993: 220]
ó	ikk	na	ikk	day	aavegaa
he	one	neg	one	day	come-fut.2m.sg
‘He will come someday or other’

In Mandarin Chinese, the numeral ‘one’ can be used between two verbs.^[6]

(20)

Mandarin Chinese (Sino-Tibetan, Sinitic)	[Li and Thompson 1981: 209]
xiào-yi-xiào
smile-one-smile
‘Smile a little!’

We identified three languages (Mparntwe Arrernte, Ewe, and Yoruba) in which the interposed element is a quantifier. See for instance (21).

(21)

Yoruba (Atlantic-Congo, Volta-Congo)		[Bamgboṣe 1966: 153]
ìwà	→	ìwà-k-ìwà
behavior		behavior-any-behavior
‘any behavior, i.e., bad manners’

In Standard Malay (Austronesian, Malayo-Polynesian), we found a reciprocal marker intervening between the two copies:

(22)

Standard Malay (Austronesian, Malayo-Polynesian)	[Hudson 1995: 93]
kami	kejar-meN-gejar
we	chase- recp -chase
‘we chased each other.’

The last interposing element we identified in our sample is a stative marker:

(23)

Uduk (Koman, Central Koman)		[Killian 2015: 212]
jǎn	‘pér-á	‘pēr
dem.med	red-stv	red
‘That is red’

Finally, in several occasions (12 types throughout 8 languages), we were unable to find information on the nature of the element interposed within DR patterns. (24) is a case in point:

(24)

Jaqaru (Aymaran, Tupe)		[Hardman 2000: 54]
t’usqi	→	t’usqi-ch-t’usqi
dust		dust-ch-dust
‘dust’		‘to be causing a lot of dust’

5.1.3 Nature of the base

In our dataset, we have 8 types of bases that can undergo DR: adjectives (25), adpositions (26), adverbs (27), nouns (28), numerals (29), quantifiers (30), verbs (31) and unknown class (i.e., elements for which we have no information, cf. 32).

(25)

Balochi (Indo-European, Indo-Iranian)			[Axenov 2006: 88]
am-ē	gwanḍ-u-gwanḍ -ēn	zāg	int-ī.
em-dem	little-and-little-attr	son	cop.prs.3sg-enc.3sg
‘This is his very little son.’

(26)

Punjabi (Indo-European, Indo-Iranian)						[Bhatia 1993: 205]
ó	ne	báár	de	báár	Taal	dittaa
he	erg	outside	gen.m.sg.obl	outside	put_off	give-pst.m.sg
‘He put (me) off right from outside’

(27)

Mparntwe Arrernte (Pama-Nyungan, Arandic-Thura-Yura)
[Wilkins 1989: 182–183]
Kele	mweteke	there	kaperte	itwe-k-itwe	ne-rlenge,
O.K.	car	two(s)	head	near- dat -near	be-ds
itne	ultake-lhe-ntye	re-nhe	iteth-ile-ke.
3sg.a	break-rfl-nom	3sg-acc	alive-caus-pc
‘So, when the two cars were bonnet to bonnet (lit. head near to near), they started the broken one (by jump starting it).’

(28)

Tshobdun (Sino-Tibetan, Burmo-Qiangic)	[Sun 2014: 634]
sŋi-ku-sŋi
day-ku-day
‘every day’

(29)

Obolo (Atlantic-Congo, Volta-Congo)	[adapted from Rowland Oke 2003: 164]
ítá-mè-ítá
trois-et-trois [three-and-three]
‘trois par trois [three by three]’

(30)

Uduk (Koman, Central Koman)				[adapted from Killian 2015: 133]
rìs	→	ris	ká	rìs
many		many	with	many
‘many’		‘very many’

(31)

Mungbam (Atlantic-Congo, Volta-Congo)				[Lovegren 2013: 197]
ù	kpɔ̀ŋ	kà	kpɔ̀ŋ	ì-kwɛ̀hɛ
cl1.top	(a) shout. ipfv	?? [7]	( a) shout. ipfv	cl9-bird.type
‘He is shouting [his battle cry] as the ikwɛhɛ bird.’

(32)

Maithili (Indo-European, Indo-Iranian)		[Asad 2015: 50]
koɪ	nɘ	koɪ
some	em	some[8]
‘someone else’

From the quantitative point of view, verbs and nouns are the most commonly reduplicated categories, followed by adjectives, adverbs and, to a minor extent, numerals. The presence of other categories is quite sporadic, as illustrated in Figure 4.

Figure 4:

Number of types per nature of the base.

This distribution is even more apparent when considering the number of examples/occurrences found per lexical category: while the type/token ratio is close or equal to 1:1 for bases belonging to minor lexical categories, the number of examples per type is much higher for major lexical categories: 91 examples for 21 types with verbs, 29 examples for 13 types with nouns, 61 examples for 10 types with adjectives, 13 examples for 8 types with adverbs.

5.1.4 Presence or absence of other reduplicative patterns

Our last formal parameter deals with the co-presence of DR and other kinds of reduplicative patterns. More specifically, we refer to prototypical cases of reduplication (i.e., partial and full reduplication) when the two copies are contiguous. Interestingly, in all the languages displaying DR, other reduplicative patterns are also attested. In (33)–(36), we exemplify cases of prototypical reduplication in four languages for which we have already presented some DR patterns: Ilocano (11), Jaqaru (24), Maithili (32), and Uduk (30).

(33)

Ilocano (Austronesian, Malayo-Polynesian)		[adapted from Rubino 1997: 117]
Naut-ot	ti	bag∼bagi =na	ken	ag-bibbi ∼ bineg	ti
adj-sore	art	red ∼body=3sbj.erg	and	I-red ∼numb	art
la∼lasag =na
red ∼muscle=3sbj.erg
‘His whole body was sore and (all) his muscles were numb.’

(34)

Jaqaru (Aymaran, Tupe)			[adapted from Hardman 2000: 53]
Qaylla-q	was	yatx-k-ushu-q	jal∼jal -k-i-wa
child-ss	walk	learn-sim-Ssub-ss	red ∼fall-inc-3>3-ss
‘Children, when they are learning to walk, fall all the time.’

(35)

Maithili (Indo-European)					[Asad 2015: 31]
O	git	sunɘIt	sunɘIt	tʰɘIk	gel.
he	song	listen	listen	tired	went.pst
‘He got tired of listening to songs’

(36)

Uduk (Koman, Central Koman)			[Killian 2015: 396]
ádɨ̄	mɨ́	tō-ánsān	’kósh-ɨ́	shúm	kɨ́
3sg	do:pfv	thing-dem.prox	kill-ad2	animal	compl
nyànyɨ́∼nyànyɨ́
many: idph ∼ red
‘He did this and killed many animals.’

5.2 Functional properties of DR

The DR patterns in our language sample can express quite a wide set of functions. More specifically, we identified 19 different functions: approximation, category-changing, continuativity, delimitation/diminution, distributivity, generality, greater plurality, habituality, identificational focus, inchoativity, indefiniteness, intensification, pluractionality, (additive) plurality, precision, predicativization, reciprocity, universal quantification. All of these functions are typical of reduplicative patterns in general (cf. Inkelas 2014; Inkelas and Downing 2015; Mattiola and Barotto to appear; Moravcsik 1978; Rubino 2005; among many others). Figure 5 provides a quantitative ovierview, illustrating the number of types/constructions per each function in our dataset.

Figure 5:

Number of DR types per function.

Not unexpectedly, the most attested functions are intensification, distributivity, continuativity and pluractionality, whereas other functions like approximation, diminution or plurality are less common (see also Section 6). In what follows, we provide definitions and examples for each of the functions we identified.

The first is approximation, i.e., cases in which DR expresses an intentionally vague concept or non-prototypicality with respect to the item being reduplicated. See for instance (37):

(37)

Ewe (Atlantic-Congo, Volta-Congo)		[Ameka 1991: 61]
tɔ́ɖe	gbɔ́	tɔɖě
uncle	side(/near)	uncle
‘a pseudo-uncle’

Rather than expressing a grammatical function, some instances of DR in our sample are employed to create a new lexeme, especially by changing the lexical category of the reduplicant. This category-changing property is found for instance in Khasi:

(38)

Khasi (Austroasiatic, Khasi-Palaung)				[Abbi 1992: 27]
tuh	→	tuh	sa	tuh
steal		steal	sa	steal
‘steal’		‘crafty’

Sometimes, DR patterns are used to express that an event is prolonged over time. Continuativity is generally defined as the case in which “a dynamic situation is ongoing and […] the agent of the action is deliberately keeping the action going” (Bybee et al. 1994: 127), as exemplified in (39):

(39)

Kuku-Yalanji (Pama-Nyungan, Yimidhirr-Yalanji-Yidinic)		[Patz 2002: 62]
maja	balka-n-balka -wa-y	warru-warru-nji
boss(master).abs(s)	talk-n -red-recp-npst	yg.man-red-com:pt
‘The boss is talking with the young men.’

Delimitation/diminution describes cases of DR that depict a situation performed in a limited or short time frame. This is the case for the X-one-X construction of Mandarin Chinese:

(40)

Mandarin Chinese (Sino-Tibetan, Sinitic)		[Li and Thompson 1981: 31]
nǐ	wèn-yi-wen	tǎ
you	ask-one-ask	3sg
‘You ask him/her a little.’

Distributivity refers to the case when a property or an event is distributed over different entities or over time, and also the case in which plural entities are conceived individually.

(41)

Mparntwe Arrernte (Pama-Nyungan, Arandic-Thura-Yura)			[Wilkins 1989: 345]
Nwerne	lesson	mape	kaltye-le-nthe-me
1pl.a	lesson	pl(grp)	knowledge-loc/inst-give-npp
nyente-me-nyente.
one- uq- one
‘We teach the lessons one-by-one. [i.e., in order, not so much one at a time]’

In Pacoh, verbs can undergo DR in order to refer to the situation denoted by the verb itself without any specific temporal or aspectual information. In other words, DR expresses the situation in general and, thus, we call this function generality.

(42)

Pacoh (Austroasiatic, Katuic)

[Alves 2006: 37]

a.

ca:	→	ca:-ʔi-ca:
‘to eat’		‘to eat in general’

b.

taʔ	→	taʔ-ʔi-taʔ
‘to work’		‘to work in general’

Greater plurality is a particular case of plurality that “typically implies an excessive number, sometimes called ‘plural of abundance’” (Corbett 2000: 30). This is, for example, the case of a DR pattern in Maithili, as exemplified in (43).

(43)

Maithili (Indo-European, Indo-Iranian)	[Yadav 1996: 92]
gam-ək	gam
village-gen	village
‘A lot of villages’

Following Comrie (1976: 27–28), we define habituality as “a situation which is characteristic of an extended period of time, so extended in fact that the situation referred to is viewed not as an incidental property of the moment but, precisely, as a characteristic feature of a whole period.” The DR pattern in (44) from Standard Malay conveys precisely this kind of situation.

(44)

Standard Malay (Austronesian, Malayo-Polynesian)				[Hudson 1995: 96]
ayah	karang-men-garang	di	tengah	malam
father	write-meN-write	at	point_of_time	night
‘Father writes at nights.’

In (45), we have a situation in which the term jɘl ‘water’ in Maithili, when reduplicated with the interposition of the element e, encodes that only water (and nothing else) is present in the pond possessed by the speaker (and her/his associates). Since this function serves to contrastively identify the referent denoted by the base, we decided to call it identificational focus.

(45)

Maithili (Indo-European, Indo-Iranian) [Asad 2015: 50]
hɘmɘra	pokhɘr	me	jɘl	e	jɘl	cʰɘɪ
our	pond	in	water	e	water	aux.prs
‘There is nothing in our pond, but water.’

Inchoativity (or inceptivity) conveys when an “action or event begins” (Bybee et al. 1994: 318). The Chácobo example in (46) illustrates this kind of situation.

(46)

Chácobo (Pano-Tacanan, Panoan)	[Tallman 2018: 880]
i-ʔ-i =baʔina=kan=á=ka	bɨʂpa	tsi	kiá	hawɨ
do∼ ep ∼do=all/each_day=3pl=nmlz:p=rel	skinny	p5	rep	3sg:gen
bakɨ́	ʂobo	nia=baya=ʔita=ʔá=k	bɨpaná	ha
child	house	leave=do&go:tr/pl=recp=nmlz:p=rel	shed	3
his=baya=ita=ʔá=na
see=do&go:tr/pl=nmlz:p=ep
‘They started to work all day and they left him thin when his son left the shed.’

In Punjabi, the reduplication of the element koii ‘some’ with the intervening negative particle ba expresses indefiniteness, since the head noun it modifies is perceived as indefinite rather than specific.

(47)

Punjabi (Indo-European, Indo-Iranian)					[Bhatia 1993: 220]
ó	koii	ba	koii	kataab	khariidegaa
he	some	neg	some	book	buy-fut.3sg.m
‘He will buy some book or other.’

As noted, we found several cases in which DR expresses intensification. This function refers to the case in which the lexical meaning of the base that undergoes reduplication is in some way augmented, that is, intensified.

(48)

Alamblak (Sepik, Sepik Hill)	[Bruce 1984: 165 quoted in Rubino 2005: 17]
hingna- marña-ba-marña -më-r
work-straight- lig -straight-rm.pst-3sg.m
‘He worked very well’

The term pluractionality has become more and more common in (typological) linguistics in the last few years. It refers to “a morphological modification of the verb (or a pair of semantically related verbs) that primarily conveys a plurality of situations that involves a repetition through time, space and/or participants” (Mattiola 2019: 164). An example of this function is reported in (49).

(49)

Chácobo (Pano-Tacanan, Panoan)		[Tallman 2018: 511]
bi=’∼bi=’	rabi=’	wa=kɨ
grab= lnk ∼grab= lnk	Rabi=erg	tr=dec:p
‘Rabi grabbed it multiple times/multiple things’

The term (additive) plurality represents the most basic plural function: it “refers to more than one real world entity” (Corbett 2000: 20). However, additive plural can have different values depending on the system it belongs to (if a language has the dual, then plural means ‘three or more’, and so on). Majhi displays a DR pattern carrying this function:

(50)

Majhi (Indo-European, Indo-Iranian)		[Dhakal 2014: 21]
bari-ka	sori-ne/sori	rəila
field-loc	pig- em /pig	be.3pl.pst
‘There were pigs in the field.’

The function ‘precision’ refers to the case in which DR points to something specific, like a localization, as exemplified in (51) for Punjabi.

(51)

Punjabi (Indo-European, Indo-Iranian)						[Bhatia 1993: 205]
ó	ne	báár	de	báár	Taal	dittaa
he	erg	outside	gen.m.sg.obl	outside	put_off	give-pst.m.sg
‘He put (me) off right from outside’

In Uduk, adjectives can be reduplicated with an intervening stative marker in order to make them function as a predicate. Thus, we call this function predicativization.

(52)

Uduk (Koman, Central Koman)				[Killian 2015: 386]
dhàlì	‘bór-á	‘bōr	làkíīn	‘dīshán̄	shíʔ-á	shīʔ
and	good- stv	red	but	now	bad- stv	red
gòmà	mùstákbàl	mò
for.cl2	future	mo
‘it was good but now it is bad for the future.’

Reciprocity obviously refers to the case in which the participants perform the same situation on each other, i.e., reciprocally. See (53) from Standard Malay:

(53)

Standard Malay (Austronesian, Malayo-Polynesian)	[Hudson 1995: 92]
kami	bantah-mem-bantah
we	object-meN-object
‘We objected each other’s (opinions).’

Finally, we label universal quantification the case in which DR patterns express all the referents denoted by the base, i.e., its entire denotation.

(54)

Kisi (Atlantic-Congo, Mel)		[Childs 1995: 192]
sɔ̀	→	sɔ̀-ó-sɔ̀
chicken		chicken-ó-chicken
‘chicken’		‘every chicken’

6 Typological generalizations and discussion

In Section 5, we described the DR patterns we identified in our language sample according to the formal and functional parameters proposed in Section 4. Here, we discuss the overall picture that emerges from our data and discuss our findings with particular reference to their typological relevance.

As mentioned in Section 5.1.2, we found quite a high number of possible interposing elements, including meaningless items, such as epenthetic elements, and meaningful items, both free and bounded. The most common interposing elements turned out to be linkers (‘and’), adpositions (especially postpositions), case markers, and negative elements.

As for major lexical categories, the most frequent bases within our dataset are – quite expectedly – verbs and nouns, followed by adjectives, adverbs and numerals (cf. Section 5.1.3).

As pointed out in Section 5.2, the many functions conveyed by DR are those usually expressed by reduplicative patterns in the world’s languages. Some of them are more typically associated with certain lexical categories (see Figure 6) and this is quite in line with expectations: (i) adjectives tend to express intensification and distributivity; (ii) adverbs tend to express intensification; (iii) nouns tend to convey delimitation/diminution, distributivity, identificational focus, and greater plurality; and, finally, (iv) verbs tend to express continuativity, pluractionality, and habituality.

Figure 6:

Number of functions for each lexical category of the base.

Interestingly, (additive) plurality (with nouns) is expressed only once. This is also not so unexpected. Recent works have shown that nominal reduplication is not so much related to plurality as previously thought: in Dryer (2013) only 8 languages out of 1,066 display full reduplication encoding plurality; and in Ivani and Zakharko’s (2019) nominal number database only 6 languages out of 237 employ some kind of reduplicative patterns to express plurality (cf. also Mattiola and Barotto to appear, which goes in the same direction).

Before collecting the data for our investigation, we expected to find some correlation between the nature of the interposing element and the value carried by the DR pattern. This was not the case: the type of interposing element does not seem to play a clear role in the functional interpretation of DR patterns. We found different kinds of interposing elements per function, as Figure 7 details. However, the limited number of constructions under observation does not allow for any statistical statement.

Figure 7:

Number of functions for each type of interposing element.

Overall, we can say that the lexical category of the base seems to play a greater role in the semantic interpretation of DR than the interposing element. In other words, discontinuously reduplicated verbs tend to express the same functions generally conveyed by contiguously reduplicated verbs, and the same applies to other major lexical categories. Therefore, ultimately, DR patterns seem to behave like standard reduplicative patterns, irrespective of the presence of the interposing element and its nature. This conclusion is reminiscent of what Stolz (2009) claims about syndetic versus asyndetic reduplication in European languages: the two strategies seem to be functionally equivalent.

Finally, probably our most important finding is related to the parameter ‘Presence or absence of other reduplicative patterns’ (Section 5.1.4): in all languages displaying DR, contiguous reduplication is also attested. Hence, there seems to exist an implication for what concerns the relationship between contiguous reduplication and DR. More specifically, the presence of a DR pattern implies the co-presence, in the same language, of a ‘standard’ (contiguous) reduplicative pattern, but the presence of the latter does not imply the co-presence of DR. This implication can be formalized in the following hierarchy:

contiguous reduplication > discontinuous reduplication.

This hierarchy is quite expected if we consider DR as a sub-type of ‘standard’ reduplication, but it is still relevant from a typological perspective and explains some properties of DR, such as the fact that DR conveys a subset of the values conveyed by standard reduplication in the world’s languages. Note that, in this subset, additive plurality is largely under-represented (we found only one occurrence), which is in line with recent investigations that – as mentioned above – challenge the central role of this function within the functional domain of (nominal) reduplication.

7 DR patterns in Italian: a case study

In this section, we offer a description of Italian DR patterns based on native-speaker intuition and (cursory) corpus search using itTenTen16 (or Italian Web 2016), a very large (nearly 5-billion token) POS-tagged web corpus available on Sketch Engine (https://www.sketchengine.eu/). It is important to stress that what follows is not a comprehensive, quantitative account (which would require a separate study) but rather an initial, qualitative description, which aims at increasing our understanding of DR.

Another reason for adding this case study is to show how the picture you gain may differ when you move from a large-scale investigation based on descriptive grammars to a specific-language investigation which goes beyond ‘scratching the surface’. Italian is actually one of the languages included in our sample and for which we did not find any kind of information about DR patterns within the grammatical description we selected for the typological investigation: Maiden and Robustelli (2007). The authors do not mention any DR pattern for Italian, either morphological or syntactic. However, they do give information about contiguous reduplicative patterns. According to Maiden and Robustelli (2007: 355), Italian displays the repetition of adjectives (55a) and adverbs (55b) to express intensification, but they add that “[r]eduplication is sometimes encountered even in nouns”, in examples like (55c):^[9]

However, we would expect Italian to display some DR pattern, especially if we consider that Stolz (2009) classifies it as a “middle-degree syndesis” language, based on parallel corpora investigation.

In order to uncover DR patterns in a bottom-up fashion, we devised the following procedure and we applied it to the itTenTen16 corpus: we designed an advanced search (regular expression) to extract three-item strings made of two identical items X separated by a third item Y. Since major lexical categories are mainly involved in DR cross-linguistically (see Section 5.1.3), we restricted our analysis to these: hence, we asked for strings where the item X was tagged as noun, verb, adjective or adverb. As for Y, we excluded items tagged as punctuation or non-linguistic material. This query gave 1,698,429 hits. We then generated a frequency list of POS-patterns and we manually explored the 100 top results.

Perhaps not unexpectedly, the top-ranked (valid) POS-patterns turned out to be [N Prep N] and [N Conj N]. The latter is typically instantiated by expressions where the conjunction e ‘and’ conjoins two plural nouns, like decine e decine ‘tens and tens’, anni e anni ‘years and years’ or pagine e pagine ‘pages and pages’,^[10] all conveying greater plurality (see for instance 56).^[11]

The [N Prep N] pattern, instead, is extremely variegated in both form and meaning: many different prepositions participate in this construction (e.g., su ‘on’, per ‘for’, a ‘at’, dopo ‘after’, etc.), which contribute to the semantics of the whole expression together with the type of noun involved. For instance, DRs with dopo ‘after’, a ‘at’, and per ‘for’ mainly convey distributivity (cf. e.g., 57) whereas those with su ‘on’ mainly express greater plurality (58), although some expressions are more lexicalized and hence more idiosyncratic, e.g., distributive porta a porta ‘door by door’ (lit. door at door) versus faccia a faccia ‘face to face, in front of each other’ (lit. face at face); or greater plural colpi su colpi ‘many many blows’ (lit. blows on blows) versus colpo su colpo ‘blow for blow’ (lit. blow on blow).

The [N Prep.Art N] pattern is also attested, but apparently valid examples are limited, like mano nella mano ‘hand in hand’ (lit. hand in.the hand), or strings like secoli dei secoli (lit. centuries of.the centuries), which are part of a larger prepositional phrase: nei secoli dei secoli (lit. in.the centuries of.the centuries) or per (tutti) i secoli dei secoli (lit. for (all) the centuries of.the centuries), both meaning ‘forever and ever’. The latter situation, which depicts a partial string preceded by a preposition, is quite common also in the [N Prep N] data. See for instance casa in casa (lit. house in house), which is actually used as di casa in casa (lit. of house in house) ‘from house to house’, again with distributive reading:

One last case of DR we found with reduplicated nominals is exemplified in (60):

Here the two identical nouns are separated by the negative marker non ‘not’ and the whole construction conveys a function of approximation: tessuto non tessuto is something that looks like a fabric but is not properly a fabric because it is not obtained by weaving or knitting. Other (recently coined) instances of this pattern would be: bollito non bollito ‘meat cooked in a low temperature water bath within vacuum sealed bags’ (lit. boiled not boiled); sapone non sapone ‘cleanser that does not contain soap’ (lit. soap not soap); talco non talco ‘talcum-free product that serves the same function of talcum powder’ (lit. talcum_powder not talcum_powder); colore non colore ‘color that is not a proper color, like black or white’ (lit. color not color) (cf. Di Donato and Masini 2022).

Let us now turn to verbs. With the exception of a couple of lexicalized and idiosyncratic instances where the two copies are separated by the negation marker non ‘not’ (vedo non vedo ‘see-through effect’, lit. see.1sg.prs.ind not see.1sg.prs.ind) and the adverb/comparative marker come ‘like/as, how’ (vada come vada ‘come what may, whatever happens happens’, lit. go.3sg.prs.sbjv not go.3sg.prs.sbjv), discontinuously reduplicated verbs are typically found in the [V Conj V] pattern (with the connective e ‘and’), which conveys continuativity:

Adjectives and especially adverbs are also used as bases in DR patterns conveying intensification. Adjectives are mostly found separated by the adversative conjunction ma ‘but’ (cf. 62). The same intervening element is found with adverbs (e.g., La situazione del piccolo George è molto ma molto critica ‘The situation of little George is very, very [lit. very but very] critical’), which are however coinjoined also by the (polyfunctional) conjunction che ‘that/than’ (e.g., peggio che peggio ‘even worse’, lit. worse that/than worse; forse che forse ‘perhaps’, lit. maybe that/than maybe), the adverb/comparative marker come ‘like/as, how’ (e.g., ora come ora ‘nowadays’, lit. now like now), and by prepositions: see the examples in (63), but also cases like di tanto in tanto ‘every now and then’ (lit. of much in much), with a double preposition.

What this brief overview shows is the wealth of constructions ascribable to DR we could extract from corpora. To conclude this section, we would like to discuss two further instances of DR that did not emerge from our corpus search. These two remaining patterns were recently discussed in the literature and both present properties that set them apart from all the cases mentioned so far, with respect to both our cross-linguistic investigation and our intra-linguistic case-study.

The first construction, studied by Masini and Iacobini (2018), consists in the repetition of a numeral ‘around’ the nominal it modifies. This [Num N Num] pattern is quite peculiar because of the presence of a noun as interposing element, a category which is absent from our typological investigation (cf. Section 5.1.2). See the examples in (64) (from Masini and Iacobini 2018: 103):

In all these examples the [Num N Num] construction is used to express precision, i.e., ‘exactly two days’ in (64a) and ‘three and no more drops’ in (64b). However, according to the authors, an additional “subjective, evaluative meaning” can be identified in many cases. In (64a), for instance, the two days spent in jail are clearly judged by the speaker as not enough for that situation (‘only two days’), whereas the seven days of mourning in (64c) are obviously believed to be too many. Thus, these two expressions add a paucity and excess reading, respectively, in addition to precision (Masini and Iacobini 2018: 104), which might lead to consider this construction as a kind of mirative, as an anonymous reviewer suggests. Even though the pattern is not very common in Italian, it seems to have some degree of productivity. We mainly find small numbers (due ‘two’, tre ‘three’, and also uno ‘one’), but we do find cases with higher numbers too, as shown in (65a) (Masini and Iacobini 2018: 105). In addition, we found full phrases instead of single nouns enclosed within the two identical numerals (65b).

The second construction which is worth mentioning is a pattern of DR that features an antonymous pair of adverbs meaning roughly ‘here and there’ and that can be schematized as: [X Adv₁ X Adv₂], where Adv₁ and Adv₂ are antonymous spatial adverbs like di qua ‘here’ – di là ‘there’.^[12] This construction, which has been extensively described by Masini and Mattiola (2022), displays quite a lot of variability since it can host different reduplicated bases in terms of both lexical category (compare the verb in 66a with the noun in 66b) and complexity: compare again the data in (66), which show the reduplication of a single word (66a), a phrase (66b), and a clause (66c).

From a functional point of view, Masini and Mattiola (2022) identify four specific functions for the [X Adv₁ X Adv₂] in Italian, all closely related to a ‘plurality/variety’ macrofunction, namely: distributivity (67), related variety (68), dispersion (69), and – to a much minor extent (one example only) – additive (standard) plurality (70).

Table 1 summarizes the results of this section by listing the DR constructions we could identify for the Italian language. Obviously, as already noted, these are only partial results based on a preliminary analysis of the data. Furthermore, we are not in a position to give details about either the precise function or the productivity of each construction, which would require a dedicated study.

In conclusion, what this cursory investigation reveals is that our typological findings seem to hold also at the language-specific level: all major lexical categories are involved in DR in Italian, although nouns seem to be the most active word class in this case; also, most of the intervening elements attested in Italian DR patterns also figure in our cross-linguistic investigation, viz. prepositions, conjunctions, negation markers. But more importantly, what emerges is that DR patterns are under-recognized and under-described even in very well-documented languages like Italian. If we rely on the descriptive grammar used for our investigation (which, by the way, is an extremely rich and informed resource), we must conclude that Italian lacks DR. But if we resort to naturally occurring data from corpora (combined with native speakers’ intuition and judgements), we can detect previously unnoticed data: far from being absent, DR is attested in Italian by a wide range of possible formal strategies.

Now, often typological investigations are based solely on the analysis of grammatical descriptions, and sometimes this is the only option available. However, in principle, typology should also be grounded in corpus-based analyses, like is being done with most language-specific studies nowadays, given the growing availability of corpora. This would have the advantage of unveiling a much broader linguistic variety and, thus, making typology much more empirically solid. Needless to say, a ‘discourse(-sensitive) typology’ is very difficult to implement and is simply not viable at present for a very high number of languages, for which we are still lacking corpora (especially of spoken language, where innovation works at its best). However, some current projects – like Multi-CAST (cf. Haig and Schnell 2016) or DoReCo (cf. http://doreco.info/) – are going exactly in this direction. We believe this is precisely the path typology should follow in the next future in order to continue pursuing its basic purpose of unveiling and explaining linguistic diversity.

8 Final remarks

In this paper, we offered what we believe to be the first typology of discontinuous reduplication in the world’s languages. After defining and delimiting our object of analysis, we presented the main formal and functional properties of DR patterns which emerged from the investigation of the languages in our sample.

We first provided a description of the main formal properties of DR. What emerged is that several different items can function as interposing element: from adpositions to stative markers, from linkers to exclusive particles. All major lexical categories can undergo DR (e.g., nouns, verbs, adjectives, adverbs), but also some minor categories (e.g., numerals, quantifiers, adpositions). From a functional point of view, DR can express a range of functions that are typically associated also with contiguous reduplication (e.g., distributivity, intensification, habituality, pluractionality, etc.).

We then drew some typological generalizations from our data. Unexpectedly, the nature of the interposing element does not seem to play a clear role in the functional interpretation of the overall pattern, which conversely seems to stem more from the nature of the base (e.g., discontinuously reduplicated verbs basically express the same functions expressed by contiguously reduplicated verbs).

The most interesting, but in some way expected, result is that we identified an implication between contiguous and discontinuous reduplication: a language which displays DR patterns will also have some other kind of contiguous reduplicative pattern.

Finally, we took a closer look at DR patterns in Italian. This language-specific investigation generally confirmed our cross-linguistic findings, but also revealed that corpus-based analyses can reveal several additional DR patterns that simply cannot be detected in large-scale cross-linguistic studies, at least for the time being. A conclusion that can be drawn is that, ideally, we need more detailed studies on single languages, not just for DR but for typological investigation in general.

Obviously, we make no claim of completeness. Rather, this work was conceived as a starting point for a wider exploration, which should ideally use a more comprehensive language sample and extend the investigation to corpora of naturally occurring language data. This would possibly allow: (i) to confirm (or not) that it is the lexical category (and not the intervening element) that plays the major role in determining the function of the pattern (and why); (ii) to confirm (or not) the proposed contiguous reduplication > discontinuous reduplication hierarchy and to inquire into its possible explanations, including the diachronic connection between the two.

Abbreviations^[14]

1, 2, 3

1st, 2nd, 3rd person

3>3

verbal grammatical third to third person ‘she>him’

-

morpheme boundary

∼

reduplicative boundary

<…>

infix

??

no glosses identified

a

subject of transitive

abs

absolutive

acc

accusative

ad2

aspect directional 2

adj

adjectivizer

adv

adverb

all/each_day

all/each day

art

article

attr

attributive

aux

auxiliary

caus

causative

cl(1/2/9)

noun class/class gender 1/2/9

com

comitative

comp

comparative

compl

complementizer

conj

conjunction

cop

copula

dat

dative

dec

declarative (mood clitic)

def

definite

dem

demonstrative

dem.med

medial demonstrative

dem.prox

proximate demonstrative

dim

diminutive

do&go

associated motion clitic

dr

discontinuous reduplication

ds

different subject

em

emphatic

enc

enclitic

ep

epenthetic

erg

ergative

exc.part

exclusive particle

fut

future

gen

genitive

gen.evt

generic event

idph

ideophone

imp

imperative

inc

incomplete aspect

ind

indicative

inst

instrumental

intens

intensifier

inter

interrogative

ipfv

imperfective

lig

ligature element

lnk

linker

loc

locative

m

masculine

n

noun

neg

negative

nmlz:p

past tense nominalizer

nom

nominative

npp

non-past progressive

npst

non-past

num

numeral

obl

oblique

p

object of monotransitive predicate

pc

past completive

pfv

perfective

pl

plural

pl(grp)

plural in group

prep

preposition

pret

preterit

prog

progressive

prox

proximal

prs

present

pt

potent case inflection

purp

purposive

recp

reciprocal

red

reduplication

rel

relative

rep

reportative

rfl

reflexive

rm.pst

remote past

s

intransitive subject

sbj

subject

sbjv

subjunctive

sg

singular

sim

simultaneous subordinate

ss

sentence suffix

Ssub

subordinate temporal tense

stv

stative

top

topic

tr

transitive

uq

unified quantity