Repetitive prefix in Agul: Morphological copy from a closely related language

Introduction

The last decade has seen a rise of interest towards contact-induced phenomena, including morphological borrowing (see e.g. Gardani, 2008; Gardani, Arkadiev, & Amiridze, 2015; Johanson & Robbeets, 2012; Matras & Sakel, 2007; Stolz, Vanhove, Otsuka, & Urdze, 2012; Wiemer, Hansen, & Wälchli, 2012). Whereas the most discussed cases of contact morphology involve borrowing from a socially dominant but genetically unrelated language, the occurrence of similar phenomena among distantly or even closely related languages cannot be excluded (on the latter, see in particular Epps, Huehnergard, & Pat-El, 2013). The aim of this paper is to present evidence for a contact-induced development belonging to the ‘closely related’ end of the continuum, namely the rise of the repetitive prefix in Agul under the influence of Lezgian, a dominant language of the area where southern dialects of Agul are spoken.

Both Agul and Lezgian belong to the Lezgic branch of the East Caucasian (Nakh-Daghestanian) family. Agul is one of the minor languages of Russia. It is spoken in about 20 mountain villages in southern Daghestan, and by those resettled on the lowlands; according to the 2010 census, there are altogether circa 29,300 speakers of Agul in Russia. Administratively, the Agul villages in the mountains are divided between the Agulskiy district and the Kurakhskiy district. There are five Agul villages in the north-eastern part of the Kurakhskiy district, which is mainly inhabited by Lezgians. From a dialectal point of view, one of the Agul villages of the Kurakhskiy district, namely Khpyuk (Huppuq’) represents a distinct dialect, while the four other villages (Usug, Kvardal, Khveredzh and Ukuz) together with the two neighbouring villages of the Agulskiy district (Richa and Bedyuk) belong to the Keren dialect of Agul. The Agul dialects spoken in the Agulskiy district include the Central (or Tpig) dialect, and the Tsirkhe, Burkikhan and Fite dialects spoken in one village each, and also the Qushan dialect, which is hardly intelligible to speakers of other varieties.

Lezgian is spoken in a much more vast territory than Agul, including several districts of southern Daghestan and also several districts of northern Azerbaijan; the total number of speakers, according to Ethnologue, amounts to almost 650,000. ¹ In the Kurakhskiy district, where the Agul-Lezgian linguistic border can be drawn, the Kurakh dialect of Lezgian is spoken. The Keren Agul villages of the district as well as Khpyuk have been in tight contact with the Lezgian-speaking villages Gelkhen, Ursun and Shimikhyur for a long time (see Figure 1 for a map of the area around the border between the two districts). Most speakers of the southern Agul dialects have a good command of Lezgian, which is the dominant language of the area and has a relatively high prestige, since the main road to the lowlands goes through the Lezgian territory, Lezgian is spoken in the administrative centre of the district (Kurakh) and shopping mainly takes place in Kurakh and Kasumkent, another big Lezgian village that is the centre of the next district on the way to the lowlands. In the second half of the 20th century, Lezgian was even taught in schools as a ‘mother tongue’ for some time, and intermarriages with Lezgian women continue to be frequent.

OPEN IN VIEWER

Figure 1.

South of the Agulskiy district and north-east of the Kurakhskiy district. ²

Other neighbours of the Aguls in the mountains are the Dargins (Dargwa) to the north and the Tabasarans to the east. Whereas Dargwa forms a separate branch of the East Caucasian family, Tabasaran and Lezgian are the languages most closely related to Agul; together the three comprise the East Lezgic subgroup of the Lezgic branch. Like East Caucasian languages in general, the East Lezgic languages have ergative case alignment, head-final word order with a basic subject–verb/subject–object–verb (SV/SOV) pattern, rich case systems and elaborate verbal paradigms. In particular, the case system typically includes a series of locative forms, comprising two slots: the localization suffix encodes the position of the figure with respect to the landmark, while the orientation suffix (which may be null) points at the direction of movement, or the absence of motion (see Daniel & Ganenkov, 2009; Kibrik, 2003 for details). Another important feature of the Lezgic languages is the use of prefixal morphemes, in particular locative preverbs, whose system is in some respect symmetric to that of the nominal locative forms.

The present paper focuses on one of the prefixes attested in Agul, which has a restricted distribution across the dialects – the repetitive prefix q(V)- meaning ‘again’, or ‘back’ in the case of motion or transfer, for example q-agune ‘saw again’ from agune ‘saw’, or qa-tin ‘give back!’ from tin ‘give!’. In the next section, the system of verbal prefixes of Agul will be briefly described. The two subsequent sections will discuss the morphological properties and the functions of the repetitive prefix in more detail. I will then turn to the hypothesis about the origin of the Agul repetitive prefix as a morphological copy from Lezgian. The final section presents the conclusion and directions for future study.

Although the existence of the repetitive prefix in southern Agul dialects has been previously acknowledged, for example by Magometov (1970, pp. 163–164), and its probable contact-induced nature was mentioned by Sulejmanov (1993, p. 162), Maisak and Merdanova (2002, pp. 260–261) and Maisak and Ganenkov (2016, p. 3590), no detailed description of this marker has been provided so far. The present account is mainly based on data from the Huppuq’ dialect (spoken in the village of Khpyuk), in particular on a corpus of spontaneous oral narratives from Huppuq’ and other Agul dialects, which was collected in the 2000s as part of the Agul Documentation Project. ³

Verbal prefixes in Agul

In this section, I will show that within the system of verbal prefixation in Agul, the repetitive prefix occupies a special position, being paradigmatically distinct from the other prefixes and possessing its own morphological slot.

There are three sets of prefixal morphemes in Agul: locative prefixes (which include two series of markers), negative prefixes (two morphemes) and the repetitive prefix, which will be the main focus of the paper. While the locative prefixes are clearly derivational and the negative prefixes are clearly inflectional, the repetitive marker is somewhat intermediate on the derivation–inflection continuum. All the prefixes of the Huppuq’ dialect (the main allomorphs only) are listed in Table 1.

OPEN IN VIEWER

Table 1.

Verbal prefixes in Huppuq’ Agul.

Repetitive	Negative	Locative
		Localization	Orientation
q- [re]	d- [neg] ‘negative’, m- [proh] ‘prohibitive’	ʔ- (∅-) [in] ‘inside a container’, ʕ- [inter] ‘inside a liquid or substance’, al- [super] ‘on a horizontal surface’ f- [apud] ‘near a landmark’, h- [ante] ‘in front of a landmark’, q- [post] ‘behind a landmark’, k- [sub/cont] ‘below’ / ‘attached to a landmark’	-č- [lat] ‘motion towards’, -atː- [elat] ‘motion from’, -ʁ- [up] ‘motion upward’, -a- (-da-) [down] ‘motion downward’	(verb root)

Locative prefixes

Although locative prefixes (also known as ‘preverbs’) are not found in all the branches of the East Caucasian family, they are certainly one of the hallmarks of the Lezgic branch. Every Lezgic language possesses a set of locative prefixes, some of which have the same form and meaning as locative case markers. It is thus plausible to assume that (some of) the locative prefixes and (some of) the locative case forms have a common historical origin in locative adverbs. In one morphosyntactic context, namely as postpositional nominal modifiers, such adverbs fused with nominals, finally becoming suffixes; in another configuration, preverbal locative markers ended up as bound prefixes (see also Haspelmath, 1993, p. 169). At least the following locative prefixes can be traced back to the Proto-Lezgic stage, according to Alekseev (1985, pp. 117–121): *al- ‘on, above’; *ʔ- ‘inside’; *h- ‘in front’; *k- ‘in contact’; *ʟː- ‘under’; *ɬːʷ- ‘near’; *s- ‘down, below’; *q- ‘behind’; *qˤ- ‘between, in a mass’; *qː- ‘out’ (see also Schulze, 1982, pp. 264–266). For the subsequent discussion, it is important to note that the prefix *q- ‘behind’ is also found among these ancient locative morphemes.

Among the Lezgic languages, Agul possesses one of the richest systems of locative prefixation, comprising two series of markers and two slots for them. The first slot is reserved for localization markers that specify a particular spatial domain where a figure is located. In the Huppuq’ dialect, there are seven localization prefixes, which express meanings such as ‘inside (a container)’, ‘in front of’, ‘behind’, ‘on a horizontal surface’, etc. – see the full list in Table 1. Prefixes of the second slot, which is closer to the verb root, mark orientation and indicate whether a figure moves in a particular direction, or rests at a spatial domain defined by the localization marker. There are four orientation markers, which distinguish between the lative (‘motion towards’) versus elative (‘motion from’) meanings on one hand, and ‘motion upwards’ versus ‘motion downwards’, on the other. The second slot may be left empty, though, so there are verbs with only a localization prefix and no orientation prefix. The absence of the orientation markers usually triggers the lative reading with dynamic verbs, and the essive reading (‘no motion’) with statives (‘to be’ and ‘to remain’).

Prefixal verbs can be derived both from roots that are also used without locative prefixes and from ‘bound’ roots that do not occur outside prefixal derivatives. The two stative roots ‘to be, be located’ and ‘to still be, remain’ belong to the latter type: with each of the two roots seven different verbs can be formed, which differ only with respect to the localization prefix, for example aa ‘is inside (a container)’, ʕaa ‘is inside (a substance)’, aldea ‘is on’, haa ‘is in front’, etc., and amea ‘remains inside (a container)’, ʕamea ‘remains inside (a substance)’, almea ‘remains on’, hamea ‘remains in front’, etc. Other bound roots with many prefixal derivatives include -arx- ‘to get to, find oneself in’, -uč’- ‘to get into (with effort), climb’, -aq- ‘to pour, scatter’, -ik’- ‘to put (with effort), thrust’, -ix- ‘to put’, -dark- ‘to turn, spin’, among others. On the other hand, verb roots ʕʷ- ‘to go/come’ or du(ʔ)- ‘to pull, drag’ occur without prefixes as well, compare forms of the unprefixed verb duaa ‘pulls, drags’ and its prefixed derivatives ʔa-ča-dʷaa ‘pulls into (a container)’, ʕa-ča-dʷaa ‘pulls into (a mass or liquid)’, al-ča-dʷaa ‘pulls on (e.g. clothes)’, ʔatːʷaa (< *ʔ-atː-dʷaa) ‘pulls from inside’, ʔa-ʁa-dʷaa ‘pulls up from inside’, ʔa-da-dʷaa ‘pulls down from inside’, etc.

These bound verb roots are attested with the largest number of derivatives. Given the seven localization prefixes and the five possible fillers for the orientation slot (including the absence of a marker), we can expect to have as many as 35 derivatives from one verb root. However, locative prefixation is not that productive: the combinability of prefixes with verbal roots is restricted, and there are no verb roots for which all of the 35 logically possible combinations with locative prefixes exist (combinations including the ‘up’ and ‘down’ orientations are especially rare). The root -ix- ‘to put’ heads the list with 32 derivatives, less than a dozen other roots have more than ten derivatives each, and only about 30 roots have more than one prefixed derivative. Table 2 shows some of the derivatives with the root -ix- ‘to put’, illustrating the three localizations ‘in’, ‘super’ and ‘post’ and all possible combinations with orientation markers (note that the root changes to -jx- after a stressed syllable). For presentation purposes, in Table 2 as well as in other examples in this section, derivatives with the ‘in’ prefix are given with the initial glottal stop, which in Agul is normally not pronounced in the word-initial position. On the whole, about 350 dynamic prefixed verbs have been identified in the Huppuq’ dialect.

OPEN IN VIEWER

Table 2.

Some prefixal derivatives with the root -ix- ‘to put’.

Localization/orientation	In	Super	Post
(no)	ʔ-ix- ‘to put inside a container’	al-ix- ‘to put on’, ‘to build’, ‘to appoint’	q-ixas ‘to put close to’, ‘to connect, hook up (lights, electricity)’
lat	ʔa-č-ix- ‘to bring inside’, ‘to hit, slap’	al-č-ix- ‘to put on the upper surface’	qi-č-ixas ‘to put behind’
elat	ʔ-atː-x- ‘to take from inside’	al-atː-ix- ‘to take from the upper surface’	q-atː-ixas ‘to take from behind’, ‘to betroth’
up	ʔa-ʁ-ix- ‘to put up and inside’	al-ʁ-ix- ‘to put above oneself’	–
down	ʔa-da-jx- ‘to throw down from inside’	al-a-jx- ‘to throw down from the upper surface’	q-a-jxas ‘to throw down from behind’

Locative preverbs in Agul do not change the aspectual characteristics of the verb – thus, Agul, as with most other languages of the family, does not belong to the ‘club’ of Caucasian languages that employ prefixation as a perfectivizing aspectual device (see Arkadiev, 2015, for a detailed treatment of the latter phenomenon). At the same time, many prefixal verbs in Agul have idiomatic meanings, with the underlying locative metaphor more or less clear in most cases. Some examples of this kind can be seen in Table 2, for example ʔa-č-ix- with the literal meaning ‘to put inside’ can also refer to hitting, and q-ixas with the localization prefix ‘post’ serves to describe hooking up lights or electricity. The stative verb qaa ‘is behind’ with the same prefix is the main means to express permanent possession, while another stative derivative faa ‘is near’ with the localization prefix ‘apud’ is used as a predicate in a temporary possessive construction. Two more illustrations are presented in (1) and (2). In (1), the verb alčarx- meaning ‘to appear, find oneself on’ with the prefixes ‘super’ and ‘lat’ is used with the meaning ‘to meet’ (literally, ‘a man appeared on me’). In (2), the verb quχ- ‘to believe, trust’ is a derivative with the prefix ‘post’ from a verb root that seems isolated; this verb does not have any locative meanings, but its connection to the localization ‘behind’ is obvious from the obligatory encoding of the person whom one trusts in the post-essive case (cf. wa-q ‘you-post’).

(1) za-l   sa  idemi alčarx.u-ne

 I-super one man  meet.pf-aor

 ‘I met a man.’

(2) zun wa-q    quχ.a-dawa

 I  you-post  believe.ipf-prs:neg

 ‘I don’t believe you.’

Indeed, the historical affinity of verbal prefixes and nominal locative cases is still clear on a synchronic level (even despite the semantic changes in prefixed verbs), due to a frequent congruence between the two sets of markers: verbs with a particular locative prefix typically go together with dependent noun phrases with the same localization marker. Thus, in (3) the prefixal combination ‘sub/cont-elat’ corresponds to the sub/cont-elative case form of ‘suffering’, and the same can be said about the ‘ante-lat’ prefix and the ante-lative case of ‘that girl’ in (4). As with the verbal locative prefixes, locative case forms also have a bipartite structure and include two (suffixal) slots, one for localization and one for orientation. Note, however, that while localization markers in the nominal paradigm are historically identical to verb prefixes, nominal orientation markers have a different form and a different set of values (only ‘lat’ and ‘elat’, but no ‘up’ and ‘down’).

(3) amma te ruš (…)  k-etː- arx.a-s-tːawa

 but  dem daughter {sub/cont-elat}-get.to.ipf-inf-cop:neg

 ʡazab.i- k-as

 suffering-sub/cont-elat

 ‘But your daughter… won’t get out of trouble.’

(4) me  t’ubal  ruš.a- h-tːi    hi-č- ik’.i-guna…

 dem ring  girl-ante-lat {ante-lat}-put.with.force.pf-temp

 ‘When (he) stretches the ring to the girl…’

For further details on locative prefixation in the Huppuq’ dialect, see Maisak and Merdanova (2002) and Maisak and Ganenkov (2016, pp. 3585–3588). In other Agul dialects, there is some variation in the number of affixes and their form, as well as in the number of slots. For example, in most dialects, there is an opposition between the localizations kː- ‘below a landmark’ (sub) and k- ‘attached to a landmark’ (cont), which are distinguished both as verbal prefixes and nominal case forms. The merger of these two localizations is characteristic only of southern varieties of Agul, including the Huppuq’ dialect (see also below on the probable contact-induced nature of this merger).

Morphology of the repetitive prefix

Productivity and frequency

Unlike locative prefixes, the repetitive marker is unrestrictedly productive and can co-occur with any verbal stem, including those that already have locative prefixes. An exception is statives, which form a small subclass of verbs with a reduced paradigm: some statives, for example verbs with locative prefixes derived from the roots ‘to be, be located’ and ‘to still be, remain’, do not co-occur with the repetitive prefix at all. Among non-locative statives, however, there are those that allow the repetitive derivation, compare qa-kːandea ‘wants again’ (< kːandea ‘wants again’) or qa-itːaa ‘is ill again’ (< itːaa ‘is ill’). On the whole, Haspelmath’s (1993, p. 174) claim about the repetitive prefix in Lezgian is equally applicable to its Agul counterpart: it is ‘so regular that it could even be considered an inflectional category of the verb’ (for more details on the Lezgian repetitive, see the Repetitive prefix: inherited or copied? section). On the inflection/derivation continuum, the repetitive can be placed between the locative prefixes (clearly derivational) and the negative ones (clearly inflectional).

The token frequency of ‘repetitive verbs’, that is, verbs with the repetitive prefix, is quite high. In the corpus of Huppuq’ texts, containing circa 92,000 words, there are 2037 tokens of repetitive verbs. The total number of non-stative verb forms in the corpus is around 25,000, so about every 12th non-stative verb form bears the repetitive prefix. On the whole, there are about 120 different repetitive verbs in the texts. A list of the top 23 verbs with more than 10 occurrences in the corpus is given in Table 3; for those verb stems that include a locative prefix, the gloss for this prefix is given with the translation. Note that, although the perfective stem of the verb ‘come’ is ad- (e.g. adine ‘came’), with the repetitive it occurs in the form aj- (e.g. q-ajine / q-ajne ‘came back, came again’).

OPEN IN VIEWER

Table 3.

Top list of verb stems that co-occur with the repetitive prefix.

Verb	Translation	No. of tokens
ad- (aj-)	‘to come’	448
ʕʷ-	‘to go/come’	418
uš-	‘to go away’	179
faj-	‘to bring’	110
aq’-	‘to do’	98
χ-	‘to bring/carry away’	84
ic’- / i-	‘to give’	82
hat-	‘to send’	61
fac-	‘to take, seize’	48
x-	‘to become’	43
ǯik’-	‘to find’	34
fajš-	‘to carry away’	28
ruq’-	‘to reach’	26
ajč’ʷ-	‘to get out’, {in}	18
ix-	‘to put inside’, {in}	15
aʁ- / up-	‘to say, tell’	13
alix-	‘to put on, build’, {super}	12
ačat-	‘to let in’, {in-lat}	11
ik’-	‘to put in (with force)’, {in}	11
alčarx-	‘to meet’, {super-lat}	11
agʷ-	‘to see’	10
uq’-	‘to sit down’	10

In can be easily seen from the table that the most frequent repetitive derivatives belong to the class of motion verbs; thus, repetitive verbs from the three verb stems – ‘come’, ‘go away’ and ‘go/come’ – account for more than 50% of all tokens. Another frequent triple is ‘bring’, ‘carry away’ and ‘bring/carry away’, with about 10% of all occurrences. These two triples are interesting in that each of them corresponds to two verb lexemes: in Huppuq’ Agul, deictic motion verbs ‘come’ and ‘go away’ have one and the same suppletive stem in the imperfective (while being distinct in the perfective and the imperative), and the same is true of the two deictic causative verbs of motion ‘bring’ and ‘carry away’. Moreover, while the imperfective for ‘bring/carry away’ is a separate non-derived stem χ-, ‘bring’ and ‘carry away’ are derived from ‘come’ and ‘go away’ by means of a preposed element faj- (> fa-). The latter is originally a converb of the stative verb faa describing temporary possession (‘have with oneself, have at hand’); hence, ‘bring’ and ‘carry away’ are historically combinations with the literal meaning ‘having at hand, come’ and ‘having at hand, go away’ (see (22)). With these derived verb stems, the repetitive prefix occurs infixally: however, ‘bring again’ and ‘carry away again’ rather seem to be derived not from already morphologized combinations for ‘bring’ and ‘carry away’, but from the original combination of faj- (> fa-) with the repetitive verbs ‘come again’ and ‘go away again’ – see (23).

(22) a. faj-ne / faj.i-ne / fajd.i-ne  (<  *fa-j      ad.i-ne)

  bring.pf-aor          apud.be-conv come.pf-aor

  ‘brought’

 b. fajš.u-ne        (< *fa-j    uš.u-ne)

  carry.pf-aor          apud.be-conv go.pf-aor

  ‘carried away’

(23) a. fa‹q›aj-ne / faj‹q›aj-ne / faj‹q›aj.i-ne (< *fa-j      q-aj.i-ne)

  <re>bring.pf-aor           apud.be-conv re-come.pf-aor

  ‘brought back, brought again’

 b. fa‹q›š.u-ne            (<  *faj      q-uš.u-ne)

  <re>carry.pf-aor           apud.be-conv  re-go.pf-aor

  ‘carried away again, took back’

From a functional perspective, the close association between motion verbs and the repetitive prefix is quite natural (see also Wälchli, 2006, pp. 91, 97). As will be shown in more detail below, one of the regular interpretations of this prefix is ‘back, backwards’, which is a salient notion in the description of the path of a moving figure (historically, moreover, this stage might have been even the initial one in the grammaticalization process).

Semantics of the repetitive prefix

Following Haspelmath’s (1993) description of a similar verb prefix in Lezgian, I have labelled the Agul morpheme q- ‘repetitive’, because its basic and most regular meaning is that of repetition: ‘do again’ (see (19) and (20) above, and also the two sentential corpus examples, (27) and (28), below). The repetitive meaning seems to be available to any repetitive verb, even if it can express any other meanings (contributed by the prefix) as well.

(27) qa-du.u-ne   ʡu-d-pu  sefer.i   penǯeg

 re-pull.pf-aor two-s-ord time(tmr) coat

 ‘We pulled his coat for the second time.’

(28) χab  q-agʷ.a-j-e     mi-s   ʡemk’

 again re-see.ipf-conv-cop dem-dat dream

 ‘Then she has a dream again.’

In a broader perspective, however, the Agul prefix belongs to a family of ‘refactive’ meanings, of which the repetitive proper is not the only one. This section will illustrate the other meanings, with examples from the Huppuq’ text corpus and with the terminological labels drawn from insightful works by Lichtenberk (1991), Wälchli (2006) and especially Stojnova (2013).

With motion verbs, the verb ‘give’ and some others, the reditive meaning ‘backwards’ is very common. Given the high token frequency of motion verbs (see Table 3), this meaning is especially prominent in the corpus, see (29) and (30) for illustration.

(29) q-aj-ne     χul.a-s,  uq’.u-ne  ustːul.i-q…

 re-come.pf-aor house-dat sit.pf-aor table-post

 ‘We came back home, sat at the table…’

(30) me  ʁʷan-ar  fa‹q›š.u-na     me-wur.i   qa-jc’.a-j-e…

 dem stone-pl <re>take.away.pf-conv dem-pl(erg) re-give.ipf-conv-cop

 me  baw.a-l-di=na     sus.a-l-di

 dem mother-super-lat=and bride-super-lat

 (The main hero sends jewels to his family with the merchants.) ‘They take the jewels (back to the village), and give them (back) to his mother and wife.’

The responsive belongs to the same ‘counterdirectional’ family of refactive meanings as the reditive, only in this case, the action is directed symmetrically to the agent of the preceding action (‘do in response’).

(31) ti     qa-ix.a-a    za-s  pul  reqː.ü-ʔ

 dem(erg)  re-in.put.ipf-prs  I-dat money road-in

 (I sent that woman some sheep wool.) ‘She sends me (in her turn) the money.’

(32) mi=ra     sa-d   q-ala-daʁ.u-ne      gi-l-as

 dem(erg)=add one-s   re-super-draw.out.pf-aor dem-super-elat

 (She – bang! – struck the man with a stick,) ‘and he struck her back.’

The additive meaning is the ‘extended repetitive’: a situation of the same type is being repeated, but involves another participant(s), or another ‘portion’ of an object (‘X also does’, ‘X does more’, etc.). Note that, in the glosses, add is the label for a very frequent enclitic, typical of East Caucasian languages, which expresses addition (‘and, also’) and is typically hosted by non-verbal phrases; see Forker (2016) on the typology of such markers.

(33) aχpːa če    dad=ra  qa-k’.i-ne,  itːar-x.u-na

 then  our.excl  father=add re-die.pf-aor be.ill-become.pf-conv

 (Soon after telling about her mother’s death.) ‘Then our father fell ill and also died.’

(34) sara fi   q-aʁ.a   zu wa-s,  sara fi.tːi-k-as

 more what re-say.ipf  I  you-dat more what-sub/cont-elat

 qa-raχ.a  zu wa-s,  ʜa-j-dewa     za-s

 re-talk.ipf I  you-dat know-conv-cop:neg  I-dat

 ‘What else can I tell you, about what else can I speak, I don’t know.’

The consecutive is close to the additive meaning, but what is ‘added’ in this case is not another participant, but the situation as a whole, which is being described as ‘supplementary’ in the series of events.

(35) χab  x.u-či     šuw=ra    q-al-č-arx.u-ne

 back become.pf-cond husband=add re-{super-lat}get.to.pf-aor

 itːa-jde

 be.ill-pt:hab

 (Telling about troubles in her life) ‘Besides, I got a husband who was ill.’

(36) me  χaǯalat=ra   qa-x.a-j-e      šuw.a-s

 dem suffering=add re-become.ipf-conv-cop husband-dat

 (The man returned and saw that his wife is gone: apart from other things) ‘the husband also had to go through this suffering.’

The restitutive, as opposed to the repetitive proper, focuses on the resultant state, and not on the action itself: it is not the action that is repeated, but rather the earlier state is restituted. Thus, in (37) the speaker describes the ‘returning’ to the state of gathering the candies on the road, while in (38) it is the state of the oxen being found that should be re-established.

(37) waʔ,  suwar, me  qːenfet-ar qa-gunt’-q’.a-s,   xe-s

 no  Suwar dem candy-pl re-gather-do.ipf-inf we.incl-dat

 te  idemi-s  c’.a-jde    šeʔ kːan-du…

 dem  man-dat give.ipf-pt:hab thing need-cop:neg:q

 (After the candies fell and scattered all over the ground, she said:) ‘No, Suwar, we have to gather all these candies again, we need something to give to that man’…

(38) aruc.u-na –  aruc.u-na    qa-ǯik’.a-j-dewa     jac-ar

 look.for.pf-conv look.for.pf-conv re-find.ipf-conv-cop:neg ox-pl

 (The oxen were lost in the mountains. They are searching for one, two, three days.) ‘They keep searching but they cannot find the oxen.’

There are also other semantic nuances that can be contributed by the repetitive prefix, including ‘continue doing’ (39), ‘do to completion’ (40) or ‘already’ (41), although these do not appear systematically.

(39) ha-ge  aχpːa qa-qatːq’.a-s-e    zun…

 em-dem  then re-recount.ipf-inf-cop I

 ‘I will later continue telling about that…’

(40) ha-le=ra   qa-ʕut’, qːučma p.u-naa, ha-le=ra qa-ʕut’…

 em-dem=add  re-eat.imp friend  say.pf-prf em-dem=add re-eat.imp

 {You’ve eaten a lot.} ‘Eat up this one, my friend, eat it up, – he said.’

(41) jeri-d-pu   is  qa-x.u-naje-f-e      sara.

 seven-s-ord year re-become.pf-pt:prf-s-cop ptcl

 ‘Seven years have passed already’ {since she died}.

Finally, repetitive derivatives from two frequent verbs with a very general meaning display idiomatic restitutive meaning (in addition to the regular repetitive one). The repetitive verb qa-xas derived from xas ‘to become’ can mean ‘to get better, recover’ (also ‘to become again’), while the repetitive verb q-aq’as derived from aq’as ‘to do, make’ is used as a transitive verb ‘to heal, cure, repair’ (also ‘to do again’). Note that such idiomaticity is not unique for Agul, as the same meanings of the repetitive from ‘to become’ and ‘to do’ are also attested in Lezgian and in the West Caucasian language Adyghe (Wälchli, 2006, pp. 95–96).

Thus, the repetitive prefix, instantiating a grammatical category in its own right, is highly integrated into the verbal system of (southern varieties of) Agul. It is unrestrictedly productive, exhibits high token frequency and displays a broad range of interpretations determined by the context and the particular verb lexeme. At the same time, there are reasons to suspect that this highly grammaticalized affix is a newcomer in the language and appeared as a result of language contact.

Repetitive prefix: Inherited or copied?

Although the systems of locative prefixes are very similar across the Agul dialects (especially concerning the localization markers), only two dialects spoken in the south of the Agul-speaking area make use of the repetitive prefix. In this section, I will argue that such areal distribution is not coincidental, but must reflect a direct influence on the part of Lezgian, a dominant language of the area where the two southern Agul dialects are spoken.

Lezgian seems to be the only other Lezgic language in which a specialized repetitive morpheme exists with a similar form and function. According to the description of Standard Lezgian by Haspelmath (1993, pp. 176–177), the repetitive marker expresses the meaning ‘again’, and represents ‘a productive verbal category that can in principle be formed from any verb’. As a prefix, it occurs in the form q- (qi-) or χ- (χu-, χü-), which are both restricted to a few verbs each. A more productive way of deriving repetitive verbs is by means of an infixal allomorph -χ- (after the first vowel of the verb). There is also a periphrastic repetitive construction with the auxiliary verb quwun ‘to do again’, which is the repetitive of awun ‘to do’ (e.g. kilig quwun ‘to look again’, lit. ‘look do-again’). The latter construction is unrestrictedly productive in Standard Lezgian (see also Mejlanova, 1970, p. 130); however, in the dialects prefixal/infixal repetitive derivatives are much more common than the periphrastic pattern (see Mejlanova, 1964, pp. 161–162; Ganieva, 2007, pp. 107–108, 2008, pp. 114–115, 2011, p. 109).

Some examples of the Lezgian repetitive verbs are provided in (42); see also Haspelmath (1993, pp. 174–176) and Moor (1985, pp. 95–103) for details on the morphological derivation. Note that the citation form of Lezgian verbs is the masdar (action noun) in -un.

(42) a. q-uwun ‘to do again’ < awun ‘to do’

  q-lahun ‘to say again’ < luhun ‘to say’

  q-fin ‘to go away’ < fin ‘to go’

  qi-jaʁun ‘to hit again’ < jaʁun ‘to hit’

 b. χ-gun ‘to give again’ < gun ‘to give’

  χu-taχun ‘to take back’ < tuχun ‘to carry’

 c. a‹χ›kun ‘to see again’ < akun ‘to see’

  ki‹χ›ligun ‘to look again’ < kiligun ‘to look’

The range of meanings expressed by the repetitive verbs in Lezgian is virtually the same as in Agul: the basic meaning is the repetitive proper (‘again’), with motion verbs and with ‘give’ the prefix adds the reditive meaning ‘backwards’, and some repetitive verbs display non-compositional meaning (e.g. ‘to heal’ alongside the regular ‘to become again’ from the verb xun ‘to be, become’). For a sentential example including two repetitive verbs, see (43).

(43) Am χta-na     wiči-n  čka.da-l  a‹χ›cuq’-na.

 she re.come-pf.conv self-gen  place-super <re>sit-aor

 ‘She came back and sat down on her seat again.’ (Haspelmath, 1993, p. 176)

The distribution of the repetitive prefix across the Lezgian dialects does not seem to be areally restricted: it is attested in all three major dialectal groups of Lezgian, including the Yarkin and the Kurakh dialects of the Kyura group (Mejlanova, 1964, pp. 97–98, 161–162), the Akhty and the Dokuzpara dialects of the Samur group (Mejlanova, 1964, pp. 216–217, 302) and the Yarghun dialect of the Quba group (A. Babaliyeva, personal correspondence, 4 July 2017). There is some variation among the dialects as to the existing and/or most productive variants of the prefix, but on the whole the repetitive seems to go back to the earliest stages of the Lezgian language, if not to the Proto-Lezgian (not to be confused with Proto-Lezgic, which is the ancestor of all the languages of the Lezgic branch).

It is most likely that the Lezgian repetitive is the result of semantic and morphological evolution of the locative prefix expressing ‘post’ localization, as suggested by Alekseev (1985, p. 121) and Haspelmath (1993, p. 174). The ‘post’ prefix going back to the *q- protoform is attested in Agul, Tabasaran, Rutul and Tsakhur, so there is every reason to expect to find it in Lezgian as well. In the system of locative cases, the ‘post’ localization is present in the majority of Lezgic languages, including Lezgian. However, as Haspelmath (1993, pp. 92–94) notes, the Lezgian ‘post’ case in -q is only rarely used with the meaning ‘behind’, but rather has more general locative semantics (expressing both location and direction, e.g. warari-q ‘towards the gate’) and also acquired a set of abstract functions, such as the encoding of the possessor or the exchange equivalent (cf. c’ud χipe-q ‘for 10 sheep’). A natural assumption then would be that as a prefix, the ‘post’ marker has also evolved from a purely locative (‘behind’) to a more general repetitive (‘again’), which became its core meaning.

Thus, in Lezgian the markers of the ‘post’ localization have changed both in the nominal and the verbal domain. Within the former, the locative suffix has extended its meaning, although it has kept its original form, and also its place in the paradigm of local cases. Within the latter, the prefix changed semantically (‘behind’ > ‘backwards’ > ‘again’), phonologically (as χ- became one of the main allomorphs) and also paradigmatically: the modern repetitive prefix no longer belongs to the set of (mutually exclusive) locative preverbs.

In Agul, the situation is different, as there is no reason to assume the proto-language origin of the repetitive marker, given its restricted distribution across the dialects. Contrary to Lezgian, in Agul both the repetitive and the (similar, but not totally identical) ‘post’ prefix co-occur in the prefixal system, belonging to different paradigmatic sets. If we take the ‘post’ prefix to be the grammaticalization source for the repetitive in Agul, their simultaneous presence in the prefixal system of the verb can be regarded as a case of ‘divergence’, in Hopper’s (1991, pp. 24–25) terms: of course, the newly grammaticalized marker does not have to oust its diachronic source completely. A crucial point, however, is that both Agul dialects that possess the repetitive prefix, namely the Keren dialect and the Huppuq’ dialect spoken in the south of the Agul-speaking area, are not only geographically adjacent to each other, but have also been in centuries-long contact with the neighbouring Lezgian villages.

Evidence in support of the contact-induced development of the Agul repetitive also comes from other phenomena in southern Agul dialects that appear to have a Lezgian origin. I will briefly describe three such cases, discussed previously by Ganenkov and Maisak (2008). The first case is nominal plural marking. While the singular in Agul is unmarked, the plural is expressed by means of suffixes. With nouns ending in a consonant, the plural marker -ar is employed in all the dialects, for example dad-ar ‘fathers’ (< dad ‘father’) or jac-ar ‘oxen’ (< jac ‘ox’). With nouns ending in a vowel, most dialects employ the plural marker -bur (-wur, -wr), for example geda-bur ‘boys’ (< geda ‘boy’) or dagi-bur ‘donkeys’ (< dagi ‘donkey’). However, in the Keren dialect, as well as in Huppuq’, the latter suffix is only used with demonstratives (e.g. me-wur ‘these’, te-wur ‘those’, etc.), while nouns ending in a vowel mark plural by means of a suffix -jar, for example gada-jar ‘boys’ (< gada ‘boy’) or degi-jar ‘donkeys’ (< degi ‘donkey’). Given that in Lezgian -jar is the only plural allomorph after vowels, the southern Agul plural marker -jar can be safely assumed to be a Lezgian borrowing.

Another example comes from the system of locative markers. As I mentioned previously in the Verbal prefixes in Agul section, in the Huppuq’ dialect the two localizations kː- ‘sub’ (‘below a landmark’) and k- ‘cont’ (‘attached to a landmark’) have merged into one ‘sub/cont’ localization in k-, both as a verbal prefix and as a nominal case form. For example, while in the Tpig dialect we have an opposition between ʁʷan.di-kː ‘(e.g. a snake is) under the stone’ and ʁʷan.di-k ‘(e.g. there are inscriptions) on the stone’, in the Huppuq’ dialect one and the same form ʁʷan.di-k will be used in both contexts. The distinction between ‘sub’ and ‘cont’ localizations reflects the original situation, so its loss in the Huppuq’ dialect is historically secondary. As noted by Alekseev (1985, pp. 52–63, 2003, p. 118), due to phonetic reasons such a merger has occurred in a number of Lezgic languages, Lezgian being one of them: in modern Lezgian, there is only a ‘sub’ locative case in -k.

The Keren dialect of Agul is not uniform in this respect, as in two villages the ‘sub’ versus ‘cont’ opposition is retained, while in three villages the merger has occurred, but only in the essive forms: for example, in Usug, ʁʷan.di-k will be both ‘under’ and ‘in contact’, but the elative cases will be distinct (cf. ʁʷan.di-kː-es ‘from under a stone’ and ʁʷan.di-k-es ‘from the surface of a stone’). This partial merger was caused by a regular phonetic process: non-aspirated stops in the final position, including /kː/, changed into the corresponding aspirated consonant (in this case, /k/). In Huppuq’, however, this phonetic process does not occur in other contexts, so the merger of the two localization markers cannot have a language-internal motivation. Influence from Lezgian is thus the most plausible cause of the ‘sub’/‘cont’ syncretism in Agul.

One more instance where Lezgian influence can be heavily suspected (although only in the Keren dialect), concerns the absolutive versus ergative opposition in personal pronouns. Whereas nouns in Agul distinguish between these two cases, the first- and the second-person pronouns lack it, for example zun ‘I (abs/erg)’, wun ‘you (abs/erg)’, čin ‘we, exclusive (abs/erg)’, etc. Only in the Keren dialect do personal pronouns have a distinct ergative form in -š, e.g. zun ‘I (abs)’ versus zaš ‘I (erg)’, wun ‘you (abs)’ versus waš ‘you (erg)’, čin ‘we, exclusive (abs)’ versus češ ‘we, exclusive (erg)’, etc. Although the origin of the ergative suffix -š is not clear, the pattern itself may reflect Lezgian influence, as in the latter language the absolutive and the ergative cases of personal pronouns make a morphological opposition, for example zun ‘I (abs)’ versus za ‘I (erg)’, wun ‘you (abs)’ versus wuna ‘you (erg)’, čun ‘we (abs)’ versus čna ‘we (erg)’, etc.

Interestingly, precisely grammatical copies from Lezgian (structural or both material and structural) are most easily identified in Agul. On the phonological level, for example, the Agul consonant system remains more complex than that of Lezgian, in that it possesses a series of post-uvular (pharyngeal, epyglottal) phonemes. Some contact influence can be suspected in the Huppuq’ dialect, though, as it is the only dialect of Agul that has lost the prosodic feature of pharyngealization (peaking on vowels or uvulars). In Lezgian, this feature has also been lost completely in most dialects, while it is preserved, although to varying degrees, in all Agul idioms besides the Huppuq’ one. As far as the lexicon is concerned, there is no extensive borrowing from Lezgian (especially as compared to the massive inflow of loan words from Russian). Lezgian words or phrases can be spotted in conversation in Agul, but mostly when the topic of conversation is events involving Lezgians, or when Lezgian speech is being quoted. In addition, loans from Lezgian are hard to identify, as the two languages share many cognates that are phonologically very similar if not identical (especially as regards noun roots).

Conclusions: The path for the Agul repetitive

In the Morphology of the repetitive prefix and Semantics of the repetitive prefix sections, I have described the morphological and semantic properties of the prefix q(V)- in Agul. Paradigmatically, this prefix belongs to a separate set of prefixal morphemes of the verb, being opposed to a set of locative prefixes and a set of negative prefixes. Unlike the locative prefixes, the repetitive is unrestrictedly productive, so it is very close to an inflectional marker: repetitive verbs can be derived from any (non-stative) verb. The following meanings of the refactive domain are commonly associated with the prefix (as a rule, each repetitive verb can express different meanings depending on the context):

There is an obvious phonetic (and also semantic, cf. ‘behind’ and ‘back’) affinity between the repetitive prefix and one of the locative prefixes, namely q(V)- encoding the ‘post’ localization. A diachronic link between the ‘post’ prefix and the repetitive prefix is very plausible, but what is the most likely scenario of development? Was the development independent, or does it reflect contact influence? The distribution of the repetitive prefix across the Agul dialects seems crucial here: the prefix is attested only in southern Agul dialects, spoken on the Agul-Lezgian linguistic border. Speakers of these dialects are mostly bilingual in Lezgian, and there are cases of Lezgian influence on the grammatical structures of these dialects. In its turn, Lezgian is a language where a productive repetitive prefix exists, which developed on the basis of the ‘post’ localization.

At the same time, in other Agul dialects, as well as most other languages of the Lezgic branch, no special repetitive prefix is attested, although the association of the ‘post’ localization marker with the expression of some meanings of the refactive domain has been reported. Thus, according to Xanmagomedov and Šalbuzov (2001), in Tabasaran (the language most closely related to Agul), prefixal derivatives with the ‘post’ prefix q- express the standard meaning ‘behind’ (q-iwub ‘put behind’, q-tibt’ub ‘tie behind, e.g. a belt’, q-ebgub ‘run behind, pursue’), and also the reditive meaning ‘back’ (q-tabkub ‘come back, return’, q-tap’ub ‘give back’). In Tsakhur, there are ‘post’-derivatives with the meaning ‘back’ (qa-les ‘come back’, qe-qːas ‘ask to give back’), and even ‘again’ (qe-zas ‘plough again’, qa-ʔas ‘do again; repair’), see (Ibragimov, 1990, pp. 124–125).

As such, the connection between the meanings ‘back’ and ‘again’ is cross-linguistically common, seen in comparison with the polysemy of the Latin and Romance (and also English) prefix re- or a widely spread grammaticalization path from ‘return, go back’ to ‘do again’ (Heine & Kuteva, 2002, pp. 259–260; Maisak, 2005, pp. 246–247; Stojnova, 2013, pp. 180–184). This is why, even though in (southern) Agul the two prefixes (repetitive q- and ‘post’ q-) are paradigmatically distinct, the diachronic connection between these two markers is beyond doubt. In principle the independent origin of the Agul ‘post’ > ‘repetitive’ cannot be excluded, but it is highly unlikely given that this development is found exactly in those Agul dialects that have been in tight contact with neighbouring Lezgian-speaking villages.

Thus, the most plausible historical scenario for the repetitive in southern Agul dialects is that this prefix emerged as a morphological copy from Lezgian. The two cognate morphemes q- ‘post’ and q- ‘repetitive’ in Agul are ‘etymological doublets’: while the former directly continues the Proto-Lezgic locative marker, the latter is a later borrowing from a closely related language, ultimately going back to the same proto-language morpheme (see Figure 2 for a schematic illustration).

OPEN IN VIEWER

Figure 2.

Diachronic scenario for the Agul repetitive prefix.

The category of repetitive in Agul can be regarded as intermediate between derivation and inflection: the formation of repetitive verbs is almost unrestrictedly (although not 100%) productive, and the meaning provided by the prefix is almost fully regular (although several interpretations are available alongside the core meaning ‘again’, or ‘back’ in reference to motion). The borrowing of such a semi-derivational prefix is in line with the common expectation that derivational affixes are more easily replicable than inflectional affixes (Weinreich, 1953, pp. 31–33), and a more recent finding that ‘inherent inflection, i.e. the inflectional categories which are more similar to derivation, such as aspect, tense, mood, gender, number and inherent cases… is borrowed far more frequently than contextual inflection, i.e. person and structural cases’ (Gardani, 2008, p. 84).

In terms of grammatical borrowing typology, the transfer of the repetitive prefix from Lezgian to Agul would be a case of ‘global copying’ (involving both form and function) in the terms of Johanson (1999), or MAT[ter] borrowing as opposed to PAT[tern] borrowing in the terms of Sakel (2007). Similar kinds of replication in the prefixal system have been reported in the literature: for example, some Slavic and Baltic spatial/aspectual prefixes were ‘globally’ copied to a number of Romani varieties, Istroromanian and also the Finno-Ugric language Livonian (Arkadiev, 2005, p. 2015: 232–257, To appear). In the case of such closely related languages as Agul and Lezgian, the borrowing process has probably been facilitated by their structural similarity, in particular in the system of verbal prefixation.

As pointed out by one of the anonymous reviewers of this paper, one cannot exclude the possibility that what we are dealing with here is not a direct borrowing of a Lezgian repetitive prefix, but rather a contact-induced grammaticalization (or rather ‘regrammaticalization’) in Heine and Kuteva’s (2003) terms. It might be the case that the Agul repetitive prefix is in fact a native Agul morpheme, namely the locative ‘post’ prefix, which has undergone the same semantic shift from ‘backwards’ to ‘again’ under the influence of the homophonous Lezgian marker. This resulted in the divergence between a source locative marker and a new repetitive prefix. Such a scenario, although not impossible, presupposes two very different paths of development for the two cognate locative morphemes. The Lezgian ‘post’ prefix q- must have evolved from a locative into a repetitive, having lost the source meaning while preserving its identity as a single morpheme. At the same time, the Agul ‘post’ prefix q- continues as such in all Agul dialects except the southern ones, where it splits into two morphemes, of which the second one copies the meaning of the Lezgian repetitive prefix. Unfortunately, we do not have any historical data on older stages of Agul or Lezgian, so it is difficult to tell whether such contact-induced change might have happened when the Lezgian prefix still displayed its original locative meaning, or whether the speakers could identify the two prefixes only based on the phonetic similarity between them. The alternative scenario of material copying, for which I have argued above, should rather presuppose the lack of clear functional associations between the Lezgian repetitive and the Agul ‘post’ marker, and hence the borrowing of the former into Agul as a distinct morpheme.

With a lack of historical evidence, however, it becomes difficult if not impossible to verify these two alternative hypotheses. At some point, a new morpheme appears in southern Agul dialects, which may be either a borrowed one (Lezgian q- ‘repetitive’ > Agul q- ‘repetitive’), or a native one, influenced by the semantics of a homophonous Lezgian prefix (Agul q- ‘post’ > Agul q- ‘post’ + Agul q- ‘repetitive’). One fruitful line of future research may be the investigation of other Lezgic languages that show associations of their ‘post’-prefixes with the meanings ‘backwards’ or even ‘again’. Such research might make it clear to what extent ‘post’-prefixes show signs of grammaticalization into repetitive prefixes across the Lezgic branch, and also whether the two markers can become paradigmatically distinct as a result of language-internal development, and not of borrowing, as (presumably) happened in Agul.