- 2 In addition, the recorded corpus contains one description of ancient marriage practices by <B> and a </B> (...)
1All Berber languages have a category indirect object, which is marked differently from other objects. Moreover, in most of them, indirect objects, different from other objects, allow for a construction known as “clitic doubling”, in which the same referent is expressed both lexically and pronominally. In this article, I shall study the contexts of clitic doubling in the Berber variety spoken in the oasis of Figuig in eastern Morocco. This language has been described by several authors, most notably Saa (2010, originally 1995), Kossmann (1997), Ben-Abbas (2003), Sahli (2008) and Benamara (2013). While doubling of indirect objects is sometimes mentioned, none of these sources provide an account of the distribution of the construction. There is no tradition of writing Figuig Berber, but a number of texts have been published, mostly in an academic context (e.g., Benamara 2011; Kossmann 2000: 104-125). The present article is based on one text genre, traditional narratives,2 and it is impossible to say to what extent the features analyzed here are specific to this genre or represent more general tendencies, also found, for instance, in conversational style.
- 3 In Figuig, i is homonymous with the locative preposition i ‘in’. The difference can be seen in pro (...)
2In Figuig Berber, like in the other Berber languages, the indirect object has different expressions before nouns and with pronouns. Before nouns, it is expressed by a prepositional phrase using i,3 while pronominalization is realized by means of a special pronoun. This pronoun is part of the verbal clitics which, depending on their syntactic context, may stand after or before the verb. The following examples illustrate this basic structure:
- 4 The following abbreviations are used: ao: Aorist; as: Annexed State (état d’annexion); do: direct (...)
(1) t-εəyyəḍ i mmi-s n wətna-s <B>4
3sg:f-call:pv io son-3sg of sister-3sg
‘she called the son of her sister’
(2) i-ṛaḥ n wuday i-εəyyḍ=as=dd i-nna=yas <A>
3sg:m-go:pv to Jew:as 3sg:m-call:pv=3sg:io=vent 3sg:m-say: pv=3sg:io
‘he went to a Jew, he called him and said to him’
3The indirect object pronouns are morphologically different from direct object pronouns, and resemble the series of bound pronominal forms used after prepositions, compare the forms of the third person singular bound pronouns (simplified, for a full overview, see Kossmann 1997: 177-185):
Table 1: Bound 3sg pronouns
|
Direct object
|
Indirect object
|
after the preposition (a)kid- ‘with’
|
3sg:m
|
ss ~ t ~ i
|
as
|
-s
|
3sg:f
|
tt ~ it
|
as
|
-s
|
4Based purely on morphology, it would be possible to analyze most indirect object pronouns as prepositions, with an allomorphy between i and a, i.e., as = a-s (cf. Bentolila 1981: 214). This is problematic on the morphosyntactic level, as the indirect object pronoun has different syntactic behavior than prepositions with pronominal suffixes. In the first place, prepositions with pronominal suffixes are the last member of the clitic cluster, following other clitics (example 3), while indirect object clitics are the first in the cluster (example 4), cf.
(3) iwa t-uṛəw=dd=akid-əs <A>
well 3sg:f-give.birth:pv=vent=with-3sg
‘well she got a child with him’
(4) t-uṛw=aš=dd Qəṛnfəṭṭa <A>
3sg:f-give.birth:pv=2sg:m:io=vent Goldhorn
‘she has given birth for you to Goldhorn’
5Moreover, indirect objects allow for a construction, impossible with any other preposition, in which the preposition i is followed by a free pronoun, e.g.
(5) awy=axdd wala i nəšni idžən <O>
bring:ao:imptv:pl=1sg:io also io we one:m
‘bring (us) one to us too!’
(6) uš=i=tət=dd i nətš! <D>
give:ao:imptv:sg=1sg:io=3sg:f:do=1sg:io io I
‘give it to me’
- 5 This definition is inspired by the description of ‘affected object’ in Hausa in Frajzyngier & Munk (...)
6Indirect objects convey that the object is involved in the event but only indirectly or partly affected by it.5 This has different implications with verbs of different semantic classes. In transfer verbs, the indirect object is the participant to(wards) which the transfer takes place, i.e., basically the recipient, e.g.
(7) y-uš=as ləḥwayəž nn-əs <Z>
3sg:m-give:pv=3sg:io clothes of-3sg
‘he gave her her clothes’
7This is different from locational constructions, in which the location towards which the transfer takes place is not involved in the event, cf. the difference between the following two examples with awəy ‘to carry toward’. In example (8), the transfer takes place towards a person, who–as the recipient–is involved in the action, while in example (9), the object tiddart ‘house’ is not involved, and the directional preposition l is used:
(8) uwy-ənt=as=dd səbεa n tməllalin <A>
bring:pv-3pl:f=3sg:io=vent seven of eggs:as
‘they brought her seven eggs’
(9) t-iwy=it l tiddart nn-əs <O>
3sg:f-bring:pv=3sg:f:do to house of:3sg
‘she brought it to her house’
- 6 Verbs that imply a dialogue, such as sstən ‘to ask’, are constructed with a direct object, however
8In verbs of speaking, the indirect object expresses the person towards whom the speech act is addressed,6 e.g.
(10) i-nna=yas <passim>
3sg:m-say:pv=3sg:io
‘he said to her’
(11) t-uyyəy a=dasən=t-əfḍəḥ. <D>
3sg:f-refuse:pv nr=3pl:m:io=3sg:f-reveal:ao
‘she refused to reveal (it) to them’
- 7 With one verb, this has become lexicalized. The verb ɣrəs ‘to slaughter’ takes an indirect object. (...)
9Verbs that describe actions manipulating only part of the object typically take the indirect object.7 When the manipulated part is also mentioned, such expressions can often be translated into English as possessives.
(12) qqim i-tḥəssn=as i… i… i-tḥəssn=as
dur 3sg:m-cut.hair:ipv=3sg:io 3sg:m-cut.hair:ipv=3sg:io
i Tyəmlult <C>
io Tayemlult:as
‘he started to cut ehh to cut Tayemlult’s hair (lit. to cut to Tayemlult)’
(13) i-nəkḍ=as=din aqəlqul nn-əs i… i lmuddən <M>
3sg:m-cut.off:pv=3sg:io=there head:fs of-3sg io muezzin
‘he cut off the muezzin’s head (lit. he cut off his head to the muezzin’
10When the object is manipulated as a whole (or conceptualized as such), the direct object is used, as shown by the use of the direct object aqəlqul ‘head’ in example (13).
- 8 fhəm in the cognitive sense of ‘to understand’ is constructed with a direct object, e.g. i-fhəm ay (...)
- 9 Indirect objects are found both when the object is the source and when it is the content, e.g. ləb (...)
11Indirect objects are also used for the objects of non-visual perception events, such as “to understand (acoustically)”,8 “to feel” and “to hear”.9
(14) s tmaṛa ay das=fəhm-ən <D>
with difficulty foc 3sg:io=understand:pv-3pl:m
‘it was with difficulty that they understood her’ (context: they listen to what is said by a girl whose tongue has been cut off)
(15) yawkan i-səll=as ann n nmalik <A>
then 3sg:m-hear:pv=3sg:io that of king
‘then the king heard her’
12Visual perception is constructed by means of a direct object, however:
(16) t-əmmatər yuma-s <A>
3sg:f-see:ao brother-3sg
‘and she saw her brother’
13Many verbs do not have an indirect object as an integral part of their event scheme, but can still be constructed with one. In such cases, the indirect object refers to a participant indirectly involved in, or affected by, the action. Most of such “ethical datives” have a benefactive or a malefactive connotation (cf. also Rapold 2010). Both connotations occur in the following passage, in which a father catches a bird for his son. The stepmother takes away the bird and kills it:
- 10 ṭṭir is a code-switch with Moroccan Arabic; the normal Figuig Berber word is abərḍal.
(17) ayu n ṭṭiṛ10 i-ṭṭf=as=dd ukk uyənn əhh i
this of bird 3sg:m-hold:pv=3sg:io=vent io that:as io
mmi-s
son-3sg
iwa t-əkkər ayu n tməṭṭut n p̣p̣a-s
well 3sg:f-rise:pv this of woman:as of father-3sg
t-əssəṛɣ=as=s t-əyr=as=s
3sg:f-burn:pv=3sg:io=3sg:m:do 3sg:f-throw=3sg:io=3sg:m:do
ikk ləmsi. <O>
in fireplace
‘this bird, he took it for (i.e., to the benefit of) that one, ehh for his son.
well, then the wife of his father burned it (to his detriment)
and threw it (to his detriment) into the hearth’
14Often the term “dative” is used to refer to what is called the indirect object here (e.g., Rapold 2010). This indicates that the transfer function is the central meaning, and that other meanings are associated to this. In Berber, the indirect object can be defined by means of a reasonably coherent meaning: the indirectly or partly affected object. As this seems to cover the entire usage range of the category, there is no need to establish a more restricted prototypical meaning.
15It is possible to have several indirect objects expressed with one single verb, e.g.
(18) εəyyḍ-əmt=idd i tqəyyarin <B>
call:ao-2pl:f:imptv=1sg:io io girls:as
‘call for me (to) the girls!’
- 11 There are no examples in the corpus of a construction with two pronominalized indirect objects wit (...)
16In example (18) there are two indirectly affected participants: the girls to whom the calling is directed, and the first person, to the benefit of whom the action is to be performed. It is not possible to have several indirect object pronouns in one single clitic complex. In the case that two different indirect objects are both pronominalized, one of them has to be demoted to a prepositional phrase using di- ‘in’ (Kossmann 1997: 217),11 e.g.
(19) day ɣərs=idd=di-s <A>
only slaughter:ao:imptv:sg=1sg:io=in-3sg
‘just slaughter it for me’
17The verb ɣrəs ‘to slaughter, to cut the throat to’ is constructed with an indirect object marking the slaughtered object. In example (19), there is a second indirect object, indicating the beneficiary of the slaughtering. In order to express both parti-cipants pronominally, one of the objects is put into a di- phrase.
18Two indirect object phrases, one pronominal and one with a noun phrase, often cooccur in a single clause, while referring to the same indirect object, e.g.
(20) iwa uš-ənt=as ukk uyənn n əhh uyənn n
well give:pv-3pl:f=3sg:io io that:as of that:as of
twəssart ayu n nəεwərt. <O>
old.-woman:as that of boy
‘well, they gave that boy to that ehh that old woman’
19In example (20) the old woman is referred to both pronominally (as) and by the prepositional phrase ukk uyənn n twəssart ‘to that old woman’.
20This phenomenon is known in syntactic literature as “(indirect object) clitic doubling” (cf. the discussion in Souag 2014). Although it may be more correct to speak of “coreferent pronominal and lexical expression of indirect objects”, I will use here the shorter and catchier term doubling.
21In most cases, the prepositional phrase is within the same intonation contour as the verb. The lexical indirect object can be followed by other parts of the clause, e.g.
(21) qqim i-ssiṛiḍ=as ukk uyənn n tməṭṭut
dur 3sg:m-dress:ipv=3sg:io io that:as of woman:as
ayənn n ilmawən <O>
that of hides
‘he started to dress this woman with these hides’ (lit. ‘make wear these hides to the woman’)
22The construction also occurs within subordinate clauses, e.g.
(22) i das=i-ɣrəs ukk unn n tzərzərt
when 3sg:io=3sg:m-slaughter:pv io that:as of gazelle
i-mmatər… <A>
3sg:m-see:ao
‘when he had slaughtered the gazelle, he saw that…’
(23) iwa y-awəy ann n tənn əhh
well 3sg:m-carry.to:ao that of that:sg:f
das=ala=n-əssεiš i lmalik əhh amənsi <A>
3sg:io=nr=sr-feed:ao io king dinner:fs
‘well, and he married the one ehh that would feed to the king ehh dinner’
23One can conclude that the lexical indirect object is part of the core of the clause in this construction, and not some type of extraposition or afterthought (cf. also Ouali 2011: 131-133).
24Clitic doubling only refers to the use of two referential expressions of the indirect object in the core of the clause. This is different from topicalization, which is peripheral to the clause. Berber is basically Verb-initial, but has a topicalization position that precedes the sentence. The topic is syntactically, and often intonationally, kept apart from the core (cf. Lafkoui 2011: 119ff.), and the topic is referred to pronominally in the core, e.g.
(24) yəmma-tnəx əhh ləhḍəṛt nn-əs t tazdatt <D>
mother-1pl speech of-3sg pred thin:sg:f:fs
‘our mother, ehh, her voice is delicate’
(25) yah ann n mmi-s t-əssrus=i <A>
well that of son-3sg 3sg:f-put.down:ipv=3sg:m:do
‘well, her son, she always laid him down’
(26) wətna-s y-uš=as lqaləb n ssəkkwar <D>
sister-3sg 3sg:m-give:pv=3sg:io sugar.loaf of sugar
‘his sister, he gave her a sugar loaf’
25This is different from clitic doubling with indirect objects. While pronominal reference in topicalization is also found with direct objects and prepositional objects, this is impossible in the core of the clause, where only indirect object doubling is allowed.
- 12 While the construction is often mentioned in passing in descriptions of Berber varieties, only few (...)
26Souag (2014) provides a comprehensive cross-dialectal overview of indirect object clitic doubling in Berber, as well as a historical account of its development.12 As he shows, the incidence of the phenomenon shows important variation within the Berber language family. In some languages it is almost obligatory with lexically expressed indirect objects, and can be considered a kind of agreement, e.g., in Siwa (Souag 2014). In other languages, such as Niger Tuareg, it is ungrammatical (Kossmann 2011: 122). Most Berber languages are in between: they allow for both constructions. Souag, mostly basing himself on published text corpora, concludes that, within the group allowing for both constructions, there are important dialectal differences as to their frequency. Some languages have it very often, e.g., Kabyle, while it is quite rare in others, such as Tashelhiyt (cf. also Rapold 2010: 353-354). Finally, studying the usage of clitic doubling, he concludes that, in addition to other correlations, there is an important lexical factor at work: the verb ‘to say’ is significantly more often used with clitic doubling than other verbs.
27Inspired by these results, I decided to take a look at Figuig Berber, a language which allows for indirect object clitic doubling, but does not require it. To this purpose, I use two copora:
- 13 Data from dictated texts, mainly by story teller <Y>, have sometimes been used as supplementary evide</Y> (...)
28a. The K-corpus: Traditional narrations transcribed by the present author as a fieldwork corpus between 1990 and 1993. I have used the texts for which I have tape recordings, while dictated texts were left out.13
29b. The B-corpus: Traditional narratives edited by Hassane Benamara in his Contes berbères de Figuig (Sud-est marocain) (Benamara 2011). The actual text of the narratives is not based on recordings (Benamara 2011: 5, n. 2), and sometimes strongly edited (Benamara 2011: 5). Still, the author managed to remain close to oral style, and the way the stories are told is similar to what I have in my recordings.
- 14 The use of f may be another archaism in Benamara’s style. Fouad Saa (p.c.), who has the same diale (...)
30The K-corpus consists of 4.5 hours of recordings, amounting to close to 38,000 words, the B-corpus counts about 24,000 words (clitics are counted as separate words). The text genre is basically the same in both corpora. As for the dialect, Figuig Berber is not entirely homogeneous (Ben-Abbas 2003). The K-corpus mainly consists of stories from the village Zenaga, with the addition of a few stories told by a speaker from the village Elmaiz. The B-corpus is entirely from Zenaga, but shows a number of differences with the language of my recordings and of my spokesmen during fieldwork. This shows especially in the somewhat more archaic vocabulary used by Benamara (e.g. the use of aməšli instead of ləfḍuṛ for ‘noon meal’) and the allomorph f before nouns of the implicative preposition, which is s in my materials (Kossmann 1997: 223).14 Similar details of linguistic variation can be found within the K-corpus. The speakers in the K-corpus are of different age and occupation. Quite an important number of the stories were told by elderly women, recorded for me by someone of their families. Another part of the corpus is spoken by younger male story tellers. In spite of these differences in detail, the two corpora can be considered representative for the language of traditional narrations in Figuig.
- 15 Of course, only sentences that have ini with a lexical indirect object were counted. While ini is (...)
31The study of clitic doubling in the two corpora shows one very strong result. While doubling is common in the K-corpus, it is extremely rare in the B-corpus. This is most clearly shown by the incidence of clitic doubling with the verb ini ‘to say’.15 In the K-corpus, doubling is almost consistently used with this verb, with only two instances without it. In the B-corpus, on the other hand, only seven instances of doubling occur with this verb:
Table 2. Clitic doubling with ini ‘to say’
|
Doubling
|
%
|
no Doubling
|
%
|
total
|
K-corpus
|
74
|
97%
|
2
|
3%
|
76
|
B-corpus
|
7
|
9%
|
74
|
91%
|
81
|
32Something similar is found with other verbs. In verbs with animate indirect objects (inanimates are too infrequent in the texts to draw conclusions), about one third have doubling in the K-corpus. In the entire B-corpus, only eight of such instances are found:
Table 3. Clitic doubling with verbs other than ini ‘to say’
|
Doubling
|
%
|
no Doubling
|
%
|
total
|
K-corpus
|
43
|
38%
|
69
|
62%
|
112
|
B-corpus
|
8
|
7%
|
100
|
93%
|
108
|
33The differences between the two corpora are very clear: While indirect object clitic doubling is a common construction in the K-corpus, it is almost absent in the B-corpus. I assume that this is a stylistic difference. As mentioned above, Hassane Benamara has written down the stories on the basis of his own notes, but not with the help of recordings. His textualization therefore implies the low-pace, highly conscient way of language use typical of the writing process. Apparently, clitic doubling is suppressed to a high degree by this process. One cannot exclude that the author consciously chose not to use clitic doubling, maybe considering it an uneconomical way of expression. However that may be, his usage clearly shows that clitic doubling is not necessary in the language, and that a writer who approaches a vivid oral style in writing can freely leave out doubling.
34This stylistic explanation for the difference between the two corpora is corroborated by the distribution of the few doubled indirect objects in Benamara’s corpus. Six out of fifteen instances, including four instances of ini, come from one single tale, Nr. 47 (Lila d Σmeṛ, p. 168-178). As Benamara explains in his notes, this particular tale was “recueilli” by Mahmoud Alehyane. It is not clear whether this also means that (part of) the textualization comes from Alehyane, but this is very well possible.
35It is hardly possible, nor interesting, to study the distribution of clitic doubling in the B-corpus, as it is almost entirely absent. The K-corpus, on the other hand, provides a good basis for a more detailed analysis of the uses of clitic doubling.
- 16 Only found in texts from one speaker, Y, who dictated a number of stories to me.
36There are a few contexts in which doubling is impossible. In the first place this is the case of reflexive indirect objects. In Figuig Berber, reflexives are made by means of a noun phrase using the nouns iman (only used for reflexives) or lxaḍər (lit. something like ‘mind, liking’)16 followed by a possessive. In spite of being lexical nouns syntactically, they are pronominal in reference, and their construction always includes a pronominal element. Apparently double pronominal reference is not possible here, and therefore doubling is not allowed, e.g.
(27) nəttata t-awəy=dd igg man nn-əs əhh -ləfrist <F>
she 3sg:f-carry.to:ao=vent io self of-3sg prey
‘and she brought prey for herself’
(28) t-əkkər t-ənna Žamila i lxaḍəṛ nn-əs
<Y, dictated>
3sg:f-rise:pv 3sg:f-say:pv Jamila io mind of-3sg
‘and Jamila said to herself’
- 17 There are too few instances of concrete inanimate indirect objects in the corpus to make any state (...)
37In the second place, doubled indirect objects are not attested with abstract or otherwise non-visual indirect objects,17 e.g.
(29) y-awəy=dd Ṛəppi idžən n əəh n nmalik
3sg:m-carry.to:ao=vent Lord one:m of of king
i-səll i nndir <A>
3sg:m-hear:pv io moaning
‘the Lord brought a ehh king who heard the moaning’
38On the other hand, as shown above, clitic doubling is almost consistently found with the verb ini ‘to say’. There are only very few instances where there is no doubling with this verb. The two following sentences represent all the cases that were found in the recordings:
(30) an i das=səll-ən middən, i… aḥbib
until when 3sg:io=hear:pv-3pl:m people friend:fs
i-qqaṛ ukk ḥbib <A>
3sg:m-say:ipv io friend:as
‘until when people heard it, and one friend said to the other’
(31) sad=ini-x ikk išš n tbuṭaṭṭ <A>
fut=say:ao-1sg io one:f of fortune-teller:as
‘I shall say to a fortune-teller’
39Consistent doubling is not found with other verbs of speaking and calling, such as ɛəyyəḍ ‘to call’, which freely allow for both constructions.
40There are 112 instances of lexically expressed indirect objects in the K-corpus that are neither in a context obstructing clitic doubling, nor with the verb ini. Among these, 43 have clitic doubling, while 69 have not. Although the number of instances is not extremely large, it allows one to look more closely at the distribution of the two constructions. Semantic sub-usage of the indirect object does not seem to play a role in this distribution. All sub-classes seem to allow for both constructions. This is true for usages where the indirect object can be considered part of the argument frame, such as transfer verbs, e.g.
(32) iwa uš-ənt=as ukk uyənn n əhh uyənn n
well give:pv-3pl:f=3sg:io io that:as of that:as of
twəssart ayu n nəεwərt <O> (doubling)
old.woman:as this of boy
‘well they gave this boy to this ehh old woman’
(33) uš-ən=tən ikk əhh id... id=bab
give:pv-3pl:m=3pl:m:do io pl=possessor
n-sən <O> (no doubling)
of-3pl:m
‘they gave them to ehh their possessors’
41It is also true for usages where the participant is more peripheral to the event, as in ethical datives, e.g.
(34) i-yy=as=tt i tfunass nn-əs əhh ikk
3sg:m-do:pv=3sg:io=3sg:f:do io cow:as of-3sg in
məẓɣan <E> (doubling)
ears
‘he put it concerning his cow in the ears; i.e. he put it in the ears of his cow’
(35) ṛaḥ-ən ttašr-ən ikk idžən n užəllid <D>
(no doubling)
go:pv-3pl:m steal:ipv-3pl:m io one:m of king:as
‘they went stealing to (the detriment of) a king’
- 18 In exophoric uses, i.e., when referring to entities that are present in “the situation surrounding (...)
42The analysis of information structure, on the other hand, allows us to trace a distributional pattern, although as a tendency, and not as a strong condition. Clitic doubling is mostly found in contexts where the referent is already known to the hearer, and present in the preceding discourse. In Berber narrations, such a situation can be overtly expressed by means of an anaphoric deictic expression. In Figuig, this expression is mostly a pre-nominal element ayu, ann or ayənn,18 linked to the noun by means of the possessive preposition n. It should be noted that the use of a deictic is by no means obligatory for referring to an old-information referent.
43New referents in the discourse, on the other hand, only rarely have clitic doubling. New information is often indicated by a marker of indefiniteness, mostly idžən n… (feminine: yišš n…). Like with old information, the use of a marker is not obligatory with new information.
44When looking at contexts that are explicitly old or new information, as shown by the use of a(yə)nn and idžən, the following picture emerges:
Table 4. Doubling and no doubling in explicitely
new and old information contexts
|
Doubling
|
%
|
no Doubling
|
%
|
total
|
new information: idžən
|
1
|
10%
|
9
|
90%
|
10
|
old information: a(yə)nn
|
16
|
67%
|
8
|
33%
|
24
|
total
|
17
|
|
17
|
|
34
|
45Examples:
46New information with doubling:
(36) an i i-wtu idžən n nbaḷu
until when 3sg:m-hit:pv one:m of ball
i-ṛṛẓ=as ikk išš n taεṛabt
3sg:m-break:pv=3sg:io io one:f of Arab:sg:f
- 19 This sentence is pronounced without any hesitation in a single intonation contour. The story telle (...)
-taməṛmiṭṭ nn-əs <A>19
saucepan:fs of-3sg
‘until when he hit the ball and broke to an Arab woman the saucepan’
47New information without doubling:
(37) iwa y-isi=tt məskina ukk yis nn-əs
well 3sg:m-take:pv=3sg:f:do poor:f in horse:as of-3sg
i-εəyyəḍ ikk idžən n nmənɛul n wuday <A>
3sg:m-call:pv io one:m of damned of Jew:as
‘well he took her, the poor one, on his horse and called a (damned) Jew’
48Old information with doubling:
(38) qa a=dasən=ɣərs-əx ukk unn n
entirely nr=3pl:m:io=slaughter:ao-1sg io that:as of
-wufrišən <A>
rams:as
‘I will slaughter all those rams’
49Old information without doubling:
(39) i-ɣrəs qaε ukk unn n tulyišin <A>
3sg:m-slaughter:pv entirely io that:as of bad:pl:f
‘he killed all these bad women’
50Clearly, clitic doubling is highly unusual when new information is explicitly expressed, while with explicitly old information doubling is more frequent than its absence.
51When including cases where the question of old and new information has to be inferred from context, a similar pattern appears:
Table 5. Doubling and absence of doubling
according to information contexts (recorded texts)
|
Doubling
|
%
|
no Doubling
|
%
|
total
|
new information
|
2
|
6%
|
30
|
94%
|
32
|
old information
|
33
|
58%
|
24
|
42%
|
57
|
general non-specific
|
0
|
0%
|
6
|
100%
|
6
|
total
|
35
|
37%
|
60
|
63%
|
95
|
52For seventeen instances (8x with doubling, 9 times without), I could not establish with confidence the information status of the dative element. “General non-specific” refers to cases that either give a general statement, which would be true of any referent, or a statement in which the identity of the referent is irrelevant, e.g.
(40) u=dam=n-əsmiḥ mta t-ṛeyyḍ-əd i ḥədd <O>
neg=2sg:f:io=1pl-forgive:np if 2sg-stop:pv-2sg io -anybody
‘we don’t forgive you if you stop for somebody’
53Thus, a pattern appears, in which doubling is strongly disfavored with new information, and slightly favored with old information. It should be noted, however, that in the latter case both constructions are possible, and sometimes occur in almost identical contexts. Thus, in a story where an ogress kills her own daughters, the first killing is described with doubling:
(41) t-əyy=as=dd amm=u ukk uyənn n əhh n
3sg:f-do:pv=3sg:io=vent like=prox io that:as of of
yəlli-s əhh taḍəṛɣalt. šštəxxxx <D>
daughter-3sg blind:sg:f:fs shhhtekhhh
‘(she took a hot stick and stung it into her daughter) she did (to her) like this to that ehh blind daughter, shhhtekhhh’
54The second killing, a few lines later, is described without doubling:
(42) dəxx t-əssitf=in ukk unn n
then 3sg:f-make.enter:pv=3pl:m:do io that:as of
tuɛdimt. šštəxxxx <D>
handicapped:sg:f
‘(she took other hot sticks) and then she stung them into that handicapped girl, shhhtekhhh’
55Information structure is not the only feature that shows a correlation with the use of doubling. Several authors have defined this difference as pertaining to definiteness (Souag 2014; Guerssel 1995: 115-116; Ouali 2011: 131, the latter two apparently on the basis of native speaker intuitions). While there is much overlap between new/old information and indefinite/definite interpretation, the two are not equivalent. In fact, the texts abound in examples where definite nouns convey new information, e.g.
(43) ṛuḥ at=t-əqql-əd manəš i-lla
go:ao:imptv:sg nr=2sg-look:ao-2sg how 3sg:m-be:pv
i-tekk i mmi-m <A>
3sg:m-do:ipv io son-2sg:f
‘go and look what he is doing to your son’
- 20 For a recent account which problematizes the notion of definiteness in Moroccan Arabic, see Maas ( (...)
56In this example, mmi-m,‘your son’, has not been mentioned before by the cited speaker–as part of a warning coming out of the blue for the addressee it can be considered entirely new information. However, mmi-m is clearly definite, in the sense that it has unique, specific, reference known to the addressee, as the woman in question has only one son. In Figuig Berber, definiteness is not expressed grammatically. This does not necessarily mean that it is absent from the mind of the speakers, as almost all speakers of Figuig Berber are also fluent in Moroccan Arabic, which has overt expression of definiteness.20 A categorization according to definiteness leads to the following results:
|
Doubling
|
%
|
no Doubling
|
%
|
total
|
new info, indefinite
|
1
|
5%
|
20
|
95%
|
21
|
new info, definite
|
1
|
9%
|
10
|
91%
|
11
|
old info, definite
|
33
|
60%
|
22
|
40%
|
55
|
general non-specific (indefinite)
|
0
|
0%
|
6
|
100%
|
6
|
unclear information status, indefinite
|
1
|
50%
|
1
|
50%
|
2
|
unclear information status, definite
|
7
|
41%
|
10
|
58%
|
17
|
total indefinite
|
2
|
7%
|
27
|
93%
|
29
|
total definite
|
41
|
49%
|
42
|
51%
|
83
|
total
|
43
|
|
69
|
|
112
|
57The results are almost identical to those of the information structure account. Indefinites and new information are only marginally used with dative doubling. Within definites, the slight preference for dative doubling with old information contexts is not mirrored by the figures for definiteness, which show an equal distribution over the two constructions. On the basis of these data, it is impossible to decide whether definiteness or information structure contributes most to the choice of construction; however, because of the stronger correlation between dative doubling and ayənn phrases (see above), the information structure account may present some more insight.
- 21 This analysis is comparable to that proposed by Mettouchi (2011: 482), who stresses the importance (...)
58On the basis of these results, it is interesting to have a fresh look at ini ‘to say’. As shown above there is a very strong tendency to have clitic doubling with this verb, a tendency also found in other Berber languages. The question is to what extent this is specific to quotative verbs and the constructions they involve (cf. Souag 2014 for a proposal), or a result of the information structure contexts where ini is mostly found in the text genre under scrutiny. On first sight, the lexical analysis seems to stand this test: Although ini is mostly used with indirect objects already present in the discourse, quite some attestations with clitic doubling involve protagonists that were not yet mentioned in the story. There is a caveat to this. In most cases, first mention of a protagonist implies the introduction of a new character into the mental image of the hearer. There are, however, cases where the presence of such a protagonist can be inferred from the larger context of the events described.21 This seems to be the case with most instances of “new” information in ini. Many of these refer to persons that are expected to be around, e.g.
- 22 Note that this does not imply that the king only has a single servant; a more interpretative trans (...)
(44) (the king made a ring the size of the foot of a certain girl he is looking for)
i-kkər i-nna=yas i tnəxdamt nn-əs:<O>
3sg:m-rise:pv 3sg:m-say:pv=3sg:io io servant:as of-3sg
‘and he said to his servant:22 (whom would fit this ring?)’
59The servant is new information–no servant was mentioned before in the story–, but her presence can be deduced from the expected cultural context. Of course, all such cases could be considered “definite”.
60The corpus provides only very few examples of ini constructed with an overtly indefinite indirect object; in such cases dative doubling is still more often found than lack of doubling (for which, see ex. 33), e.g.
(45) i-ṛaḥ wuday i-ffəɣ i-nna=yas
3sg:m-go:pv Jew:as 3sg:m-go.out:pv 3sg:m-say:pv=3sg:io
ikk idžən n uməddukəl nn-əs <M>
io one:m of friend:as of-3sg
‘the Jew went out and said to a friend of his’
61One may therefore question that the exceptional status of ini is purely a lexical feature, unrelated to the wider usage of the construction. Instead, it seem that its status is partly due to the type of contexts in which this verb is normally found. This does not mean that lexical information does not play a role at all. As shown above, verbs other than ini mark old information a bit more often by clitic doubling, but it is also possible not to have it. With ini, on the other hand, old information always leads to doubling. Thus, what is a tendency with other verbs is exceptionless with ini. It should be noted that this is only the case within the stylistic register reflected by the K-corpus. As mentioned above, the B-corpus ini only rarely has clitic doubling, irrespective of lexical features, definiteness, or information structure.
62The use of indirect object clitic doubling in Figuig Berber is determined by a number of independent factors. The first factor is style, overriding everything else: there are some stylistic usages where doubling is strongly disfavored. This is found in the textualizations by Benamara, and may be related to the slow and speech-conscious process involved in writing down a language that has no tradition of writing. When looking at spoken registers, clitic doubling is quite common, and other factors play a role. In the first place, some syntactic contexts never show doubling, esp. pronominal-like contexts, such as reflexives, and abstract indirect objects. In the remaining cases, there is a clear tendency related to information structure and/or definiteness: Clitic doubling is common when expressing old information and definites, while it is strongly disfavored in the expression of new information and indefinites. In addition to this, there is a lexical effect: what is a tendency with most verbs in the spoken corpus is exceptionless with ini ‘to say’, which, in the spoken corpus, always has doubling in contexts with old information.