1. Introduction

2These deictic distinctions are expressed in a number of morphosyntactic categories. In the first place, almost all Berber varieties have adnominal deictics, that is, deictic elements that are added directly to the noun. For example, in Nador Tarifiyt in (1).

(1)	Nador Tarifiyt [MA-Z] (Mourigh & Kossmann 2019: 173–175)
	aryaz	‘the man, a man’
	aryaz-a	‘this man here’ (proximal)
	aryaz-in	‘that man over there’ (distal)
	aryaz-ənni	‘the aforementioned man’ (anaphoric)

4In addition, deictic elements are combined with pronominal bases in order to create demonstrative pronouns. In most varieties, there is a clear connection between the adnominal deictic elements and the elements as found in demonstratives, as for example in (2).

(2)	Nador Tarifiyt [MA-Z]
	w-a	‘this one over here’ (m:sg:prx)
	w-in	‘that one over there’ (m:sg:dist)
	w-ənni	‘the one just mentioned’ (m:sg:ana)

5While demonstratives of this type are found in all Berber varieties, the connection with the adnominal forms is not always straightforward, as will be shown below.

6Finally, a number of adverbial elements can be combined with deictic elements. These can be more or less lexicalized. Some examples are given in (3).

(3)	Nador Tarifiyt [MA-Z]
	da	‘here’ (proximal)
	din	‘over there’ (distal)
	diha	‘over there’ (far distal)
	dinni	‘over there’ (anaphoric)

7It has been proposed that the adnominal deictics are historically derived from constructions of a noun followed by a pronoun. Thus, Galand (2005: 191) writes about adnominal deictics:

• 3 ‘Originally the construction simply involved the recapitulation of the noun (or nominal element) b (...)

À l’origine de la construction, il y a simplement reprise du nom(inal) par un pronom, comme en touareg (Ahaggar) : aləs wa-rəġ « homme celui-ci » = « cet homme-ci ». Mais le plus souvent cette origine est perdue de vue, le pronom est réduit à l’état d’élément invariable et il fonctionne simplement comme une marque qui accompagne le nom : chleuh argaz-ad, kabyle argaz-agi, nefusi atəRas-uh « cet homme-ci ».3

8In the citation above, it is not entirely clear if Galand considers forms like argaz-ad to be phonetically reduced outcomes of *argaz wad, or that he considers ad to be a pronoun in itself (as suggested by a slightly different formulation in Galand 2010: 156). The first analysis (*argaz wad) would put us way beyond the timeframe of reconstructible Berber, as almost every Berber variety has adnominal deictics that are added directly to the noun. The few varieties that do not have this construction have fossilized forms that show the previous presence of adnominal deixis, such as Ayer Tuareg [ni-t] ămer‑ă ‘at this moment, now’; Siwa [eg] asf-a ‘today’; Zwara [li-z] ass-u ‘today’. Moreover, it is important to keep in mind when studying Tuareg examples like aləs wa-răɣ ‘this man’, that almost all Tuareg varieties also have constructions with adnominal deictics, such as Ahaggar [al-t] aləs-a ‘the man we are talking about’ (Prasse 2010: 116), and it is very well possible that the Tuareg construction with demonstrative-like pronouns (wa-răɣ) is a local innovation (Prasse 1972: 193).

9The second interpretation of Galand’s analysis would be about a refunctionalization of a now-defunct independent pronominal element into a bound morpheme, presumably the “supports de détermination” (pronominal elements that constitute the head of a modified or determined NP); on this question, see Section 5 for further discussion.

10In any case, the adnominal construction with bound deictic elements like in aləs-a, clearly has a long history. Adnominal deictics originally had number agreement to the noun (see Section 2). One could speculate that Galand – who, as far as I know, never commented on this phenomenon – would have considered this an argument in favor of their synchronic status as pronominal elements.

11Most Berber systems have been described as distinguishing at least two deictic points of spatial reference (proximal and distal; exophoric reference in the terminology of Diessel 1999) and, in addition, a dedicated expression of anaphoric deixis (endophoric reference), marking that the referent is already mentioned, or deducible from context. Some varieties distinguish more spatial points of reference, adding for example a difference between ‘near distal’ (or ‘near the hearer’) and ‘far distal’. This can be the case all over the deictic system of the variety, but can also be restricted to certain subsystems. Thus, in Tarifiyt [ma-z], locational adverbs differentiate between general distal din and far distal diha, while it has only one distal element (-in) in the adnominal and demonstrative subsystems. There are a few varieties that only have two spatial deictics. Here the split seems to be between an exophoric proximal and a second deictic used both for distal and anaphoric reference (cf. Kossmann 2015 for Figuig [ma-z]). One should also mention the case of Tetserret [ni], a non-Tuareg variety from Niger, which has an adnominal element ad, which is largely neutral in its deictic reference, opposed to an element un, which tends to be related to the expression of distance (Lux 2013: 443). As argued by Lux, fixed expressions point to an earlier proximal value of ad, e.g. ell-ad ‘here’.

13This article will provide a partial analysis of the history of the morphology of deictic systems in Berber. As this is a study of morphology and not of semantics, I have chosen to retain the common denominations of these series. Thus, in spite of Mettouchi’s insights, I will stick to ‘proximal’, ‘distal’, ‘anaphoric’, even though there is no doubt that this is just an oversimplified character­ization of much more complex semantic and pragmatic facts.

14This article will focus on three subjects. The first subject is the reconstruction of the singular forms of the proximal deictic. The second subject is the expression of number together with the expression of deixis. The third subject is a series of deictics, which is only sporadically attested in Berber, but will be proposed here as being ancient. This article does in no way pretend to provide a full analysis of all aspects of Berber deictic forms. For example, there are more series, especially those including a non-proximal element n (e.g. Tarifiyt [ma-z] -in and -nni), which will only be mentioned where relevant, and left out of our general efforts in reconstruction. This omission is in no way a dismissal of these forms as innovative. Similarly, the history of elements specialized in expressing temporal deixis will not be studied here.

15One variety that illustrates the systems that will be studied here is Ghadames [li] (Ayt Waziten dialect; Lanfry 1968; Kossmann 2013b). In this variety, there are three basic series, which have different forms depending on number (4). In addition, the proximal and the distal deictics have extended forms using a suffix -ăt.

16The origin of the singular proximal forms -o and -odăt will be treated in Section 2; the plural forms will be treated in Section 3, while the anaphoric series -e/-id will be the subject of Section 4.

2. The singular forms of the proximal deictic

17Berber varieties show three different basic forms of the proximal deictic (Destaing 1922; Naumann 2001). Zenaga [mt] will be treated below.

-a

-o and –u

-ad

18Forms in a are found in Tuareg, in Kabyle [al], and in a number of Central Moroccan, Senhaja de Sraïr [ma], and Zenatic varieties, as well as in Awjila [li]. Forms in -u and -o are found in a sub-set of Zenatic and in Ghadames [li]. Forms in ad are typical of Moroccan Berber: Tashelhiyt [ma], most Central Moroccan dialects [ma], Ghomara [ma], and some Senhaja de Sraïr [ma] varieties. They also occur in Tasahlit [al], in Zenaga [mt], and in Tetserret [ni].

19Ghadames [li] points to a fourth option, -od. In this variety, the proximal and distal deictics have extended forms ending in ăt. In the distal, this ăt is simply joined to the deictic: -ănn-ăt. In the proximal, the basis seems to be od rather than o, and we get -od-ăt. The extra d also appears in the extended form of the locational adverb, da, which is dadăt ‘here’.

20In addition, a few varieties have proximal forms ending in h. The first one is Jebel Nefusa (Jado) [li], which has a binary distinction between proximal and distal forms. The consonant h is found in both deictic series, for example, in the adnominal suffixes prx -uh ~ -uha; dist -ih ~ -iha (Beguinot 1942: 120). The second one is Zwara [li-z], where h-final forms are restricted to proximal demonstratives (Mitchell 1953: 376–377); h does not occur in the few fossilized forms with adnominal deixis such as ass-u ‘today’ (Mitchell 2009: 80). Like in Jebel Nefusa, this variety only expresses two degrees of deixis, with the proximal demonstratives wuh (m:sg), yih (m:pl) (f: tuh/tih), and the non-proximal demonstratives wəddin (m:sg), yidin (m:pl) (f: təddin/tidin).

21In Siwa [eg], final h is typical for the distal demonstratives, e.g., wih ‘that one’, and in many other varieties, h functions as (part of) the distal element (cf. Tarifiyt [ma-z] diha ‘over there’). One way to explain why h functions as a proximal in Zwara [li-z], while it seems to have distal semantics elsewhere, is the following – perforce speculative – scenario. Originally, both Jebel Nefusa [li] and Zwara [li-z] had two (or more) forms, one of which was non-proximal, and ended in -h, while another was proximal, and did not end in -h. In both varieties, the final h was generalized to the proximal deictic. This is the situation as currently found in Jebel Nefusa (Jado). In Zwara, this expansion of -h to the proximal deictic did not extend to all fixed expressions, as witnessed by ass-u ‘today’. In a later development, Zwara lost the non-proximal form ending in ­-h, whose function was taken over by forms in -ddin, which may or may not have been part of the larger deictic system already. As a result, the only remaining form ending in -h in Zwara is the proximal, even though h originally comes from a non-proximal form.

22It has been long remarked that the d in proximal deictics is similar in form with the directional clitic dd, which, generally speaking, expresses movement towards the speaker (on which, see Bentolila 1969; Kossmann 2014; Taine-Cheikh 2015, and many others; for a critical reassessment, Mettouchi 2015). The comparison is strengthened by the fact that in many varieties, distal and/or anaphoric deictics contain the consonant n, which can be compared to the directional clitic nn expressing movement away from the speaker. Thus, for example, Tashelhiyt [ma] has a highly symmetric system (5).

(5)	Tashelhiyt [MA]
		directional	deictic
	prx	dd	ad (also a)
	dist	nn	ann

24Therefore, it is better to consider forms with a and o/u as alternatives to ad rather than derivations from it (Galand 2002: 208 [1969]; 2010: 156; Taine-Cheikh 2010). Whether d is really identical to the directional clitic, and, if so, which directionality we should assume, are questions that lie outside the scope of this article (cf. Gutova 2021: 309 for a discussion with references).

25This brings us to the next question: how should we understand the dialectal alternation between a(d) on the one hand and u/o on the other? At this point, the advances in Berber historical phonology of the last few decades come to our rescue. As shown in Kossmann (2001), stressed *ăɁ became o in Ghadames [li] and u in a number of Zenatic varieties and in Jebel Nefusa [li]. The evidence is based on correspondences found in verb conjugations that have an original glottal stop in Zenaga, the only Berber variety that has preserved this phoneme (Taine-Cheikh 2004).

27As for the varieties that have a(d), there is a distinct possibility that they also go back to *aɁ, as a is the regular correspondent there. However, a reconstruction *a is also possible, and for the individual varieties, it is impossible to decide which one to choose. In fact, the original forms may have been different for different varieties.

28It is relevant that the varieties that have o or u (< *aɁ) as their adnominal deictic have deictic forms with -a in locational adverbs. A number of varieties of this type are listed in (6).

(6)		adnom:prx(:sg)	‘here’	source
	Ghadames [LI]	-o	da	Lanfry (1968)
	Jebel Nefusa [LI]	-uh(a)	dah	Beguinot (1942)
	Ouargla [AL-Z]	-u	da	Delheure (1987)
	Figuig [MA-Z]	-u	da	Kossmann (1997)
	Ayt Seghrushen [MA-Z]	-u	da	Bentolila (1981)

29This suggests that, at some moment in their history, these varieties had both a deictic with *aɁ and one with a; the form with *Ɂ was generalized to adnominal and demonstrative deixis, while the form without *Ɂ was retained in some spatial adverbs.

30This idea is reinforced by the existence of a further set of demonstrative pronouns in Ghadames [li], which have sg a instead of o. This is the case of the pronoun ‘another’, e.g., m:sg wa-yiḍ and the interrogative ‘which one’, e.g., m:sg wa-din (Kossmann 2013b: 49). These forms are likely derived from a basis *-a rather than *-aɁ.

(7)	Zenaga [MT] (Taine-Cheikh 2010)
		prx	med	dist
	adnom:sg	äđ	iʔđ	ān
	dem:m:sg	äđ	iɁđ	ān
	dem:f:sg	täđ	tiɁđ	tān
	neutral	äyđ	äɁđ	ān
	locational adverb	đāđ	đäɁđ	y

32As one can see, Zenaga has a deictic form with a glottal stop, but this is in the medial series, which is characterized by a non-low vowel. Zenaga i can go back to proto-Berber *ə, *i, *u, *e, but not to *ă and *a (Kossmann 2002), so Zenaga iɁđ cannot be directly cognate to *aɁ or *ăɁ.

33The forms of the locational adverbs in Zenaga are difficult to explain, as the origin of the long vowel in đāđ ‘here’ is unclear. One may conjecture that they somehow go back to a locational basis da- followed by the deictic element, and that *da-ad became đāđ, while *da-iɁd became đäɁđ. The latter vowel coalescence rule might also be used to explain the neutral medial pronoun äɁđ (< *a-iɁđ ?), but the rule *a-a > ā needed for đāđ does not have a counterpart elsewhere in the language. In fact, the origin of other long vowels in Zenaga seems to be different, and therefore this explanation of đāđ is, at best, pure speculation.

34There are two ways to explain the relationship between the Ghadames [li] (etc.) forms with reconstructible glottal stops and the Zenaga [mt] system. In order to do so, we will also take into consideration the Zenaga medial deictics, which, arguably, correspond to the Ghadames anaphoric deictics in -e (see Section 4 for a comprehensive discussion).

35In the first scenario, an early phase of Berber had a system that included the following deictic singular forms (forms with n have not been included):

*-a

*-ad

*-aɁ (or *-ăɁ)

*-iɁ (or *-eɁ)

(8)	Scenario 1	Ancestral system	Zenaga [MT]	Ghadames [LI]
	prx-1	*-a	lost	(maintained in adverbs)
	prx-2	*-ad	-äđ	lost
	prx-3	*-aɁ	lost	-o
	med / ana	*-iɁ	*-iɁ + d >> -iɁđ	-e

37In the alternative explanation, there would not have been an element *-aɁ in proto-Berber, and the subset of deictics that concern us here could be reconstructed as follows:

*-a

*-ad

*-iɁ

38In Zenaga [mt], *-a was lost, and the consonant d was joined to *-iɁ according to the same developments as proposed in the first scenario. In Ghadames [li], Jebel Nefusa [li] and (part of?) Zenatic, the glottal stop of *iɁ was extended to proximal deixis (except in adverbs), and, in an unrelated development, *ad was lost. The second scenario is summarized in (9).

(9)	Scenario 2	Ancestral system	Zenaga [MT]	Ghadames [LI]
	prx	*-ad	-äđ	lost
	prx-2	*-a	lost	*-a + Ɂ >> -aɁ > -o adverbs maintain *-a
	med / ana	*-iɁ	*-iɁ + d >> -iɁđ	-e

39Both scenarios are predicated upon the presence of a glottal stop in the original form of the Anaphoric series in Ghadames [li]. This is impossible to prove or disprove outside Zenaga [mt], as (stressed) *əɁ and likely also *iɁ (or *eɁ) merged with e or i everywhere (cf. van Putten 2016: 27), so Ghadames -e could go back both to *iɁ (or *e) and to *e.

40These scenarios are, of course, speculative, and I have no strong opinion about which of them should be considered more likely.

42We may safely assume that there was a semantic difference between proximal forms without a formative, forms with d and, if reconstructible, forms with Ɂ. As there is no variety that has clearly preserved this original system, it is impossible to decide what difference this could have been exactly.

3. Number marking in deictic forms

3.1 Introduction

43The majority of Berber varieties do not have gender or number marking in adnominal deixis. Thus, in Tarifiyt [ma-z], -a, -in, -nni can be combined with singular and plural nouns, and with feminine and masculine nouns without any changes, as, for example, in (10).

(10)	Nador Tarifiyt [MA-Z]
		prx	dist	ana
	‘donkey’ (m:sg)	aɣyuř-a	aɣyuř-in	aɣyuř-ənni
	‘donkeys’ (m:pl)	iɣyař-a	iɣyař-in	iɣyař-ənni
	‘donkey’(f:sg)	taɣyuč-a	taɣyuč-in	taɣyuč-ənni
	‘donkeys’ (f:pl)	tiɣyař-a	tiɣyař-in	tiɣyař-ənni

44Demonstratives, on the other hand, normally have gender and number marking. Gender is marked in the initial consonant of the pronoun, while number is marked in the vowel following this initial consonant, or by other means (on which, see Section 4).

45In the case of Tarifiyt [ma-z], the demonstrative can be analyzed as the combination of a pronominal element, which includes gender and number marking, and the same deictic elements that are used for adnominal deixis (11).

46The Ashtuken Tashelhiyt [ma] system is even more balanced, with clearly distinguished gender marking, number marking and deictic marking (12).

(12)	Ashtuken Tashelhiyt [MA] (Aspinion 1953: 92–94; forms adding initial ɣ not included)
		prx	med	dist	ana
	adnom	ad	nna	ann	lli
	dem:m:sg	w-a-d	w-a-nna	w-a-nn	w-a-lli
	dem:m:pl	w-i-d	w-i-nna	w-i-nn	w-i-lli
	dem:f:sg	t-a-d	t-a-nna	t-a-nn	t-a-lli
	dem:f:pl	t-i-d	t-i-nna	t-i-nn	t-i-lli
	neutral	ay-a-d, ay-a	ay-nna	ay-a-nn	ay-lli

47In varieties like Tashelhiyt [ma], morphemic analysis is simple, and it makes sense to distinguish a pronominal basis with gender/number marking from a second part that consists of the specific morphemes that indicate deixis.

3.2 Number marking in adnominal deixis

Table 1 — Varieties that have a number opposition in adnominal deictics: distribution

Variety	Expression of number	Source
Zenaga [MT]	only proximal adnominal deictics	Taine-Cheikh (2010)
Senhaja (Ketama, Taghzout) [MA]	most adnominal deictics	Gutova (2021)
Ghomara [MA]	all adnominal deictics	Mourigh (2015)
Blida Atlas [AL]	only distal adnominal deictics	El Arifi (2016)
Eastern Kabyle (Ayt Mbarek) [AL]	proximal and distal adnominal deictics	Genevois (1955)
Tasahlit (Ayt Mhend) [AL]	only proximal adnominal deictics	Rabhi (1994)
Ghadames [LI]	all adnominal deictics	Lanfry (1968); Kossmann (2013b)
Awjila [LI]	all adnominal deictics	van Putten (2014a)
El-Fogaha [LI]	only proximal adnominal deictics	Paradisi (1963)

50Number marking in adnominal deixis is found both in varieties that have final d in the proximal and varieties without d. Thus Zenaga [mt], Senhaja (Ketama) [ma], Ghomara [ma], and Tasahlit (Ayt Mhend) [al] have forms in d, while Eastern Kabyle [al] (Ayt Mbarek), Ghadames [li], and Awjila [li] lack d. One may therefore conclude that the presence of number marking is independent of the presence or absence of d. Table 2 presents the relevant forms (unless stated otherwise, proximal forms are listed).

Table 2 — Varieties that have a number opposition in adnominal deictics: proximal forms

Variety	sg	pl
Zenaga [MT]	-äđ	-iđ
Senhaja (Ketama) [MA]	-ad, -adah	-id, -idah
Senhaja (Taghzout) [MA]	-ada	-idi
Ghomara [MA]	-ad, -adin, -adinət	-id, -idi, -idinət
Blida Atlas [AL]	[distal:] -adin (among other forms)	[distal:] -idin (among other forms)
Eastern Kabyle (Ayt Mbarek) [AL]	-a	-i
Tasahlit (Ayt Mhend) [AL]	-a, -ad	-i, -id
Ghadames [LI]	-o, -odăt	-i, -idăt
Awjila [LI]	-aya, -e, -ək	-iya, -iyək
El-Fogaha [LI]	-dda	-di, -ddi

51It seems that *iɁ became i everywhere in Berber; therefore, all plural forms of the proximal could (but need not) go back to *iɁ.

53In many varieties with number marking in adnominal deictic elements, the same opposition appears in non-proximal deictics, as seen in (13)-(15) (for Kabyle [al] forms see below).

(14)	Senhaja (Taghzout) [MA]
		prx	med	dist
	sg	-adi	-ayəs	-an, -ayən, -ayənnah
	pl	-idi	-iyyəs	-iyyən, -iyyənnah

(15)	Ghomara [MA]
		prx	med	ana
	sg	-ad, -adin, -adinət	-an, -ani, -anit	-ahən
	pl	-id, -idi, -idinət	-in, -ini, -ininət	-ihən, ihin

54In Ghadames [li] (see Section 1), the proximal o/i marking corresponds to a marking sg -ănn ~ ‑onn, pl ‑inn in the distal.

55The situation in Awjila [li] (16) is difficult to analyze, as the exact system behind the attested forms is unclear. While it is conceivable that -aya is an extended form of -e, the form -ək may in fact be related to Siwa [eg] forms with addressee agreement (van Putten 2014a: 121; Souag 2014a: 548; 2014b: 43).

(16)	Awjila [LI]
		prx	med	dist
	sg	-aya, -e, -ək	-idin	-iwan
	pl	-iya, -iyək	-	-idanin

56From the discussion above, one may conclude that in adnominal deixis number marking is wide-spread, being found in varieties of Mauritania, northern Morocco, eastern Algeria, and Libya. This suggests that it is an old phenomenon in Berber.

4. The Ghadames anaphoric series and its cognates elsewhere

4.1 Introduction

57Ghadames [li] is quite different from most Berber varieties in that it has two series of (adnominal and demonstrative) deictic elements that both end in a vowel in the singular. The Ghadames system is summarized in (17), leaving aside the extended forms in (d)ăt (Lanfry 1968; Kossmann 2013b).

58While the distal marker is characterized by the consonant nn, the proximal and the anaphoric marker are distinguished by having different vowels.

59The anaphoric series in Ghadames, with its variation -e (sg), -id (pl) has a number of possible cognates in other eastern varieties. The exact meaning of the forms is sometimes different, and rarely well-described, so one should assume a certain degree of systemic reshuffling, either in Ghadames, or elsewhere, or both.

61In Siwa [eg], the original adnominal deictic series was lost, but in the demonstratives a system similar to Awjila [li] and Ghadames [li] appears (18). Siwa has three degrees of distance: proximal, medial, and distal. In the medial set, a difference is made according to the addressee (Naumann 2001: 70; Souag 2014a; 2014b; Schiattarella 2016: 33). These medial demonstratives derive from the incorporation of the preposition ɣuṛ ‘at’ (Souag 2014a: 541; Souag & van Putten 2016: 192), and will not be provided or discussed here (forms following Souag 2014b: 36).

(18)	Siwa [EG]
		prx	dist
	dem:m:sg	wa, waya	wih
	dem:f:sg	ta, taya	tih
	dem:pl	wi, wiyya	widin

62The similarity to Ghadames [li] is obvious: Ghadames ana:pl -e corresponds to Siwa [eg] dist:pl ­‑i (+ h), while ana:pl -id corresponds to Siwa dist:pl ‑id (+ in).

63There are a number of Berber varieties that have ‑i as a proximal or medial marker, both in the singular and in the plural. It is possible that they represent cognates of the Ghadames Anaphoric series. The following cases have been identified:

1. South-Eastern varieties of Central Moroccan Berber (Ayt Atta, Mauri 2020; Dades region, Willms 1972: 175): Proximal -i.

(19)	Ayt Atta [MA] (Mauri 2020: 6)
	nkkʷni	n-kk	abrid-i
	we	1pl-pass:ao	road-prx
	‘(and) we’ll take this road’

64Ayt Atta [ma] also has an alternative proximal marker, ‑a, which combines with a small subset of nouns mostly used in adverbial contexts, e.g., ass ‘day’, ass-a ‘today’. Willms (1972: 175), writing about neighboring dialects, describes the forms with i as belonging to the “zweite Entfernungsstufe” (“second degree of distance”), which probably refers to near-hearer deixis, whereas ‑i is a simple proximal marker in the variety described by Mauri.

2. Jebel Nefusa (Jado) [li] has two deictics, both neutral as to number: proximal ‑uh(a) and distal ‑ih(a) (Beguinot 1942: 120), where the final h(a) may be some kind of fossilized suffixed element. The distal form ‑i (+ h) could be cognate to Ghadames [li] ‑e.

3. Mali Tuareg has a number-neutral proximal adnominal deictic ‑i, where other Tuareg varieties have ‑a (Heath 2005: 242). The same form ‑i is used in Ayer Tuareg [ni-t] for near-hearer deixis in some fixed expressions (Prasse et al. 2003: 966). It should be noted that in this case the correspondence with Ghadames [li] ‑e is not straightforward. Tuareg has an opposition /e/ vs. /i/, and assuming a neutralization of this opposition in the relevant forms is ad hoc.

4. One may add Zenaga [mt], which has ‑iɁđ as the number-neutral medial deictic, both in adnominal deictics and in demonstratives. One way to analyze this is that its vowel is cognate to the Ghadames [li] anaphoric form. The adjunction of đ may be considered a generalization of the final consonant of the plural form, probably also inspired by the final đ of the proximal äđ (see Section 2).

4.2 Greater Kabylia and the Ghadames anaphoric series

65Varieties from Greater Kabylia [al] provide highly relevant forms when it comes to this question. As mentioned above, Mettouchi (2011) has shown that the description of deixis as only related to distance and anaphora is too simplistic for Kabyle [al]. The chosen labels should therefore be taken as representing different morphological series, and not as full (or even correct) descriptions of their usage.

67In addition, there are forms with -ən, which appear in adverbial expressions like ass-ən ‘that day’, imir‑ən ‘that time’ (Basset & Picard 1948: 94–96; Chaker 1983: 178–179). Note that non-lenited ġ [g] in -aġi (etc.) comes from *yy, so all forms with -aġi represent *-ayyi.

68With demonstratives the system is different. In the following, only the masculine demonstratives will be provided, which start in w-. Feminine forms are the same, except that w- is substituted by t-. In addition, there are neutral demonstratives based on ay-. They sometimes have combinatory restrictions with deictic elements, but this does not seem to be relevant to the issues at stake here.

69We shall start with the system in At Iraten [al] in (21), as described by Chaker (1983). In this variety, demonstratives have four degrees, which will be labeled here as proximal, medial, distal and anaphoric. In addition, there is an “indefinite” form, whose uses remain unexplained, and which is rare (Chaker 1983: 156; see also Naumann 2001: 41, who argues that the entire indefinite series should be detached from the demonstrative paradigm; see also the discussion in Section 5 concerning supports de détermination). Chaker’s description of the uses of the demonstratives has been contested by Naumann (2001) on the basis of the texts annexed to Chaker’s work.

70The systems presented in the Kable [al] varieties of Irjen (Basset & Picard 1948; Boulifa 1897: 21–23) in (22), At Mangellat (Vincennes & Dallet 1960: 93–95; Dallet 1982) in (23), and Iɛemṛanen (Aoumer 2013) in (24) are similar, although they present less distinctions than At Iraten Kabyle as described by Chaker. Sometimes the various descriptions provide different meanings for the same form; one may safely assume that this is at least in part due to the lack of precision inherent to this categorization.

(22)	Irjen [AL]
		prx	indef	dist	ana
	dem:m:sg	wa, waġi, waġini, waġik̇nint	win	wihin, wihina	winna
	dem:m:pl	wi, wiġi, wiġini, wiġik̇nint	wid	wihid, wihidak	widak

(23)	At Mangellat [AL]
		prx	dist	ana
	dem:m:sg	wa, waġi, waġini, waġikana	wahi, wahin, wahinna, wahikana, wihin, wihinna	win, winna
	dem:m:pl	wi, wiġi, wiġini, wiġikana	wihidən, wihidak, wihikana	wid, wud, widən, widak

71In general, the singular forms of the demonstratives correspond to the forms of the adnominal deictics: proximal wa corresponds to ‑a; medial winna corresponds to ‑nna, one of the variants of the distal; distal wihin corresponds to distal ‑ihin; and anaphoric winni corresponds to anaphoric ‑nni. Forms like win may be analyzed as including the residual distal marker -ən attested in words like ass-ən ‘that day’.

72The situation is different in the plural. Let us summarize the situation, as in Table 3, leaving out some lengthened variants that are irrelevant to number marking, and adding some morphological analysis. By means of <.> I will mark a division of constituent parts in these forms based on the comparison, distinguishing the vowel part (a/i in the proximal; i elsewhere), the number-neutral part of the deictic component, the number marking, and what seem to be further extensions. The use of <.> is to be understood as a historical, not a synchronic analysis of the forms; thus Iɛemṛanen widakihin is presented as w‑i.d.ak.ihi.n even though the elements dak and (i)hin are no doubt morphologically simplex in this variety.

Table 3 — Historical morphological analysis of demonstratives in four Kabyle varieties

At Iraten [AL]	prx	indef	ana	dist	med
dem:m:sg	w-a	w-i	w-i.n	w-i.hi.n	w-i.nna (w-i.n.nna?)
dem:m:pl	w-i	w-i.d, w-i.d.ak, w-i.ġa.d	w-i.d.nni w-i.d.ak.nni wi-ġa.d.nni	w-i.hi.d w-i.d.ak.ihi.n w-i.ġa.d.ihi.n	w-i.d.ak.i.nna w-i.ġa.d.i.nna
Irjen [AL]	prx		indef	dist	ana
dem:m:sg	w-a		w-i.n	w-i.hi.n	w-i.nna (w-i.n.nna?)
dem:m:pl	w-i		w-i.d	w-i.hi.d	w-i.d.ak
At Mangellat [AL]	prx		ana (1)	dist	ana (2)
dem:m:sg	w-a		w-i.n	w-a.hi w-a.hi.n w-i.hi.n	w-i.nna (w-i.n.nna?)
dem:m:pl	w-i		w-i.d	w-i.hi.d.ak w-ihi.d.ən	w-i.d.ak w-i.d.ən
Iɛemṛanen [AL]	prx		là1	là2
dem:m:sg	w-a		w-i.n	w-i.hi.n
dem:m:pl	w-i.yi		w-i.d.ak	w-i.d.ak.ihi.n

73We find the following markers related to number:

(i)	sg n vs. pl d
	At Iraten	ana, dist (possibly med)
	Irjen	indef, dist (possibly ana)
	At Mangellat	ana1 (the situation with dist and ana2 is more complicated)
	Iɛemṛanen	là1 (the situation in là2 is more complicated)

(ii)	sg ø vs. pl d.ak
	At Iraten	indef, ana, dist, med
	Irjen	ana
	At Mangellat	dist, ana2
	Iɛemṛanen	là1, là2

(iii)	sg ø vs. pl ġa.d
	At Iraten	indef, ana, dist, med

(iv)	sg nna vs. pl ø or nni
	Irjen	ana
	At Mangellat	ana2

(v)	sg ø vs. pl nni
	At Iraten	ana

75This is not at all obvious for (i), that is, the opposition of sg n to pl d as found in the At Iraten and Irjen forms w-ihin / w-ihid, and in the Irjen Indefinite form w-in / w-id. An element d is also easily recognized in the composite elements d.ak and ġa.d, which are only used in the plural.

76At Mangellat dist w-ahi(n), w-ihidən is a bit more complicated, as the n is found both in the singular and in the plural forms. It should be remarked that in neither number the presence of n is obligatory: At Mangellat also has sg w-ahi and pl forms like w-ihidak, w-ihikana, without the final n. One remarks the same adjunction of n to the plural forms in the ana2 form w-idən (corresponding to sg w-inna).

77A different situation is found in Iɛemṛanen w-ihin, w-idakihin. In these forms, the main marker of là₂ deixis is ihin in both numbers, with the addition of (i)dak to mark the plural. This is best understood as a functional change in which the final n is no more interpreted as a number marker, but constitutes a stable part of the deictic element. Note that in adnominal deixis, the singular form has been generalized (‑ihin), so the analogical pressure towards this reinterpretation is particularly strong.

78Thus, Greater Kabyle [al] demonstratives present evidence for a number opposition expressed by means of n in the singular and d in the plural. In At Iraten and Irjen forms like w-i.hi.n / w-i.hi.d the number markers follows the deictic element; hence n/d cannot be part of the initial pronominal element in the demonstrative.

4.3 Varieties from the peripheries of Kabylia

79Varieties more to the geographical periphery of Kabylia show systems that can be analyzed to some degree as going back to a similar system, even though a large number of – mostly transparent – innovations blur the picture considerably.

80We shall look at three eastern varieties, two of which are normally included in Kabyle [al], while the third is part of Tasahlit [al].

81The first system is that in Ayt Mbarek. Here we find the forms given in (25) (only masculine forms cited).

(25)	Ayt Mbarek [AL] (Genevois 1955: 47)
		prx	dist	ana
	adnom:sg	‑a	‑nanha	‑ən
	adnom:pl	‑i	‑ninhi	‑ən
	dem:m:sg	waha	wanha	win
	dem:m:pl	wihi	winhi	widak

82This variety has a final element sg ha pl hi in the distal, as well as in the demonstrative forms of the proximal. The adnominal distal is composed of a deictic marker n, followed by a (deictic) number marker a/i, followed by a repetition of the deictic marker n, followed by h and a repetition of the number marker a/i. The distal demonstratives lack the initial n, but are the same otherwise. It is difficult to see how this four-part form came into being, but it likely constitutes an innovation.

83On the other hand, the anaphoric demonstratives display forms well-known from Greater Kabylia, a singular form ending in n and a plural form ending in d extended by ak.

84In the Ihbachen dialect, we find the forms given in (26).

(26)	Ihbachen [AL] (Rabdi 2004: 72–74)
		prx	dist	ana
	adnom	‑a, ‑ayyi, ‑ayyina	‑ihin, ‑ihinna	‑nni, ‑n, ‑in,
	dem:m:sg	wa, wayyi, wayyina	wihin, wihinna	winn, winna
	dem:m:pl	wiyya, widakayyi	widakihin, widakihinna	widaknni

85This system has a number of interesting features. The forms of the proximal demonstrative show an intriguing variation between singular ayyi and plural iyya. While there is little doubt that these are extended forms based on a/i, the final vowel of the extension shows the opposite correlation to number: sg ayyi, pl iyya. While the singular form wayyi corresponds perfectly to other Kabyle forms such as waġi (< *wayyi), the final a in the plural wiyya is unexpected. One may speculate that this pattern was inspired by nominal plural patterns with forms like sg aţaksi pl iţuksa ‘car’ (Rabdi 2004: 64), where the final vowel also changes sg i to pl a.

86The demonstrative distal forms are reminiscent of those in Iɛemṛanen Kabyle [al], where the part ihin is common to both numbers, while the plural demonstrative adds (i)dak to express the plural. However, while the element dak is restricted to non-proximal deixis in Iɛemṛanen, it can occur in all deictic degrees in Ihbachen, and thus has become a general marker of plurality.

87The Tasahlit variety of Ayt Mhend (Aokas) [al] described by Rabhi (1994: 48–50) and Berkai (2013: 87–90) presents us with an extremely complicated system. Among these two sources, Berkai (2013) focuses on the demonstratives, and only provides limited information about adnominal deixis.

88In this variety, number marking in the deictic is complemented by gender marking. Both the adnominal and the demonstrative system have forms distinguishing gender and number. For ease of understanding, the gender markers have been detached by hyphenation from the other parts of the deictic in the presentation in (27).

89As one can see, this is a variety where the proximal optionally has the final d (see Section 2).

90Gender marking is according to agreement with the gender of the noun, and based on a paradigm m:sg ‑aka‑, f:sg ‑atta‑; m:pl ‑akni‑; f:pl ‑akti‑. These elements are transparently derived from proximal and distal presentative forms, such as akaya ‘here he is’, akni ‘here they (m) are’, aknan ‘there they (m) are’ (Berkai 2013: 90, 575).

92There is no demonstrative counterpart to the gender-neutral adnominal deictic ‑nnanha.

93In comparison to the adnominal forms, the distal demonstratives have the number sensitive initial vowel a/i preceding the deictic element, which is exclusive to the proximal forms in adnominal deixis. The same element is found in the anaphoric demonstratives as described by Rabhi. In Berkai’s materials, win(n) and tin(n) are singular forms (see the discussion in Berkai 2013: 90).

94Otherwise, the proximal and the distal forms are mostly the same as in adnominal deixis; the most important differences are the optional presence of an added ha/hi in the proximal, and the presence of an initial vowel a/i in the distal forms. The anaphoric forms are quite different in the demonstrative series. In the first place, there are lengthened forms with ha/hi preceding the deictic element. In the second place, some variants have a dedicated plural marker dak following the initial vowel.

95As regards the n/d number opposition, Tasahlit only provides restricted evidence. The forms with d in the gender-sensitive forms of the proximal are both singular and plural, and the d here is no doubt related to the proximal marker d, which is also found in short forms. The only clear indication of an ancient number opposition using d as a plural marker appears in the anaphoric plural demonstratives widak, tidak, widakənn, tidakənn.

96At the western side of Kabylia, the variety spoken in the Blida Atlas [al] also shows adnominal and demonstrative systems (29) highly different from the Greater Kabylia forms (El Arifi 2016: 741–743 and s.v.). El Arifi only distinguishes two degrees of deixis, proximal and distal. In the adnominal forms, only the distal forms have number opposition. El Arifi’s data stem from a large number of dialects; as it is not clear for every single form to which dialect it belongs, I will present them together in the table below, providing the dialectal information, as far as given by the source, in footnotes.

97As in Greater Kabylia, the feminine forms differ mostly by having initial t- rather w-. One notes however an enigmatic dem:f:pl form twidi (Ayṯ Ṣaleḥ; next to expected tiddi).

99Most important for our purposes, however, are the distal demonstratives without extentions, which show a variation sg -in pl -id very similar to what is found in Greater Kabylia.

5. Discussion

100The Kabyle [al] and Tasahlit [al] forms show great variability in their formation of adnominal and demonstrative deictics. Many of these elements and variants can be explained as extensions and grammaticalizations that became fixed to more simple forms. Abstracting away from these extensions and grammaticalizations, we find a number of basic patterns:

1. The proximal series has a number distinction sg a, pl i. This is found in adnominal deixis in a number of varieties, and – with few exceptions – consistently in demonstratives. By analogical extension, the a/i distinction was introduced to the other demonstratives in many varieties, especially in Morocco (Basset 1933). In these varieties, they should now be synchronically analyzed as part of the pronominal first half of the demonstrative.
2. In Kabyle [al], non-proximal deixis tends to start in a number-neutral vowel i. In demonstratives, varieties from Greater Kabylia show a distinction sg n, pl d. While this distinction has been altered elsewhere in Kabyle, and in Tasahlit [al], remnants are found all over the region, esp. in non-proximal plurals containing d. In the varieties that preserve the distinction, the number-sensitive elements may follow other deictic markers (e.g., w-ihin, w-ihid). This implies that the n/d marking is not part of the initial pronominal element of the demonstrative.

101This situation is at many points similar to what is found in Ghadames [li], which has proximal ‑o/‑i (corresponding to (1)) and anaphoric ‑e/‑id (corresponding to (2)). The most salient difference is the absence of n as a marker of the singular in the Ghadames [li] ‑e/‑id set.

Summarizing:

a. There are scattered indications all over Berber for a deictic series with (singular) *‑e (> -i). The Kabyle [al] correspondent seems to be ‑in (which could go back to *en)

b. Kabyle [al] and Ghadames [li] point to a number distinction in this series, where the plural is id.

102In view of Zenaga [mt] iʔđ (sg=pl), probably from sg *iɁ pl *iɁd (see Section 2), one may assume that this series included a glottal stop. Therefore, we may conclude that proto-Berber had a deictic series that was something like sg *‑eɁ(n) pl *‑iɁd, in addition to other deictic series. Parallel to one of the reconstructions of the proximal, one could also posit that proto-Berber had forms with and without a glottal stop in this series. As the difference between */i/, */e/, on the one hand, and */iɁ/, */eɁ/ on the other has not been maintained anywhere but in Zenaga, there is no way to prove or disprove this.

103It should be noted that Reesink (1979: II, 297) provides a different interpretation of the history of the Kabyle [al] demonstratives:

• 33 ‘One remarkable feature in Kabyle is the lack of parallelism [between the adnominal deictics and t (...)

Ce qui frappe en kabyle, est le manque de parallélisme [between the adnominal deictics and the demonstratives, and the singular and the plural. MK] (…). [A]insi les pluriels wihidən, widak ne correspondent pas, morphologiquement, aux sing. wahin(na) et winna : ils semblent, historiquement, dérivés d’un ancien pronom *wad, plur. *wid-, augmentés d’une particule déictique (-ak) ou d’orientation (-n).33

104The status of Reesink’s reconstructed pronoun *wad, *wid remains unclear. According to Reesink’s own analyses, the proximal (adnominal) deictic -a goes back to -ad, and includes a cognate to the directional deictic dd. It is difficult to see how the proximal pronoun would include the proximal deictic d and then get augmented by a distal deictic element n, as in sg wahin, which, if I understand Reesink correctly, would go back to something like *wad-hi-n. More importantly, forms like wihidən show that the element d is not adjacent to the element wi. Therefore, positing a pronominal element wid does not solve the morphological problem.

105In Berber studies, demonstratives are analyzed as a complex of several elements: a gender-marking pronominal element w/t, an element consisting of a vowel, and a specifically deictic part, which clarifies what kind of deixis is meant. The gender-marking pronominal element is absent in adnominal deixis, while the vowel element and the deictic part occur both in adnominal deixis and in demonstrative pronouns. Thus, forms like ad and ann in Tashelhiyt [ma] (argaz-ad ‘this man’, argaz-ann ‘that man’) are interpreted as containing a vowel element a and the explicit deictic markers d, resp. nn.

107Things are more complicated if one wants to include the second deictic series represented by Ghadames [li] -e/-id in this analysis. Indeed, there exist supports de détermination based on i (also e, cf. Prasse et al. 2003: 147), which express that the referent is explicitly indefinite (Galand 1974). This is found, for example in the Iwellemmeden Tuareg [ni-t] phrase in (30), where the indefinite support de détermination i is modified by a relative clause:

(30)	Iwellemmeden Tuareg [NI-T] (Prasse et al. 2003: 964)
	gámmăy-ăɣ	i	dəy	i-llíl-ăn
	look:ipv-1sg	indef:m	1sg:io	ptc-help:pv2-ptc
	‘I am looking for somebody (i) to help me’

(31)	Kabyle [AL] (Reesink 1979: II, 302)
	wi	t-ufi-ḍ,	awy	i	t	iḋ
	whoever	2sg-find: pv-2sg	carry:ao:ipt:sg	1sg:io	3sg:m:do	hither
	‘whoever (wi) you find, bring him to me’

(32)	Figuig [MA-Z] (Kossmann 1997: 201)
	wi	xəf	dd	y-us	a	ss	awy-əx
	whoever	on	hither	3sg:m-come:pv	irrealis	3sg:m:do	carry:ao-1sg
	‘whoever (wi) it fits, I will marry her’

110Therefore, I think one should remain cautious in conflating the supports de détermination related to (in)definiteness with the forms found in the deictic system. While the two categories may have been somehow related originally, this seems to be something that precedes the oldest reconstructible stage of Berber. In any case, whatever the final verdict on the relation between supports de détermination and deictics, such an analysis should include the *‑eɁ(n)/*‑iɁd series.

6. Conclusions

111On the basis of the foregoing discussion, one can conclude that adnominal deictics and demonstratives in Berber originally had a number of highly different series.

112Proximal deictics are characterized by an element sg a, pl i. This element can be followed by a further proximal extension. Depending on the variety, this is d or (possibly) *Ɂ. As a result, two or three major types of proximal marking are reconstructed:

a. a/i without a further marker

b. a/i + d

c. a/i + Ɂ (possibly; depending on the reconstruction)

113No doubt there were meaning differences between these forms, but they can no more be determined.

114In addition, there was a set of non-proximal deictics consisting of an element sg *eɁ(n) pl *iɁd. The reconstruction of the singular is uncertain. This set of deictics can be combined with further specifications of deixis, such as the element ihi.

115There were also further deictics, such as those based on a non-proximal element n(n). These have not been studied here, but as they are found almost everywhere in Berber, they are undoubtedly ancient.

116Most modern varieties have done away with the *eɁ(n)/*iɁd set of deictics. The reason behind this is obvious in many cases. If a variety has ad and id as proximals, where d functions as a gender-neutral proximal marker, there is strong analogical pressure against having a different element id in the non-proximal, dedicated to the expression of number. In varieties that lack final d in the proximal, the pressure is less strong, and it is especially here that we find remnants of the second series.

Acknowledgments

117I wish to thank Marijn van Putten and Lameen Souag for their discussions of, and comments on, an earlier draft of this article. I also wish to thank the reviewers and the editor of the volume for their pertinent and very helpful comments. Of course all errors and flaws in the argument are entirely my fault.

Abbreviations and transcription

118The following abbreviations are used:

adnom	Adnominal
ana	Anaphoric
ao	Aorist
dem	Demonstrative
dist	Distal
do	Direct Object
ext	Extended
f	Feminine
indef	Indefinite
io	Indirect Object
ipt	Imperative
ipv	Imperfective
m	Masculine
med	Medial
pl	Plural
prx	Proximal
ptc	Participle (= subject relative form)
pv	Perfective
pv2	Long Perfective
sg	Singular

119The following codes are used to refer to the countries where the cited varieties are spoken:

[AL]	Algeria
[AL-T]	Algeria (Tuareg)
[AL-Z]	Algeria (Zenatic)
[EG]	Egypt
[LI]	Libya
[LI-Z]	Libya (Zenatic)
[MA]	Morocco
[MA-Z]	Morocco (Zenatic)
[MT]	Mauritania
[NI]	Niger
[NI-T]	Niger (Tuareg)

120Transcriptions of northern Berber forms have been adopted to the Kabyle standard as used in Dallet (1982). The most important characteristic of this standard is that the mostly phonetic lenition of non-geminated plosives is not written, while the rare remaining non-geminated plosives are written by means of a dot above the letter. Thus <ġ> stands for [g], <k̇> for [k], <ṫ> for <t>, and <ḋ> stands for [d]. Different from Kabyle orthography, schwa will be written <ə>. Some sources do not write schwa; in that case it has not been added. The transcriptions of Tuareg and Ghadames follow the conventions used, among others, in Kossmann (2011) and Kossmann (2013b), respectively. Zenaga transcriptions follow the transcription system used by Catherine Taine-Cheikh. In direct quotations, the transcriptions are as in the original.