The Typology of Progressive Constructions in Arabic Dialects

: This paper describes and analyzes the strategies used in Arabic dialects to encode the progressive aspect, in order to show their commonalities and what sets them apart from one another, as well as to situate them within cross-linguistic tendencies. Drawing on a wide variety data, the paper shows the different ways the progressive aspect was (or is being) grammaticalized in Arabic dialects in light of the typologically common paths of grammaticalization. These paths involve for some dialects the reorganization of their aspectual categories, and thus provide an interesting perspective to look at the evolution of aspectual systems in Arabic varieties.


Introduction:
The PROGRESSIVE ASPECT (PROG) "conveys the idea that an event is progressing dynamically over a time frame opened up by an utterance" (Mair 2012). It is a subcategory of the IMPERFECTIVE ASPECT (IPFV) that has a more specific meaning (Comrie 1976:25), as represented in fig. 1. As such, there is an asymmetric entailment between the two aspects, whereby a progressive verb form is automatically imperfective, but an imperfective verb from-by virtue of its wider semantic range-can have a progressive reading as well as other readings.
Progressiveness differs from imperfectivity in semantic (Comrie 1976:33) and typological as well as historical (Mair 2012) respects. For instance, while progressive verb forms cannot have habitual readings, imperfectives do not exclude such a reading (Comrie 1976:34), as shown in the French example in (1) which has two possible readings in the absence of contextual indicators.
(1) FRENCH IMPERFECTIVE [Comrie 1976:34] Chaque jour à cinq heures le poète écrivait un poème Every day at five hours the poet write.IPFV.PST.3SG a poem a. 'Each day at five the poet wrote a poem.' HABITUAL b. 'Each day at five the poet was writing a poem.' EVENT-IN-PROGRESS (1) contrasts with (2) where the progressive marker en train de restricts the reading of the verb to the EVENT-IN-PROGRESS reading exclusively. Another semantic restriction that applies to progressive verb forms-but not imperfective ones-is that the use of PROG is usually reserved for eventive predicates, that is activities, achievements and accomplishments, but not states. Comrie (1976:35) attributes this semantic incompatibility to a "contradiction between the stativity of the verb and the non-stativity essential to the IPFV PROG Figure 1: Deo's (2019) representation of PROG progressive." For instance, he is painting is an acceptable utterance no matter the context, whereas he is believing in God sounds odd except if licensed by special discourse conditions.
In terms of historical developments, it is generally accepted that the PROGRESSIVE category is historically "younger" than other aspectual categories (Deo 2019) and is thus "rarer in the world's languages than the PERFECTIVE/IMPERFECTIVE distinction" (Mair 2012). This follows from semantic considerations given that progressive readings are able to be expressed by imperfective verb forms, while progressive verb forms are semantically more restricted and cannot, as such, express the wide range of imperfective readings.
As for typological considerations, languages are divided into three categories from the perspective of their overt marking of the PROGRESSIVE aspect (Deo 2019): (3) PROG-MARKING IN THE WORLD'S LANGUAGES: a. Zero-PROG designates a language that does not have a dedicated progressive marker and needs "contextdependent strateg [ies] for encoding and recovery of sub-meanings" (Deo 2019) of the imperfective verb form. b. Emergent-PROG designates a language that has developed a progressive marker but does not use it obligatorily.
As such, it has "partially context-dependent and partially explicit marking" (ibid) of progressiveness. c. Categorical-PROG designates a language that has a dedicated progressive marker that is categorically used and as such has an "explicit marking strategy for encoding and recovery of sub-meanings" (ibid).
These are not simply independent synchronic states according to which we can classify languages but are historically ordered stages of a four-stage cycle (the PROGRESSIVE-to-IMPERFECTIVE cycle). The fourth stage (generalized-PROG) involves the weakening of the semantics of the progressive marker to include other imperfective readings, which leads to a language becoming zero-PROG once again.
The goal of this paper is to provide a study of Arabic dialects with regards to their progressive marking strategies. The diversity of Arabic dialects is often obscured in typological studies by the use of the umbrella term "Arabic," and the fact that this term usually refers to the Modern Standard variety. This does not do justice to the immense richness of this dialect continuum with respect to any linguistic feature one chooses to study, and deprives linguistic theory from a plentiful source of data. In this paper, I will describe the historical underpinnings of the development of the PROGRESSIVE category in Arabic dialects ( §2), survey a representative sample of Arabic varieties with respect to their encoding of PROG, classify them according to the typology in (3) as well as to their various strategies for expressing the PROGRESSIVE aspect ( §3), and show where some of these dialects are situated historically on the PROGRESSIVE-to-IMPERFECTIVE cline ( §4).

The progressive in Semitic and Arabic:
The PROGRESSIVE aspect is not fully grammaticalized in old West Semitic languages where the main distinction is between PERFECTIVE and IMPERFECTIVE aspect. We can safely say that the older stages of Arabic also lacked a grammaticalized progressive form, judging by the Classical Arabic (CA) and Central Semitic data: The morphological encoding of the progressive arises at some point in the history of Arabic but is not inherited from the proto-language, which simply has an imperfective verb form conveying multiple readings that can be disambiguated by contextual clues. Cross-linguistically, the imperfective has three different readings (Deo 2015:4), all of which are available in Classical and Modern Standard Arabic, as well as in some dialects.
The following examples from Gulf Arabic 1 illustrate the range of readings available to the imperfective verb from in these dialects.
(4) GENERIC/HABITUAL READING: [Johnstone 1967:143]  While Gulf Arabic at the time it was documented by Johnstone (1967) does not seem to have had a special marker for the EVENT-IN-PROGRESS reading, 2 many dialects have innovated various ways to do encode it at different times in their histories. These parallel innovations led to other changes in the tense-aspect-mood (TAM) systems of these varieties.

The rise of progressive marking and the reorganization of TAM systems:
In his book on Semitic verbal systems, Cohen (1984) shows how the innovative verbal systems of Arabic dialects arose by means of trying to encode the "concomitance" or simultaneity of two actions, be it in the imperfective or the perfective. In Classical Arabic (CA), while the prefix conjugation still bears all the readings of the imperfective in (4)-(6), the active participle-when used as a verb-is the main form expressing simultaneity in the realm of the imperfective (this applies only to dynamic transitive predicates). As for stative verbs, "le participe désigne un état dont ne sont pas envisagés les termes," that is, continuous aspect (Cohen 1984 :277). By our definition, then, CA is a zero-PROG language since its strategy for conveying the EVENT-IN-PROGRESS reading is nevertheless context dependent: That reading is only accessible for certain types of predicates in the right context, which is not what emergent-PROG and categorical-PROG languages look like, as shown in (3). Since the active participle is not a grammaticalized progressive from in CA, it can take on different meanings depending on the lexical aspect of the verb it combines with-just like the imperfective-where dedicated progressive markers in languages that have them can generally only combine with dynamic predicates (Comrie 1976: 35). 3 Arabic dialects, by contrast, have grammaticalized more aspectual distinctions, and have developed various strategies to encode the progressive aspect. This development is actually responsible for many Arabic dialects' TAM systems today, since the integration of the of auxiliaries to mark finer aspectual distinctions has led to the reorganization of those systems. Specifically, it is the grammaticalization of the progressive aspect that led to the synchronic states of affairs that we witness today. For instance, many dialects now have a marked distinction between modal and non-modal forms, because their older progressive forms became indicative imperfectives (e.g., b-imperfectives in Levantine Arabic), in the diachronic PROGRESSIVE-TO-IMPERFECTIVE shift. It is precisely this kind of development that will be thoroughly explored in §4, after looking at the synchronic situation of progressive marking in Arabic dialects in §3.
Already in the 1980's, Cohen makes very insightful remarks on the various types of dialects that exist with regards to their TAM systems as well as on the grammaticalization paths of progressives. In this paper, I will add more data to Cohen's survey, and analyze it from a typological perspective. In fact, Cohen reports many markers as optional or not fully grammaticalized, so this paper also provides an update on these dialects, on top of being a classification project in the larger domain of tense/aspect typology.

The Progressive in Arabic Dialects:
This section provides a twofold classification of Arabic dialects with respect to their PROG-marking. First, it considers to what degree there is a grammaticalized PROG-marking strategy in each of the dialects that I have surveyed. Second, it shows what different strategies these dialects use in order to encode the EVENT-IN-PROGRESS reading.

Typological Classification:
For the typological classification, I use Deo's (2015) terminology shown in the introduction, where dialects are classified in three different categories depending on their PROG-marking. While not perfect, the criteria that I used to differentiate between different types of dialects are the following: • Zero-PROG means that the dialect in question has no specific way of encoding the PROGRESSIVE aspect and the EVENT-IN-PROGRESS reading always needs to be recovered from contextual clues. • Emergent-PROG means that the dialect in question has (a) specific way(s) to encode the PROGRESSIVE aspect but it is optional such that the imperfective morphology can still convey the EVENT-IN-PROGRESS reading, that is, this reading is not usually blocked when the imperfective is used. 4 • Categorical-PROG means that the dialect in question has a way of encoding PROGRESSIVE aspect that is always used in those contexts and thus the EVENT-IN-PROGRESS reading is usually blocked for the imperfective verb form.
In Table 1, I classify Arabic dialects in three groups according to the criteria explained above. Each dialect whose description and data were available to me is followed by a reference to a numbered example with the relevant data that the reader is invited to look at. Some of these examples are in the appendix in §6, and the rest are shown in the main text for illustration purposes.  (7), Lahej, Aden, Ḥabbān, Abyan. (39) The Arabic spoken in Wādi Ḥadramawt in Yemen is a good example of a zero-PROG language: 6 While (7a) is a generic statement about Salim and Salih going to different schools, an event that has no specific ties to the time of the utterance, (7b) refers to an ongoing action at the time of the utterance rather than a habitual or generic event that obtains repeatedly. These two readings are different as shown by their English translations with a simple present and a present progressive respectively, but they are expressed in this dialect with the same unmarked imperfective verb form. In this sense, the EVENT-IN-PROGRESS reading is recovered by the hearer.

EVENT-IN-PROGRESS
By contrast to Wādi Ḥaḍramawt and dialects similar to it, Khartoum Arabic and other emergent-PROG dialects have innovated a special progressive marker-here the active participle of the verb gʿd 'to sit'-which is used as an auxiliary before an imperfective verb to express ongoing action at the time of utterance, as in (8c). This marker, however, is seemingly optional, as the EVENT-IN-PROGRESS reading can still be expressed (and retrieved) without it in (8b), where only the imperfective preceded by the indicative marker b-is used. 7 This same form is also the one that is used for the habitual statement in (8a).

EVENT-IN-PROGRESS
The last dialect group which the majority of varieties that I surveyed belongs to is the categorical-PROG one. These dialects have innovated a special marker for the progressive, and they cannot regularly use the bare imperfective for such a reading because that form is blocked in such a context. One of these dialects in Tunis Arabic, which is reported to have two progressive markers: the auxiliary made of the active participle of qʿd 'sit', and for transitive verbs that do not require a specific preposition, the preposition fī 'in' can be used before the direct object. While in (9a), the bare imperfective can only convey the reading where the subject writes letters habitually, the use of the preposition fī 'in' in (9b) indicates that the event is ongoing at the time of the utterance. This strategy is interesting because it does not really modify the verb like the auxiliary does in (9c); rather, it is a syntactically restricted marker that can only be used with transitive verbs due to its prepositional nature. (9c) shows the use of the auxiliary qāʿid with an intransitive verb, and (9d) shows the use of qāʿid in conjunction with the preposition fī with a transitive verb. While common in transitive sentences, the cooccurrence of these two progressive markers is not obligatory. 9 Having explored the three groups of dialects with regards to the level of grammaticalization of their PROG-marking, we can now turn to the different strategies they recruited for their PROG-markers.

The sources and morphosyntactic forms of the Arabic Progressive constructions
Cross-linguistically, there are four major lexical sources for progressive constructions (Bybee et al. 1994: 128-129). They are listed here from the most common to the least common.
7 The bare imperfective in this dialect is restricted to modal and embedded contexts and the preverb b-seems necessary in indicative contexts, except if other preverb like the future marker ḥa or auxiliaries like the progressive marker gʿd are present (Dickins 2011). 8 If a transitive verb requires a specific preposition before its object, then that preposition is kept and the auxiliary qāʿid is used to encode the progressive. However, McNeil reports (2017:171) that when that preposition is ʿla 'on', then it can optionally be replaced by fī. 9 McNeil (2017:177) reports that in transitive sentences, fī is always obligatory and qāʿid may be omitted, saying "the use of qāʿid without fī for transitive verbs is ungrammatical" (ibid:187). My native speaker intuitions differ in this respect and while I admit that fī is preferred in transitive contexts, I would not go as far as claiming that its omission is ungrammatical.
(10) LEXICAL SOURCES FOR PROGRESSIVE MARKING: a. Locative elements in the verbal auxiliary, e.g., a posture verb like 'sit' or a verb expressing "the notion of being in a location without reference to a specific posture like 'stay'," or locational pre/postpositions. b. Copula functioning as an auxiliary and a non-finite form. c. Movement sources such as 'come' and 'go'. d. Reduplication of one of the verb's syllables. e. Other sources include verbs meaning 'keep on' or continue as auxiliaries, etc.
The lexical sources in (10) can be perfectly integrated as affixal progressives (e.g., a preverb) yielding a synthetic progressive construction, or they can be periphrastic (Mair 2012: 809). Both morphosyntactic types relate to the sources described here, but it is easier to recognize the lexical source in periphrastic constructions since there is less phonological reduction. For example, the Levantine progressive marker ʿam is from the hyperbolic participle ʿammāl 'doing/ making': The non-reduced participial form is still available in many dialects, and while it is semantically bleached, its origin is more easily recoverable compared to the reduced form ʿam or even ʿa in other dialects. Table 2 gives an overview of emergent or generalized-PROG dialects. These are organized according to the different strategies described in (10), showing for each dialect the lexical source of its progressive marker as well its morphosyntactic type, followed by a reference to an example with the relevant data for each dialect.
Table 2 shows that Arabic dialects most often form progressives with posture verbs, gʿd and gls 'sit' being the most common. This kind of observation is usually obscured in the typological literature because Arabic tends to be treated as a monolith, typically categorized as a zero-PROG language (Deo 2015: 41). As Table 2 makes clear, however, the Arabic dialects in fact align with (10a), which is the most common strategy for marking progressive aspect cross-linguistically. So, far from being zero-PROG, Arabic dialects are robustly progressive-marking, and, although each encodes the progressive in a different way, they are collectively in line with known cross-linguistic tendencies.
While most of the dialects have recruited a locative form or some other common strategy for the progressive, Levantine and Taza Arabic use verbs meaning 'do' or 'make', which is arguably unusual. 11 Another dialect-the Bāṭina sedentary dialect of Oman-stands out as well since it has the option of using the active participle of gls 'sit' as an auxiliary for progressive readings (11a), but also has another, highly unusual, strategy: Al-Balushi (2016:119) reports that if the bare imperfective is used in the VSO word order as in (11b), only the habitual reading is available, whereas its use in the SVO word order as in (11c) conveys the progressive reading. In this section, I have shown the variety found in Arabic dialects with regards to the different sources each dialect recruited for its PROG-marking strategies as well as how grammaticalized each strategy is according to dialect. Having looked at the synchronic state of affairs, in the next section I will explore how the synchronic situation developed historically. In particular, I will describe Arabic dialects from the perspective of the PROGRESSIVE-to-IMPERFECTIVE shift. Cohen (1984) had already noticed that many dialects introduced a new aspectual subdivision by innovating progressive markers, which resulted in the reduction the semantic scope of the bare prefix conjugation (imperfective) that was inherited with all its possible readings shown in (4)-(6). He also noticed a now well-known grammaticalization path, namely the PROGRESSIVE-to-IMPERFECTIVE shift (Deo 2015), shown here in (12).

The Progressive to Imperfective Shift
(12) THE PROGRESSIVE-TO-IMPERFECTIVE SHIFT [Deo 2015:20] a. XIPFV To move from (12a) to (12b), the language must first recruit a source to "innovate a new functional category" (Deo 2015:20), going from having a single imperfective form X that has multiple readings and not marking the progressive at all to marking it optionally with the new form Y. This transition happens gradually in "speech events which require disambiguation" (Deo 2015:23) between, e.g., generic and ongoing actions, such that some of the ways speakers disambiguate may become very frequent. The most "conventionalized and reliably frequent" (ibid) form Y ends up being chosen for this function.
From (12b) to (12c), this innovative form Y becomes categoricalized, meaning that its use becomes obligatory in progressive contexts. At this stage, forms X and Y coexist, and each has its own context of use, Y being restricted to EVENT-IN-PROGRESS readings. This categoricalization process is pushed by pragmatic principles: The form Y is more marked (imperfective verb + progressive marker) than the form X (imperfective verb). Therefore, if a speaker uses X, then the hearer can infer that they do not mean Y, given the availability of Y as a more restricted and thus more informative option. Thus, the hearer concludes that if the speaker had meant the more specific meaning of Y, they would have used the corresponding form Y (cf. note 4). "The conventionalization of this implicature pattern is the categorical-PROG system" (Deo 2015:24).
Finally, from (12c) to (12d), the form Y starts being applied for events other than progressive ones, until its semantic content becomes so broad that it contains all of the readings of the imperfective, and the language ends up with the Y form only. Form X may disappear completely, or it may be relegated to specific uses, as in many Arabic dialects where the bare imperfective now has a modal or subordinate function. It is worth mentioning here that, while (12a) and (12d) represent two different diachronic stages, they are identical from a synchronic perspective. Both refer to languages that only have one form expressing all imperfective readings, but generalized-PROG is a diachronic term, whereas zero-PROG is a synchronic one.
While there is no consensus on how generalization of PROG to imperfective happens (cf. Deo 2015:21-22 for discussion), what is clear is that speakers tend to overextend the use of PROG in certain discourse situations. Because the PROG-marker is occasionally licensed in non-canonical EVENT-IN-PROGRESS contexts, learners may eventually generalize that these are canonical contexts in which PROG may occur. For instance, when one says Bill is smoking these days, something like EVENT-IN-PROGRESS is meant, but it has a sense of iteration, not of a single event in progress (i.e., Bill is in the habit of smoking again). Contrast this with the more canonical Bill is smoking right now. The latter is infelicitous if he is not currently smoking, while the former may be uttered even if Bill is not smoking at the moment of speech. When we say Bill is smoking these days, the meaning is habitual, but it is habitual in a particular, progressive sense (i.e., different from Bill smokes or Bill is a smoker). From such contextually licensed, extended uses of PROG, speakers may generalize that PROG is licensed in habitual contexts of all kinds. While the recruitment a PROG marker in the first place (from (12a) to (12b)) was motivated by speakers' desire for specificity, there is an opposing force at work, by which the original functional domain of the PROG marker is inevitably broadened by contextually licensed extensions of its use. This type of process contributes to the change from (12c) to (12d), whereby a categorical-PROG grammar becomes a generalized-PROG one.
This process of overextension is at work in categorical-PROG dialects of Arabic as well, such as Tunis Arabic. Though Tunis Arabic's progressive marker qāʿəd is incompatible with habitual meanings of the type he gets drunk every day, (13) shows that qāʿəd is licensed in a special type of habitual reading, that of a "restricted habit," in which the adverbial phrase h-al-ayyēmēt 'these days' yields an iterative progressive sense.

DEM-DEF-days
'He is getting drunk every day these days.' The process of generalization seems to be happening in Ṣanʿāʾ Arabic in the 90's, as documented by Watson (1993:72-79). alli some-3MPL say.IPFV.3MSG allaði 'Some say alli and some say allaðī' GENERIC While (14a) has a bare imperfective for a habitual reading and (14b) a prefixed imperfective for an event-in-progress reading, (14c) is a generic statement using both types of imperfectives. The preverb bi-recruited for the progressive marking is not limited to that function anymore: It is used for a non-progressive event such as the one in (14c). This is interesting because it looks like (14) is a good example of the transition between (12c) and (12d): Here YPROG (prefixed imperfective) and XIPFV (bare imperfective) still coexist, but YPROG is in the process of generalizing towards YIPFV, as evidenced by the fact that it is coordinated with the XIPFV in the same meaning in (14c).
It is valuable to be able to see these transitions with data such as (14), where there clearly is a generalization happening in speech communities. This generalization is also traceable historically for other dialects. As mentioned in §2.2, the PROGRESSIVE-to-IMPERFECTIVE grammaticalization path is responsible for certain Arabic tense-aspect systems we see today. For instance, the indicative preverb b-of Levantine dialects as shown in (15) comes from a historical progressive form, 12 for which Cohen (1984:294) found attestations in older texts, as in (16)  In the same family of dialects, the Arabic spoken in Cyprus went through the cycle only once, and is now synchronically a zero-PROG dialect but diachronically a generalized-PROG dialect. As such, all indicative imperfectives in this dialect have the preverb p-(<*bi) and all the readings of the imperfective are available to this form as shown in (17).
(17) CYPRIOT ARABIC [Borg 1985:76,173]  The same historical argument could be made for dialects of Moroccan Arabic that have an obligatory preverb ka-/ta-(probably derived from the copula kān) with imperfective verb forms whose readings are "compatible with generality, repetition, and concomitance" (Caubet 2011). These dialects also innovated optional auxiliaries for progressive readings (cf. (27) in the appendix).
This also goes for Juba Arabic whose indicative gi= preverb almost certainly originates in a progressive marker according to Manfredi (2017:97 (7), others went through the whole cycle, generalizing their progressive marker to a plain imperfective marker, as Vanhove (2011) reports for the Jabal Yazīdī variety. Similarly, in Ṣanʿāʾ Arabic as discussed above, the bi-prefix is used with stative verbs (Watson 1993: 81), a sign that it has generalized completely.
In Tripoli Arabic, which has the same progressive strategies as Tunis Arabic, it seems that the progressive marker fī 'in' is extending to habitual contexts, such as the one in (20b). As can be seen in the data in (20), it seems that fī is generalizing into an imperfective marker, perhaps with specific nuances of durativity, 16 while the auxiliary gāʿid is developing copular meanings. 17 The most surprising aspect of this survey is the discovery that some dialects might be able to start generalizing their progressive marker before they become categorical-PROG languages. One prediction of the theory about the PROGRESSIVE-to-IMPERFECTIVE shift is that each step in (12) needs to be fully reached before the language begins a new transition, and this is indeed what we typically observe in Arabic dialects. However, as far as I can tell from Davey's (2016) description, it seems that Costal Dhofari goes against this typological tendency. On the basis of the fact that the seemingly progressivizing preverb b-is optional for the EVENT-IN-PROGRESS reading (compare (21c) and (21d)), 18 by the criteria given in §3.1 we would categorize this dialect as emergent-PROG. However, as can be seen in (21b), the preverb b-is also optionally used in habitual contexts, which is a quality typical of generalizing-PROG languages. So Coastal Dhofari contradicts the grammaticalization path as predicted by our framework.

on-us 'He is lying to us' EVENT-IN-PROGRESS
It is unclear however, if this is truly an originally progressivizing prefix that is generalizing: The morpheme b-in the Peninsula has been the object of controversy, and there seem to be at least two preverbs b-with different etymologies in the area (Restö 2011). It is very well possible that the data of Coastal Dhofari does not at all contradict typological tendencies; rather, it has two different optional homophonous but etymologically distinct b-preverbs. This would of course make it an emergent-PROG and not a generalized-PROG dialect, if one of those b-is a progressive marker. Another option is that b-is neither a progressive nor habitual marker but some other kind of preverb that is optionally attached to the imperfective verb form. Peninsular dialects are reported to have b-preverbs with volitive nuances (Bettega 2019: 140). More attention on the different contexts where b-is (not) used in this dialect will shed light on its meaning and tell us whether Coastal Dhofari really goes against theoretical predictions.

Conclusion:
In this paper, I have shown that Arabic dialects display a sort of microcosm of linguistic typology. Studying certain features in Arabic from a typological perspective shows how much variety exists in this branch of the Semitic family that is often lumped together as a single language. Even dialects in the same geographical region can vary widely as regards the developmental stage of their PROG-marker, as seen in Yemen and Oman. Moreover, some dialects provide precious transitional data which gives us an opportunity to better understand how a language moves along the PROGRESSIVE-to-IMPERFECTIVE grammaticalization path. In addition, the Arabic data possibly displays unexpected patterns with regards to typological tendencies and theoretical predictions, as the Coastal Dhofari data might suggest.
With this variety taken into account, there is no motivation behind referring to "Arabic" as a single language. In this particular case, the typological literature has treated Arabic as a zero-PROG language (cf. Deo 2015:41), despite the fact that "Arabic" with regards to this feature, as well as other linguistic features, is by no means monolithic. Finally, this typological study contributes to larger scale efforts of documenting the world's languages tense and aspect systems, where in the future, I hope to see different Arabic dialects represented rather than the highly reductive "Arabic."