2.1 The expression of non-verbal predication

Following Higgins (1979)’s classic study and subsequent literature (e.g. Mikkelson 2011; Roy 2013) we can distinguish three major types of copular constructions, according to whether or not the subject and the “complement” XP of the copula are referential, as shown in Table 1 (adapted from Mikkelson 2011: 1810) and (1).

The primary focus here is on the predicational sub-type of copula constructions.¹ We take the predicational copula construction to be a sentence type in which the lexical or contentful predicate is some non-verbal element (Mikkelson 2011: 1805). The English examples in (2) are predicational copula constructions in which the sentential predicate is respectively adjectival, nominal and prepositional. In such clauses, the forms of be are copula verbs, that is, linguistic elements which appear in some sort of mediating or linking role between subject and predicate in predicational sentences in which the main semantic predicator is a non-verbal element. Hence we use the term predicational copula construction to include both property ascriptive examples, as in (2a) and (2b), and locational clauses, as in (2c).

It is often stated that in copula constructions the copula element is totally devoid of meaning, at least in predicative copula constructions (Hengeveld 1992: 32; Pustet 2003: 5), and, in some accounts, also in equative and specificational constructions (Partee 1987). We assume that the predicational copula has no inherent lexical semantic content but simply plays a role in semantic composition, i.e. in applying the predicate to the argument (Partee 1987; Roy 2013) and in carrying tense information, although such matters are orthogonal to our concerns here.² In languages with multiple copulas, a choice between competing forms generally corresponds to some semantic property, and hence may be thought to realize the competing values of that property, or is to constrain any such property to be present, depending on the precise details of the approach adopted.

Crosslinguistically we find a great diversity in the syntax of copula constructions. Languages differ in terms of the diachronic source and synchronic syntactic status of copula or linking elements; copulas may be full words or affixes, and common sources include pronouns, deictic particles and verbs (Devitt 1990; Pustet 2003). They also differ as to whether, and under what conditions, they require a copula clause to contain an overt copula or linking element. For example, no overt copula is required in predicational copula constructions with a present tense interpretation in Russian.

Arabic also exhibits this tense-conditioned morphosyntactic alternation between the absence and presence of a copula element in predicative copula constructions. We will use the term null copula to refer to copula-free copula constructions, without commitment to any particular syntactic analysis.

When a language has multiple copulas, a range of different factors may govern the choice of copula. Choice of copula can be determined by various clausal features such as tense and aspect or polarity, but also by the morphosyntactic category of the predicate phrase itself, as in Bambara (Niger-Congo) which exhibits a four-way choice between copula forms, dependent on the category of the predicate (Pustet 2003: 46). Equally, it may be determined by the distinction between locational and non-locational clauses, as in Kinyarwanda (Bantu) (Jerro 2015) or by other semantic or pragmatic characteristics of the predicate, clause or subject argument (see Pustet 2003: 45–53). For example, Kuuk Thaayore (Paman) (Gaby 2006: 460–477) has five verbs used as optional copula verbs (nhiin ‘sit’, than ‘stand’, wun ‘lie’, yan ‘go’ and yoongke ‘hang’). The default choice of copula in ascriptive and locative copula clauses for higher animates is yan ‘go’, with the use of a different copula introducing additional connotations, which may or may not relate to the postural sense itself. The choice of an optional copula for animate subjects in ascriptive copula clauses is determined by the canonical posture of the animate entities in question.

2.2 Split copula systems

Split copula systems implicating a semantic/pragmatic distinction between permanent or inherent properties and temporary, contingent or temporally-bounded properties are quite widely attested. Stassen (1996) notes that an alternation between a null and a locational copula encoding for nominal predicates occurs in several Carib languages (Apalai, Hixkaryana, Macushi) and the Dravidian languages Tamil, Telugu and Kannada. The locational encoding for predicate nominals is associated with non-habitual, contingent or temporary states. As is well known, Irish makes use of forms of two distinct verbs in copula constructions, the copula verb is and the so-called substantive verb bí (Stenson 1981; Carnie 1995; Doherty 1996). Prepositional and adjectival predicates, whether interpreted as permanent, inherent properties, or as temporary states, properties or locations appear with bí (glossed simply as BI) in copula constructions. In the modern language, the copula use of is with adjectival and prepositional predicates is highly circumscribed and a vestige of Old Irish (Doherty 1996: 36–7; Stenson 1981: 99).

Nominal copula constructions make use of the copula verb is, which is variously characterised as predicating inherent qualities or defining characteristics.³

What is of interest is that a clear contrast arises between a nominal copula construction with is (5)–(6) and one in which the nominal predicate is introduced by the preposition ar ‘in’ (4a), where the substantive verb bí is used. Several different characterisations of the associated semantic distinction are suggested in the literature; Stenson (1981: 94–5) takes nominal copula constructions with is to be defining or characteristic, and those with bí to be suggestive of the attainment of a state, and which is more anchored in time. Carnie (1995) and Doherty (1996) relate the contrast to the distinction between individual-level and stage-level predicates. Roy (2013) characterises the semantic distinction differently, suggesting that is is limited to maximal predicates, that is, predicates devoid of “perceptible spatio-temporal subpart properties” (Roy 2013: 90), while bí occurs with situation-descriptive predicates which are dense, that is, which hold continuously for every sub-interval of the eventuality, and habitual or generic sentences.⁴ Consistent with these various characterisations of the difference, the nominal copula construction with is in the past tense shows lifetime effects, such that (7a) entails that Seán is dead, while the prepositional nominal construction with bí does not (7b).

The distribution of ser/estar as copula forms in predicative constructions in Spanish is also sometimes characterised in terms of the distinction between individual-level and stage-level predication.⁵ Maienborn (2005a) offers a discourse-based account of the distinction within DRT, Luján (1981) (inter alia) takes an aspectual view associating estar with the feature [+perfective], and Roy (2013) proposes that copula estar occurs with dense predicates while predicative copula ser marks maximal predicates and those which are non-dense, that is, have spatio-temporal subpart properties and are not required to hold continuously for every sub-interval of the eventuality. A good overview of the facts for Spanish and the issues and challenges faced by different theoretical accounts is provided in Camacho (2012). Predicative NPs occur only with ser, unless preceded by the prepositional marker de, in which case they occur with estar, and receive a particular, transient, interpretation, as in the contrast in (8).

Many adjectives will occur felicitiously with both ser and estar in predicative copula constructions—in the former case a permanent, inherent or intrinsic property is ascribed to the subject, while in the latter case, the property might be temporary, contingent or situation-descriptive. Such a contrast is provided in (9). Similarly, the absolute/transient distinction also applies with PP predicates, in general. With locational PPs, if the subject is a movable entity, in which case the location may be temporary, locative prepositions occur with estar, as in (10).

This section has provided background to contextualise our discussion of Arabic predicative copula systems in subsequent sections. We have observed that many copula systems display a split which is grounded in a distiction between permanent, inherent or immutable properties and those which are temporary, contingent or episodic. In the following sections we first outline the picture for Arabic dialects as generally described. We see that two conditioning factors are standardly thought to be relevant to this split copula system—the clausal feature of tense and the distinction between predicative and non-predicative clause types. We then turn to Maltese, where the distribution of copula forms is more complicated, being sensitive to additional conditioning factors, including copula construction type (i.e. predicative versus non-predicative; locational versus non-locational), the clausal feature of tense, and the distinction between enduring and temporary properties.

5.1 Libyan

Consider now these examples from Libyan, which involve a verbal element which is the ACT.PTCP form of what is etymologically the root ‘sit’, but which synchronically is the verb meaning ‘stay; remain’. In Libyan, gāʕəd does not mean ‘sitting’ at all (Pereira 2008, as also reported in Rubin 2005). We have examples such as the following, where our glossing and translation are intended to maintain the insights from the original descriptive source, which in some cases is indicative of a degree of ambivalence about the analysis of these items. To this end, we have provided the original free translation in French alongside our own English rendering, and reflect the original French gloss restant as ‘stay.ACT.PTCP’, and se trouvant and étant as BE, which is to be understood as indicative of a copula function in such contexts.

An example such as (25), glossed as ‘stay.ACT.PTCP’ (restant in the original), is perhaps suggestive of a lexical predicate gaʕad ‘sit.PFV.3SGM’ with the semantically bleached lexical meaning of ‘stay; remain; continue to be (in a location)’, combining with a locational modifier to give a meaning of CONTINUE TO BE IN A LOCATION. If we associate the CONTINUE TO BE sense with gāʕəd in (25), then this might suggest (taking a conservative view) that we are dealing with some bleached lexical function of the participle form in this variety, rather than a use that necessarily has a copula function. In isolation, then, (25) is consistent with a view of gaʕad as a bleached lexical predicate with the meaning shown in small capitals in (26), where the location phrase is a selected dependent or a modifier.

Similarly, gāʕəd in (27) (glossed as ‘stay.ACT.PTCP’, reflecting restant in the original French glossing line), is part of an intransitive clause, and involves no overt locational phrase at all. This suggests a clearer lexical usage meaning stay, remain, continue to be in a location (rendered in French as ‘be there’). Clearly, since gāʕəd is used with no potentially predicative dependent, it must itself have some lexical meaning.¹³ Remaining cautious in our identification of potential copula uses, we might informally describe the use of gāʕəd in (27) as in (28).

On the other hand, the following examples of gāʕəd with locational predicates, whether glossed as stay.ACT.PTCP or BE.ACT.PTCP on the basis of Pereira’s (2008) translation, can at least equally well be seen as copula constructions in which the locative phrase itself (whether it involves a temporary or a more permanent location) contributes the main predication, as shown in (33). On this view, the verbal element is itself semantically empty of lexical content. It does not express the location, and may or may not express a continuative aspectual value.¹⁴

Examples such as (34) present their own puzzle, the issue being whether what we see here is a copula use extending beyond locational predication to use in the ascription of contingent or temporary properties, or whether what we see here is the figurative extension of a stative predicate stay, remain beyond the locational domain, just as in English John remained/stayed silent throughout this diatribe. Clearly, distinguishing between these is a more than delicate matter, but we note that (34) is glossed and translated as a copula construction by Pereira (2008).

Other examples are more questionable. In (35) Pereira in fact glosses the ACT.PTCP form gāʕəd as an adverbial ‘still’ (toujours), which might suggest that it is only a continuative aspectual value which is maintained.¹⁵ This is however still consistent with viewing it as a temporally-anchored copula.

Given this fact about Libyan, an anonymous reviewer rightly asks how we might resolve the question of whether gāʕəd in the examples above has a true copula function, or simply represents a figurative extension of stay/remain to mean something like is still? They observe that since expressions such as stay, remain and is still involve a presupposition that the state holds as a continuation of a previous state, and a simple copula such as is lacks this presupposition, we might use a context to test where such a presupposition is ruled out, to test whether gāʕəd is still felicitous. The examples in (36), suggested by the reviewer as counterparts to (30) and (34) respectively, are such contexts, and hence provide further evidence for the conclusion that we are indeed dealing with a copula function of gāʕəd.

The conclusion that can be deduced from the above array of data and usages of gāʕəd is that there seems to be clear evidence for a locative copula use of this ACT.PTCP form in this variety, apart from broader semantic bleaching of the lexical posture predicate itself. Furthermore, in relation to the copula function, there may additionally be some evidence of extension beyond locative predicative constructions.

5.2 Chadian

Chadian Arabic (for which a major source is Abu Absi & Sinaud 1968, a pedagogical/descriptive manual) is another variety in which the ACT.PTCP form gāʕid does not mean ‘sitting’ at all, as is also the case with its verbal counterpart. Rather, it is used as a locational verb ‘be present’, i.e. ‘is situated/is located/exists’, as illustrated through (37) below.

In this variety too, it is thus clear from the outset that the lexical predicate gāʕid has undergone a significant degree of bleaching and widening. Beyond such bleached uses of the lexical function of gāʕid, we find many examples of gāʕid with locational predicates, both examples which can be construed as expressing a more permanent location (38)–(39), and those expressing temporary or contingent locations (40). The subject in such structures can be both animate and non-animate. The following examples illustrate this for both declaratives and interrogatives. Once again we choose to gloss the copula function of gāʕid in this variety as BE.ACT.PTCP.

It is worth noting that the examples above are not simply semantically bleached lexical usages meaning ‘stay’ or ‘remain’. Even though they are all locational clauses, they do not have the additional “continuative” nuances which would follow on that view.

gāʕid is clearly an optional strategy with such locative PPs, occurring optionally with locational predicates of all sorts, in both declarative and interrogative clauses. As observed specifically for (41b) (Where are you? is surely asking about a contingent/temporary location), the presence of a zero copula is also available in the context of temporary locational predications, and hence a zero copula is possible for both temporary and permanent locations. The observed split distribution of gāʕid and the zero copula parallels that discussed for Maltese.

The plethora of examples above are all locative PPs. The examples which we have of non-locative uses of otherwise locative prepositions are ones with min ‘from’, which expresses origin in such constructions and hence a permanent or characteristic property, rather than a temporary (or permanent) physical location. The examples that we have ((42)–(43)) involve a null copula, which is consistent with what we seem to see here in terms of an emerging split copula system.

The above data concludes our observations about the emerging distribution of the copula function of gāʕid from the available Chadian data.

5.3 Levantine region

Above we have demonstrated that locative copula uses, with a possible extension to some non-locational predications as well, as in Libyan Arabic, is present in dialects other than Maltese, with a concomitant loss of the central lexical meaning of ‘sitting’. This lexical meaning is preserved in some other dialects. The question arises as to whether the development of the locative copula use goes hand in hand with the loss of the ‘sitting’ meaning for gāʕid ‘sit.ACT.PTCP’. We will below see that this is not a necessary prerequisite and indeed that gāʕid is synchronically emerging as a copula in the locative constructions of a number of vernaculars where gāʕid, as well as its associated verb-form, concurrently still maintain their lexical meaning ‘sitting’ and ‘sit’, respectively. The example in (44), from Palestinian Arabic (specifically Kufr al-labad, Tulkarem), illustrates the ambiguity which results synchronically from the development of gāʕid as a locative copula and the concurrent maintenance of the lexical ‘sitting’ meaning in this variety.¹⁷

Examples from the Levantine region denoting the emergence of a copula function include (45), denoting an ad hoc temporal location, in Negev Arabic. Further Palestinian data in (46) illustrates how beyond temporary locational predications (such as (44)), time-stable ones can also appear in the context of the optional use of kāʕid.¹⁸

Use of this form is equally possible in negative clauses involving a time-stable locational predication, as in (47) (the negative marker miš may come before or after the copula, but appears only once).

5.4 Kuwaiti/Hijazi – Gulf region

Other varieties which preserve the ‘sit’ lexical meaning of gāʕid include those in the Gulf, such as Kuwaiti, Qatari and the varieties of Saudi Arabia.¹⁹

The examples in (48) illustrate the ACT.PTCP with the ‘sitting’ meaning from Qatari, Kuwaiti, and Urban Hijazi, while (49) provides examples of the more desemanticised or bleached reading of ‘staying’ or ‘remaining’ in Kuwaiti and Urban Hijazi.²⁰

The examples of primary interest here are the following from Kuwaiti (pc, Duha Alaskar), demonstrating a further degree of grammaticalisation. In these cases, the distribution of gāʕid is that of a locative copula restricted to contingent or temporary locations.

The restriction to temporary locations is shown by the ungrammaticality of the following examples, if the copula gāʕid is used.

Further evidence illustrating that the use of gāʕid does not extend to properties, whether permanent, or temporary, in Gulf dialects, here represented by Kuwaiti, is the ungrammaticality of the data in (52) (pc, Duha Alaskar).

Parallel data illustrating the presence of gāʕid in temporary locations is also found in Hijazi (53). Additionally, we also find a broader extension to contingent states in this dialect, illustrated in (54).

The range of data above from some Gulf dialects supports our claim that the posture active participle gāʕid is grammaticalising as a copula, even if still in parallel distribution with the null copula strategy. While the Gulf dialects seem to be very clearly grammaticalising the presence of gāʕid in temporary locative predications, Urban Hijazi appears to have gone beyond locative constructions, and data is available where the use of gāʕid has evidently extended to other contingent states as well.

5.5 Data summary

In this section we have suggested that beyond the phenomenon of desemanticisation of the lexical predicate into a cluster of meanings in the general domain of ‘remain’ or ‘stay’ (CONTINUE TO BE AT LOCATION), gāʕid has developed a copula function within predicational locative structures across non-peripheral/core Arabic vernaculars very similar to the grammaticalisation of gāʕid as a copula in (peripheral?) Maltese. Some vernaculars permit both temporary/contingent and permanent/stable locations with gāʕid. Others, such as Kuwaiti, distinguish between temporary/contingent locations using a null copula or gāʕid, and permanent locations, where gāʕid cannot figure. We have also pointed to possible evidence for the extension of this locational copula strategy beyond cases of locational predication in Libyan and in Urban Hijazi. The distribution of what we have argued to be copula uses across this range of dialects indicates that this grammaticalisation is found both in vernaculars where gāʕid maintains, and in those where it has lost, its original lexical posture meaning of ‘sitting’. Table 3 provides a summary overview of the data presented in the subsections above.

5.6 Implications as to grammaticalisation

While the characterisation in §5.5 summarises the data which we have argued support our claim that an additional copula is emergent within the copula systems of Arabic vernaculars, in what follows we consider what the ramifications of this data are from a diachronic perspective. That is, we seek to understand how the grammaticalisation of a copula that stands in both a morphosyntactic and semantic contrast with the zero copula strategy and the pronominal copula may have come about and developed in the vernaculars. Given the lack of a historical written record for vernacular Arabic, and the fact that this innovation also does not figure in Classical Arabic texts, our methodology in addressing this question is essentially comparative, considering the variation across the different Arabic varieties, but also informed by a typological perspective.

We have argued above that data from a considerable number of different Arabic varieties supports the view that the pre-existing split copula system has undergone further complexification with the emergence of an additional copula form so that the resultant system marks the sorts of semantic distinctions among different types of eventualities which have been described for other copula systems (such as those of some Celtic and Romance languages), often under the label of the distinction between stage-level and individual-level predication.²¹

The backdrop to the innovation is a copula system with a two-way choice between the null and pronominal copula in the present tense. Following the recruitment of the posture active participle gāʕid into the system, a three-way split copula system in the present tense emerges. If we consider the distribution of data across the different dialects to be indicative of the trajectory of change in progress, the most striking observation is that all of these dialects allow the presence of the emergent copula with temporary/contingent locations. This possibility remains in free variation with the zero copula strategy. In Maltese, on the other hand, the use of the copula qiegħed in such structures is notably itself becoming the default strategy. Beyond this core, there is variability in the occurrence of the newly grammaticalised copula gāʕid in locative predicational structures: it is not the case that both time-stable or temporary locational anchorings are found with the newly grammaticalised copula in all dialects, or that inanimate subjects are necessarily found across the board. This sort of variability is of course in the very nature of change in progress.

As a very first approximation, this distribution appears to align itself with what is observed crosslinguistically from a typologically diverse set of languages. First, suppletion and renewal of copula elements is a common phenomenon, within and beyond Indo-European (Irslinger to appear: 6). Second, posture verbs are found crosslinguistically as a source of copula elements (Lesuisse & Lemmens 2018: 44, Devitt 1990: 104). Third, crosslinguistically, it is a very common pattern for languages to encode nominal predication in a distinct manner from locational predication (Stassen 1996: 482, Irslinger to appear: 38), and bodily posture verbs are frequent sources for the encoding of locational predication (Newman 2002: 7). The suggested trajectory which we envisage for Arabic, rooted in the salience of the locational element of the meaning of a posture verb is hence consistent with what we know about the diachronic development of copulas from posture verbs crosslinguistically. A case in point is the grammaticalisation of the Portuguese, Catalan and Spanish copula and auxiliary estar from Latin stāre ‘stand’ and its gradual encroachment on ser, a well-researched case of posture verb grammaticalisation, where locational predication has played a key role (see e.g. Falk 1979, Vañó-Cerdá 1982, Remberger & González-Vilbazo 2007, Brucart 2012, Carvalho 2010, and many others). The historical record here supports a trajectory in which the newly grammaticalising element (estar) first established its place alongside ser in locative constructions, and then extends to other uses, as Batllori & Roca (2012: 86) observe: “We can see that in the twelfth century there is feature syncretism concerning the use of ser and estar to express the same value only in locative constructions, whereas in the thirteenth century it [estar] extends to stage-level copulative, resultative passive, and existential sentences … a syntactic change that conveys replacement of ser by estar is taking place progressively” [in locatives and stage-level predicate copulatives] Batllori & Roca (2012: 86).

The salience of the locational element is pinpointed as a key factor in the development of copula forms in a range of languages, including Spanish and Turkish in Devitt (1990). The grammaticalisation path which is at the core of his proposal is shown in (55), which takes account of the fact that a language may go on to develop a general copula. For Turkish, which does not generally make use of a copula in the present tense, Devitt (1990) suggests that the notion of temporariness has led to the development of a modal, presuppositional flavour associated with the use of the addition of the enclitic -dir, itself derived from the posture verb meaning ‘stand’, as shown in (56) (see Devitt 1990: 113).

In the light of these considerations and the central role of locative predications in the data we have presented, we suggest that a natural hypothesis is that the grammaticalisation is triggered primarily through a semantic extension from the encoding of mere ‘sitting’ to ‘be located somewhere’, where a PP predicate is most natural. This eventually gave rise to the copula + locative PP combination, alongside the pre-existing zero copula structures. Irrespective of the (internal) distinct stages the different Arabic dialects display, in their grammaticalisation and establishment of gāʕid as a locative copula, there is clear evidence from the same dialects for further extension to a general contingent/ad hoc marker, as it comes to express particular/temporary states. Hence we suggest that the Arabic dialects provide evidence for the cline of incremental change and grammaticalisation shown in (57), although we leave open for further research a more fine-grained understanding of the temporally-anchored nature of the predicates.

It should not be taken as a deficiency to the path being posited here that a further developmental extension to a temporary state function of the copula follows the prior establishment of the locative function of the copula, independent of the variation observed in the use of the locative copula itself. Rather, it is in fact in line with observations from different Romance languages with split copula systems, where fine-grained studies (Remberger & González-Vilbazo 2007; Batllori & Roca 2012) of these languages reveal subtle differences over the choice of copula, and which do not invalidate the general trajectory proposed. While for instance Portuguese and Spanish both make use of the split between the copulas ser ‘be’ and estar ‘contingent BE’ to express a distinction between permanent versus temporary states in the context of adjectival predicates, their individual use of the copulas in locative structures differs. While Spanish makes use of estar in all locative contexts, Portuguese still uses both copulas in locative contexts, such that ser is maintained to mark permanent locations, while estar is used in the contexts of temporary physical locations (58).²²

The split in the locative constructions in Portuguese thus essentially reflects the same split that obtains in the context of adjectival predicates. A similar, if not exactly parallel split use of the copulas in locative constructions, is also true of Catalan (see Batllori & Roca 2012). The pattern of difference which we see between these Romance languages, including in particular the locative use of estar in Spanish, is relevant to the use of the new copula with all locational predicates in Maltese. These differences show that as the languages or dialects develop along the same grammaticalisation cline, different nuances or components of meaning become or remain focal. For some discussion of this in relation to Romance, see Remberger & González-Vilbazo (2007).

The path in (57) that we reconstruct as the developmental path for Arabic involves a change from an active participial of a posture predicate to a copula with various functions and domains of applicability, with variability across the dialects. This path of change, we claim, did not take place on its own. Rather, there is evidence of a distinct but related development in which the same posture predicate maintains its status as a lexical predicate, yet undergoes distinct stages of semantic bleaching as hypothesised in (59). These different stages are posited on the basis of the range of variation that exists across the dialects, where for instance we observe the loss of a ‘sitting’ reading in Libyan and Chadian, varieties which use the same lexical form to mean ‘exist, be situated’. On the other hand, Levantine and Gulf dialects make use of gāʕid with both a maintenance of the original ‘sitting’ sense, as well as the more desemanticised sense of ‘staying, remaining’. In these dialects a further bleached existential reading is however not (as yet) recorded.

If (59) is on the right track, it displays key parallels with the cline in (57) as the ‘staying, remaining’ is clearly closely related to the locative copula part of the latter path. We keep (59) distinct from (57) for our Arabic data, because the former are not copula functions of gāʕid, but rather, bleached lexical extensions, and the development of an existential use does not in principle need to be correlated with the emergence of a predicative copula. However, the fact that those dialects which do have the existential use also have the copula use is suggestive of a close connection, raising the possibility that the locative copula stage in the trajectory in (57) might actually encompass two stages, the first of which involves the bleached lexical extension to a ‘stay, remain’ meaning which also underpins the development of the existential usage. This possibility is discussed in more detail in Camilleri & Sadler (under review).