2. yV-PaRRaS/y-əFăRRăS

The most obvious formal and functional parallel is found with the yV-PaRRaS/y-əFăRRăS form. In Semitic, this form is attested in Akkadian, Modern South Arabian, and Ethiosemitic. While the vowel between the second and third radical varies between languages and between verbs, *a is probably original here (Kouwenberg Reference Kouwenberg2010: 109–12). The uses of this form in Akkadian have been studied in detail by Streck (Reference Streck1995). It amounts to an archetypical imperfective aspect, used for progressives, habituals and future events; the same uses occur in Modern South Arabian (Rubin Reference Rubin2010: 123–4; Reference Rubin2014: 142–3) and Classical Ethiopic (Dillmann Reference Dillmann1899: 153–5).

In Berber, the y-əFăRRăS form is also used for Imperfectives, and the semantic domain of the form is virtually the same as in Akkadian. Only its use for future events is not found in most Berber languages, which rather use a construction with the modal particle ad. This particle is probably an innovation that does not greatly predate the proto-Berber stage (see below), and in fact there are a couple of vestiges of the usage of the y-əFăRRăS form for marking the future. This is found in all contexts in Zenaga in addition to an auxiliary construction (Taine-Cheikh Reference Taine-Cheikh2009) and in certain types of subordinated clauses in Middle Atlas varieties (Kossmann Reference Kossmann2002: 366).

Thus, there is a perfect match, both in form and in (reconstructed) value, between Semitic yV-PaRRaS and Berber y-əFăRRăS. On the surface, it would seem evident that this form should be reconstructed for proto-Berbero-Semitic, and the idea is, of course, not new (Rössler Reference Rössler1952; Diakonoff Reference Diakonoff1988). However, for both Semitic and Berber, it has been argued that the forms are internal innovations. In the case of Semitic, the main arguments are the widely accepted absence of the yV-PaRRaS form in West Semitic languages such as Hebrew and Arabic, and its resemblance to a derived stem (called the D-stem in Akkadian studies and Stem II in Arabic studies), which also features gemination of the second radical. Thus, for example, Rundgren (Reference Rundgren1959) considers the imperfective value a reinterpretation of the original derivational meaning of the D-stem. One argument against reconstructing proto-Semitic yV-PaRRaS lies in the formal differences between the Akkadian and the Ethiosemitic forms (Kouwenberg Reference Kouwenberg2010: 117–25), especially in derived stems and quadriliteral roots. This problem does not occur in Modern South Arabian, whose reconstructible forms are very similar to those found in Akkadian.Footnote ³ Thus, reflexes of a yV-PaRRaS aspectual form are securely attested in both East and West Semitic (Akkadian and Modern South Arabian, respectively).

In Semitic studies, the debate seems to be gridlocked, with a small majority of researchers tending to reconstruct yV-PaRRaS as a proto-Semitic aspectual form (e.g. Kienast Reference Kienast2001, Lipiński Reference Lipiński2001, Huehnergard Reference Huehnergard, Deutscher and Kouwenberg2006, Weninger Reference Weninger and Weninger2011), while others follow Rundgren's lead, including such influential scholars as David Cohen (Knudsen Reference Knudsen1984–86).

Within Berber studies, comparing the Imperfective to the Semitic forms is even less accepted, and the communis opinio is that it is an ancient derived form that has become integrated into the verbal system at a late stage (Galand Reference Galand1977, Reference Galand2010: 204; Chaker Reference Chaker1995: 231).Footnote ⁴ The main internal linguistic reason for this is that there are two ways of constructing the Imperfective. With many stem and root shapes other than CCC, the gemination does not take place, and instead one finds a prefix tt-. There is little doubt that this prefix is related to, and stems from, a derived form. This is corroborated by the fact that tt- is incompatible with the causative derivation, and thus seems to have played a role in derivation (Kossmann Reference Kossmann2002). It thus makes sense, in principle, to consider the other means of forming an Imperfective as stemming from a derivation too. If one considers the situation in more detail, this line of argumentation does not look very promising. In the first place, if the putative ancient geminate derivation was anything like its Semitic counterpart, the D-stem, it would have had a strong relationship to transitive expressions (Kouwenberg Reference Kouwenberg1997). The origin of the tt- prefix, on the other hand, obviously lies in a valency-decreasing derivation, which is still attested in this form in a number of Berber varieties (Kossmann Reference Kossmann2002). Thus the two derivations underlying the Berber Imperfective would have had opposite values. A further argument against a derivational origin of the y-əFăRRăS form is the absence of residual forms. Unlike in Semitic, there is no such thing as a D-stem in Berber and, moreover, there are hardly any native verbs that have the basic stem form P-RR-S; those that are found are all evidently secondary formations, e.g. denominal verbs like Ghadames nəmmər “to bless” from tanəmmert “blessing” (Lanfry Reference Lanfry1973: 243). Even if the original derivation was lost, one would expect some lexicalized forms to be retained, and their absence suggests that no such derivation ever existed in Berber.

The main reasons that y-əFăRRăS is considered a Berber-internal innovation, however, have to do with the specific comparative and historical assumptions that the researchers adhere to. Thus David Cohen adheres to Rundgren's analysis of the Semitic forms as innovations and therefore has no reason to compare them with Berber. The case of the most influential analysis of the history of the Berber system, Galand (Reference Galand1977), is similar. Galand argues that the present situation of a non-aspectual Aorist and an aspectual Perfective (accompli) goes back to an earlier opposition where the Aorist expressed the imperfective aspect (inaccompli). What is now the Imperfective must therefore be an innovative form that took over the functions of the ancient Imperfective and ousted it into the domain of sequential and modal expressions.

Additionally, there is a typological argument: the fact that the Berber and Semitic Imperfectives are marked by gemination would be indicative of their derived, maybe even expressive, origin (Kouwenberg Reference Kouwenberg1997: 36–7; similarly, it seems, Galand Reference Galand2010: 203). Whatever one may think of this argument in theory, there is no obvious reason why it should be relevant to proto-Berber or proto-Semitic. Even if the geminated forms represent an ancient expressive formation, the development into an aspectual form may have been in place already before Berber and Semitic split.

Therefore we conclude that, in keeping with the majority view of Semitic and considering the lack of arguments against such an analysis in Berber, yV-PaRRaS can be reconstructed for proto-Berbero-Semitic in the meaning of an Imperfective.

3. yV-PRVS/y-VFRVS: introduction

The forms with the basic structure yVPRVS are structurally quite different in Semitic and in Berber. In Semitic, the question as to whether yV-PRuS, yV-PRiS or yV-PRaS is used is largely a lexical choice: most verbs only allow for one form. In Berber, on the other hand, every verb has both a y-ăFRəS and a y-əFRăS form. Instead of being a lexical choice, the two forms constitute a fully productive opposition.

In order to reach a reconstruction of the system behind these forms, one has to look in detail at the individual systems of the two language families, and attempt to make a family-internal reconstruction of these. While informed by the data from the other family, this reconstruction should not be dependent on them.

In Semitic, all three yV-PRVS forms are used in the same way. In Akkadian, they express perfective events in the past. This perfective meaning has largely been taken over by the suffix conjugation in West Semitic, but yV-PRVS still occurs in this use in vestigial forms in different West Semitic languages (Kouwenberg Reference Kouwenberg2010: 129). This meaning is therefore reconstructed for the proto-Semitic yV-PRVS form (Huehnergard Reference Huehnergard, Deutscher and Kouwenberg2006). Besides past events, yV-PRVS and derived forms are also used to express various forms of deontic modality, and the imperative is formed from the same stem, PRVS (Suchard Reference Suchardforthcoming). Both of these uses may be related to the perfective aspect.

The usage of the Berber y-ăFRəS and y-əFRăS forms will be discussed in the relevant sections below.

4. yV-PRaS/y-əFRăS

Let us first take a look at the forms with internal /a/: Semitic yV-PRaS and Berber y-əFRăS. In Semitic, the distribution of yV-PRaS is lexically determined. It consistently occurs with verbs that express a change of state in the subject, such as *yV-bsal ‘it became cooked, it ripened’ or *yV-qrab ‘he came near’. Verbs with yV-PRaS forms are therefore often referred to as the ‘stative’ or ‘intransitive’ class in Semitic grammars, but the occurrence of transitive forms like *yV-lmad ‘he learned’ and *yV-lbas ‘he put on (clothing)’ makes ‘middle’ a better term (M. Cohen Reference Cohen and Cohen1955).

Unusual semantics occur with forms of two irregular verbs in Akkadian: idû ‘to know’ and išû ‘to have’. Unlike other yV-PRVS forms, these so-called prefixed statives express a present state: thus, īdē (from *yi-ydaʕ or *yi-wdaʕ) means ‘he knows’, not ‘he came to know’; īšū (from *yi-ysaw or *yi-wsaw) means ‘he has’, not ‘he acquired’ (Von Soden Reference Soden1995: 127). The reconstruction of the stem vowel in these verbs is uncertain, as the *a-vowel of *yi-y/wdaʕ may be conditioned by the following pharyngeal, while īšū could also go back to *yi-y/wsuw; semantically, though, these verbs align well with the middle meaning of other yV-PRaS forms, supporting a reconstruction with *a. It may be that these frequent verbs preserve a stative meaning that was shared by all yV-PRaS forms at a pre-proto-Semitic stage (Kouwenberg Reference Kouwenberg2010: 468).

In Berber, y-əFRăS is used to express a category that will be called Perfective here. The Perfective is used in two basic contexts. The first usage is the expression of a dynamic, punctual, closed event, almost always in the past – this will be called the dynamic use. The second is the expression of a state, without any temporal connotations – this will be called the stative use. With verbs whose semantics inherently imply processes, the stative use can be interpreted as a result of an earlier action. On the other hand, Perfectives can also be used for states that are not conceived of (nor conceivable as) results (cf. the overview in Kossmann Reference Kossmann, Frajzyngier and Shay2012: 79 ff.); thus, a verb like ‘to go round, to surround’ may refer to somebody walking around an oasis (which would be dynamic), an army encircling an oasis (which would be dynamic or resultative), and to mountains surrounding it (which would be stative; cf. Kossmann Reference Kossmann1997: 352). Many Perfective verbs in Berber are patientive ambitransitives, that is, when used dynamically, they are transitive, and the agent is the subject of the construction, but when used as states, they are intransitive and the patient of the transitive construction functions as the subject of the intransitive construction.

The analysis of the Perfective is among the most hotly debated issues in Berber linguistics. In his classic article about the uses and history of the verbal system, Lionel Galand (Reference Galand1977) proposed that the Perfective (accompli) is inherently dynamic, and that the stative use can be interpreted as a resultative. Although couched as synchronic, it seems that the analysis is to a large extent historical in nature, as the author concedes that verbs of permanent state are different, because they have a different history. Other writers (e.g. Chaker Reference Chaker1995: 63–82) have taken issue with this analysis and point to the fact that, according to native speakers’ intuitions, the stative uses of the Perfective do not have a resultative connotation. The analysis as a resultative would be an analytical tool to reach a unified description of the Perfective rather than a semantic description of its uses. As these writers do not problematize the existence of other uses that are punctual and dynamic, the picture they draw is essentially one of polysemy: the Perfective has two different usages, whose meanings may be connected, but are not simply sub-parts of a larger, overarching, meaning. Finally, still other researchers take the opposite stance from Galand, and consider stative to be the quintessential meaning of the Perfective, and the dynamic uses to be either sub-uses of the stative (e.g. Prasse Reference Prasse2009: 237), or historical derivations from it (Mettouchi Reference Mettouchi, Naït-Zerrad, Vossen and Ibriszimow2004).

Most of these analyses are put forward as synchronic accounts: on the level of the languages as spoken nowadays, the Perfective is essentially dynamic (Galand) or stative (Prasse), or there is polysemy. It is, however, possible, and often silently intended, that these scenarios be rephrased in a historical way. If one takes the modern situation as largely polysemous, one may assume that one of the two meanings was derived from the other in some sort of historical process. Thus one could venture two hypotheses:

1. 1. The Perfective was originally a punctual dynamic form. In certain contexts, it could be used as a (dynamic) resultative. From this resultative use, a new interpretation evolved, which made the process behind the resultant state less important, while the stative meaning achieved prominence. Possibly enhanced by the verbalization of verbs of permanent state, the ancient resultative evolved into a pure stative without connotations of a preceding process.

2. 2. The Perfective was originally a marker of a state. With verbs that inherently denote a process, this entailed a resultative reading. On the basis of this resultative interpretation, the Perfective acquired a dynamic usage, which denoted the process as such, and no more just the state from which it resulted.

Both hypotheses are plausible, and no doubt attested in other language families. One notes, for instance, that the development of a stative form into a Perfective is paralleled by the West Semitic Perfect, which stems from an ancient stative form. In order to choose between one or the other scenario, one needs external evidence.

When comparing Semitic and Berber, there are important differences and commonalities. The most important difference is that in Semitic yV-PRaS is largely a lexical class, while y-əFRăS is an aspectual stem in Berber. Thus, while most Semitic yV-PRaS verbs are not opposed to a yV-PRu/iS verb of the same stem, in Berber the opposition between y-əFRăS and y-ăFRəS is relevant to all verbs. The main commonality lies in the meaning. Even though middle and stative are not the same, they are semantically close; moreover, the Semitic forms *yi-yda ʕ and *yi-ysaw (assuming the reconstruction with a is correct) could be interpreted as residual cases of a formerly more common stative meaning of the form. In this connection, it should be noted that in the earlier stages of Semitic (as attested in Akkadian), stative meanings are expressed by means of a conjugated verbal adjective (the so-called suffix conjugation), which is held to be a Semitic innovation. This brings us to what we consider the most plausible scenario in order to tie together the Semitic and the Berber forms. Originally, yVPRaS was a productive way to express a state; as in Berber, this included states resulting from the action described by dynamic verbs. In Semitic, a new form emerged, based on verbal adjectives, that took the place of the yV-PRaS stative. Particularly common cases of yV-PRaS were retained, but shifted their meaning to become a perfective, e.g. *yV-lmad *‘he is learnèd’ > ‘he learned’, *yV-qrab *‘he is near’ > ‘he came near’; cross-linguistically, such a change from a stative to a perfective – by way of an anterior or perfect, i.e. ‘he has learned’, ‘he has come near’ – is very common (Bybee et al. Reference Bybee, Perkins and Pagliuca1994: 63–9, 81–7). This resulted in a lexicalization of yV-PRaS: what used to be a productive opposition became a lexical feature associated with verbs that were particularly prone to a stative/intransitive reading. In Berber, on the other hand, the stative meaning was maintained. y-əFRăS did, however, acquire meanings beyond the stative, as the original resultative meaning associated with states from inherently dynamic verbs was extended to mean a simple perfective without any association to state or result.

5. yV-PRuS and yV-PRiS/y-ăFRəS

The third set of forms that concern us here is yV-PRuS, yV-PRiS / y-ăFRəS. In Semitic, yV-PRuS and yV-PRiS belong to the same tense as yV-PRaS. The stem vowels -u- and -i- are characteristic of transitive verbs, which only rarely have yV-PRaS forms. Based on the designation of yV-PRaS as a middle form (see above), we may refer to yV-PRuS and yV-PRiS as active. In Akkadian and Classical Arabic, yV-PRiS tends to be associated with verbs that express punctual action, like *yV-sriq ‘he stole’ (Aro Reference Aro1964: 191). In other languages, yV-PRuS has largely replaced original yV-PRiS forms, as in Hebrew and Aramaic, or the two forms have merged phonologically, as in Ethiosemitic.

Intransitive yV-PRuS and yV-PRiS forms also occur. A class of intransitive yV-PRuS verbs is found in all classical Semitic languages; this mainly contains atelic verbs of motion, like *yV-rqud ‘he skipped around’, and other verbs expressing atelic activities (Aro Reference Aro1964).Footnote ⁵ In Akkadian, there are also adjectival verbs with yV-PRiS forms, like i-kbit ‘it became heavy’, while adjectival yV-PRuS occurs in both Akkadian, like i-tqun ‘it became certain’ and Arabic, like ya-kbur ‘(that) it be(come) great’. These verbs are usually associated with primary adjectives of the PaRiS and PaRuS patterns (Kouwenberg Reference Kouwenberg2010: 60–65). In Akkadian, such adjectives form the Stative of their associated verbs, e.g. kabit ‘it is heavy’, taqun ‘it is certain’; in West Semitic, they have developed into Perfects, e.g. Classical Arabic kabur-a ‘it was/became great’. It seems likely that the -i- and -u- vowels in the yV-PRVS form of these verbs were taken over from the adjectives that formed the core of their paradigm. Discounting these adjectival verbs, then, we may conclude that yV-PRiS was originally associated with transitive verbs, especially those with punctual semantics, and yV-PRuS was associated with both transitive verbs in general and atelic intransitive verbs. The main difference with yV-PRaS is that the latter is used with verbs where the subject is greatly affected, while the subjects of yV-PRiS and yV-PRuS verbs are not particularly affected. This distinction between middle yV-PRaS and active yV-PRiS or yV-PRuS is clearest in the cases where a middle verb coexists with an active verb from the same root, e.g. *yV-xrab ‘it fell into ruin’ vs. *yV-xrib ‘he ruined’.

In Berber, y-ăFRəS is the vowel scheme of a form commonly called the Aorist (aoriste; cf., however, Prasse Reference Prasse1972–73, who uses imparfait). The Aorist is found in a number of well-defined contexts, and its presence is to a large degree predictable from syntactic context.

In the first place, the Aorist is used with normal imperatives and injunctives. It should be noted, however, that habitual imperatives are formed with the Imperfective. In the second place, it is found after the modal particle ad, which marks that the event described in the following verb is not (yet) realized. Again, it is also possible to use the Imperfective after ad for habits.Footnote ⁶ In the third place, it is used in sequential constructions, in which the Aorist signals that a preceding earlier Mood or Aspect is still valid. The details of this construction vary between languages (Galand Reference Galand, Jungraithmayr and Müller1987): in some languages any aspect or mood can be continued by the Aorist (e.g. Tashelhiyt), while others have restrictions.

The Aorist thus has a wide range of only slightly related uses, most of which can be predicted from the syntactic context. This led Galand (Reference Galand, Jungraithmayr and Müller1987) to the conclusion that it is best analysed as a non-aspect, a form which has no inherent aspectual value, but which takes on the aspectual value set by its preceding context. In the case of the sequential use, this preceding context is the earlier verb form; in the case of the non-realized mode, it is the particle ad that provides the non-realized meaning. While this analysis is largely accepted by scholars working on North African varieties of Berber, it was strongly criticized by Prasse (e.g. Reference Prasse2009: 259 ff.). Whatever one's opinion about the synchronic analysis, it is clear that the system as it stands now is not very ancient. The construction with ad is almost certainly an innovation from shortly before the proto-Berber period, ad being either an ancient pronominal form (Vycichl Reference Vycichl1992) or an ancient conjunction (Prasse Reference Prasse1972–73 III: 238–9) – or both – that developed into a preverbal marker, probably in a similar way to how French developed injunctives of the type qu'il vienne ‘may he come’.

All authors seem to agree that the current distribution of the Aorist is different from earlier on, and that it has been ousted from its earlier functions by innovative forms. Galand (Reference Galand1977) identifies this function as an ancient Imperfective (inaccompli). In his view, the modern Perfective would have been more or less similar in usage to what is found today, while the y-əPăRRăS form would be an innovation. Once this innovation had happened, the Aorist lost most of its functions, and was left merely as a non-aspect or, alternatively, became a modal rather than an aspectual marker.

The actual uses of the Aorist – which is modal rather than aspectual, or neither – do not seem to provide many arguments to tie it to an ancient Imperfective rather than to another aspectual meaning. The main reason for Galand to consider it an ancient Imperfective seems to be the fact that he already reconstructs a Perfective (y-əFRăS), while he considers the attested Imperfective (y-əFăRRăS) an innovation; thus, in his reconstruction there is a gap at the imperfective meaning, which could be filled by the Aorist. Once one chooses different scenarios concerning the history of y-əFRăS and y-əFăRRăS, there is no intrinsic reason to consider y-ăFRăS an ancient Imperfective anymore.

As argued above, we reconstruct y-əFăRRăS for proto-Berbero-Semitic, and consider y-əFRăS to be an ancient stative form that only later expanded to express punctual events. From this point of view, the natural locus for the Aorist would not be the Imperfective (which is already accounted for), but the Perfective. Modal usages, such as imperatives, would be part of the Perfective meaning, which fits in well with the fact that orders typically demand a complete action and exhibit little temporal structure.

The reconstruction of the Berber Aorist y-ăFRəS as a perfective form, which included modal uses such as the imperative, conforms well with the situation in Semitic. As we saw, in proto-Semitic, yV-PRu/iS can be reconstructed as a dynamic perfective; moreover, imperatives are formed with PRVS forms.Footnote ⁷

7. A note on the prefix vowel

So far, we have ignored the issue of the proto-Semitic prefix vowel, simply giving the prefixes as yV-. The reconstruction of this vowel is controversial. Several different systems are attested (see Table 5). In Akkadian, the prefix vowel varies with person, with a occurring in the first person singular prefix a- (Old Akkadian ʔa-) and the second person prefix ta-, but i occurring in the first person plural prefix ni- and the third person prefix i- (Old Akkadian yi-). In Northwest Semitic, one and the same prefix vowel is used for every person, but different verbs take different prefix vowels: *a occurs before -PRiS and -PRuS and *i before -PRaS (-PaRRaS not occurring in Northwest Semitic). This distribution is commonly known as the Barth–Ginsberg Law, after its discoverers (Barth Reference Barth1894: 4–6, Ginsberg Reference Ginsberg1932–33: 382–3, Reference Ginsberg1939: 318). The Barth–Ginsberg Law also operates in some colloquial dialects of Arabic (e.g. Najdi, Ingham Reference Ingham1994), but Classical Arabic has a as the prefix vowel in nearly all cases; Classical Ethiopic, on the other hand, has ə everywhere, which must come from *i or *u. Given the alternations seen in both the Akkadian and the Northwest Semitic prefixes, this homogeneity probably results from one prefix vowel being levelled.Footnote ⁹

Table 5. The Semitic subject prefixes

Which of these distributions, if either, should be reconstructed for proto-Semitic? Barth believed the stem-based alternation to be original; more recently, this view has been defended by Testen (Reference Testen1992, Reference Testen1994, Reference Testen2000), who invokes possible traces of the Barth–Ginsberg Law in Akkadian. Others prefer to project the Akkadian prefixes back to proto-Semitic and see the Barth–Ginsberg Law as a later innovation (Hetzron Reference Hetzron1976; Hasselbach Reference Hasselbach2004; Huehnergard Reference Huehnergard and Gianto2005). As both accounts have their advantages and difficulties, the issue has reached something of a stalemate.

If the proposed connection between proto-Semitic yV-PRaS / proto-Berber y-əFRăS and proto-Semitic yV-PRuS, yV-PRiS / proto-Berber y-ăFRəS is accepted, this provides external evidence that can resolve this question. As proto-Berber *ə corresponds to proto-Semitic *i or *u and proto-Berber *ă corresponds to proto-Semitic *a, we may reconstruct the proto-Berbero-Semitic Perfective as yaPRuS and yaPRiS and the proto-Berbero-Semitic Stative as yiPRaS. This suggests that, in some prefixes at least, the Barth–Ginsberg Law was already present in proto-Berbero-Semitic, and, by extension, in proto-Semitic.