Pluractionality in Hittite

2.1 The suffix -ške/a-: previous research and open issues

The existence in Hittite of a suffix -ške/a- has been pointed out since the beginning of Hittitology (cf. Hronzý 1917; see Oettinger 1979: 315–29 for the morphology).^{[3] [4]} Ever since its individuation, the puzzling behavior of the suffix has fueled an intensive scholarly debate (see Cambi 2007 for a comprehensive overview of previous scholarship). Several are the issues that the study of this suffix has raised: what are the functions that the suffix performs? Can one single out a core meaning of the suffix as opposed to more secondary and/or marginal ones? Does the suffix operate within the domain of lexical or grammatical aspect?

Before turning to discuss different approaches to the study of -ške/a-, we summarize some salient aspects of its distribution. In the first place, it must be stressed that the suffix is optional, since unsuffixed verbal forms can freely occur in contexts in which they have the exact same meaning as -ške/a- forms (cf. Daues 2009; Dressler 1968: 84−85).^[5] Since optionality is a key feature of derivation as opposed to inflection, the optional use of the suffix -ške/a- easily points to its derivational nature. However, there are contexts in which the occurrence of the suffix is virtually obligatory. These mostly feature

distributive expressions such as UD-at UD-at ‘day after day’, ITU-mi ITU-mi ‘month after month’, GE₆-ti GE₆-ti ‘night after night’, MU-ti MU-ti ‘year by year’, lammar lammar ‘moment by moment’, uddanī uddanī ‘word by word’ […] with 1-an 1-an ‘one by one’. (Hoffner and Melchert 2008: 320)

Conversely, according to Bechtel (1936: 62) and Cambi (2007: 121−122) -ške/a- is incompatible with adverbs meaning ‘X times’. This view is however partly unwarranted: whereas it is clear that the majority of adverbials of the type ‘X times’ occur with simple verbs, a few -ške/a- forms do occur in such contexts (see e.g., example (21a) below). Another constraint concerns the lexical aspect of the base verb, inasmuch as the suffix is unavailable to stative verbs (cf. Bechtel 1936; this is interpreted as a case of neutralization by Cambi 2007). Finally, suffixed forms of the supine (a non-finite verbal form) are systematically employed in an ingressive construction ‘begin to X’ in combination with the verbs dai- ‘put’ and tiye/a- ‘step’ (Hoffner and Melchert 2008: 322, 338).

These facts are well known, and they have been used by various scholars in support of different views about the nature of -ške/a-. As already pointed out by Cambi (2007), approaches to -ške/a- can be roughly divided into three groups: those that view the suffix as marker of actionality, those that view it as mostly aspectual, and those that treat it as belonging to both domains.

The first school of thought regards the suffix -ške/a- as affecting the lexical semantics of verbs and as operating on their lexical aspects. Advocates of this view deny a connection with imperfectivity and instead propose that the suffix is predominantly employed to indicate iterativity or distributivity. Individual accounts differ in their details, as is shown by the somewhat diverse terminology adopted by different scholars: ‘iterative-durative’ (Gusmani 1965: 79), ‘iterative’ (Pedersen 1938: 132; Sommer and Ehelolf 1924: 21–22), ‘distributive’ (Neumann 1967: 24), ‘iterative-durative-distributive’ (Friedrich 1960; Rosenkranz 1966), ‘durative-distributive’ (Kammenhuber 1969: 217), ‘iterative-durative-intensive’ (Kronasser 1966). As we discuss below, Dressler’s (1968) account can also be grouped together with actional approaches.

This approach, that has recently been labeled the actional hypothesis (cf. Cambi 2007), has been argued against by several scholars who instead suggest treating the suffix as a full-fledged marker of imperfective aspect. Under this second view, that has been called aspectual hypothesis, the suffix -ške/a- is the marked member of a privative opposition, since it can only encode imperfectivity as opposed to base forms, which are neutral with respect to their aspectual interpretation and can have either a perfective or an imperfective reading (Bechtel 1936; Cambi 2007; Puhvel 1991).

The two approaches are not entirely incompatible, and a more nuanced view, called hybrid hypothesis by Cambi (2007), has been advocated whereby -ške/a- forms operate at the interface between lexical and grammatical aspect, and are associated both with imperfective (progressive, continuous, habitual) and perfective (inceptive) aspect (Hoffner and Melchert 2002, 2008; Melchert 1998).

Among actional approaches, Dressler’s (1968) contribution deserves a special mention, as this work stands out for its particularly innovative proposal. Dressler’s approach can be considered actional in that he essentially views -ške/a- as affecting the verbs’ lexical semantics. However, with respect to other actional approaches, Dressler shifts away the focus from the aspect vs. actionality distinction proper, and rather suggests that the suffix can be interpreted as instantiating verbal plurality. Dressler’s account is extremely insightful, in that he proposes a new comprehensive model that easily encompasses all detected usages of the suffix in a principled way. Unfortunately, this view has found little echo in recent literature.^[6] We believe that this line of reasoning deserves more attention, as we discuss extensively in Section 4.

In Table 1 we summarize some of the different existing descriptions of -ške/a-. The purpose of this table is not to provide a comprehensive comparison between earlier approaches to -ške/a-, but merely to highlight that while virtually all authors agree in assigning certain functions to the suffix, e.g., durative and iterative, the status of other functions, e.g., inceptive, remains debatable.

2.2 Approaches to aspect

In Section 2.1, we have seen that most of the scholarly debate concerning -ške/a- has revolved around the issue of whether the suffix operates as a marker of either lexical or grammatical aspect. However, these two notions are on turn highly complex and controversial, and different authors adopt at times inconsistent definitions of these in their assessment of the usage of -ške/a-. This results in a great deal of confusion across different contributions, since it is not always clear that the authors employ terms such as ‘aspect’, ‘imperfective’, ‘iterative’ to refer to the same linguistic notions. An interesting case in point is the status of the notion of ‘iterative’. As a matter of fact, proponents of the actional hypothesis argue that -ške/a- operates at the level of actionality precisely because it encodes iterativity, which in their view is a typical Aktionsart. This assumption is however not trivial, since in other frameworks iterativity is viewed as belonging to the domain of aspect instead (see e.g., Tatevosov 2002: 333). This shows that far from being theory neutral, one’s understanding of the function of -ške/a- is heavily dependent upon the underlying theory of aspect.

In the remainder of this section, we briefly survey current approaches to aspect and actionality, in order to ground our own analysis on an explicit and well-defined terminology.

As is well known, aspect constitutes one of the mostly debated components of verbal semantics, and it has sparked the interest of scholars from across all different linguistic traditions.^[7] As pointed out by Sasse (2002), approaches to aspect can be sorted out into two categories: unidimensional and bidimensional. Unidimensional approaches “proceed from the assumption that there is only […] a single conceptual dimension in terms of which aspectual phenomena […] can be analyzed and described” (Sasse 2002: 202). By contrast, bidimensional approaches postulate a clear-cut distinction between two dimensions, which are commonly referred to as lexical aspect (also actionality or Aktionsart) and grammatical aspect respectively (see e.g., Bertinetto 1986; Bertinetto and Delfitto 2000).

Lexical aspect, or actionality, refers to the inherent temporal structure of events (see also Ježek 2016: 121−126). In his seminal work, Vendler (1957) first established the classification of verbs into four classes based on their lexical aspect (Table 2). Vendler’s classification is based on the features of duration, dynamicity, and homogeneity. This basic taxonomy has been further refined and new sub-types have been proposed, including ‘semelfactives’, e.g., cough, and ‘degree verbs’, e.g., ripen (see Bertinetto and Civardi 2015; Botne 2003; Croft 2012).

Grammatical aspect, instead, refers to the “ways of viewing the internal temporal constituency of a situation” (Comrie 1976: 3). Therefore, whereas lexical aspect is lexically stored in the verbal meaning, grammatical aspect is in principle independent from verbal semantics, and is a grammatical category that may receive overt morphosyntactic marking in the world’s languages. The most basic aspectual distinction is that between perfectivity and imperfectivity. The former can be broadly conceived as an external view on an event as a whole, whereas the latter implies that the action is viewed from within (cf. Comrie 1976). Readings typically associated with imperfectivity include e.g., habitual and continuative (cf. Bybee et al. 1994).

Bidimensional approaches are thus based on the distinction between lexical and grammatical aspect and view the two as being fully independent from one another. The bidimensional approach has prominently featured in the discussion of aspect in ancient IE languages (see Inglese forthcoming) and, as the various positions summarized in Section 2.1 show, it also underlies most of the debate on the interpretation of Hittite -ške/a-. It is not our goal here to challenge the status of bidimensional approaches in general. However, one should keep in mind that bidimensional approaches set out from a number of non-trivial assumptions, chiefly the more or less tacit assumption that there exists a strict distinction between the domains of lexicon and grammar. Notably, the need of such sharp grammar vs. lexicon distinction has been convincingly challenged by frameworks such as Construction Grammar, which propose a gradual continuum from the ‘lexical’ to the ‘grammatical’ pole (cf. Croft 2001; Goldberg 1995). It follows that, in Construction Grammar, a more nuanced understanding of the divide between ‘lexical’ and ‘grammatical’ aspect is required, whereby aspectual notions commonly seen as belonging to either of the two are placed in a single conceptual domain. In this respect, Construction Grammar pursues an essentially unidimensional approach to aspect.^[8]

In the remainder of this paper, we follow Dressler (1968) and pursue a different approach to the description of the behavior of -ške/a-. As we argue, there is little point in maintaining a distinction between those functions of -ške/a- that are more drawn to either lexical or grammatical aspect, and we propose that all functions can be accounted for in terms of pluractionality instead. Under this approach, typical aspectual features such as “continuative” are conceived as belonging to the same conceptual domain as more actional-like ones, such as “iterativity”. In this respect, even though setting out from a different angle, the perspective that we adopt here goes in the direction of Croft’s (2012) unidimensional approach to aspect.

4.1 Functions of -ške/a- as a PM

In this section, we survey the functions of -ške/a- that are attested in our corpus and classify them according to the typology of PMs laid out in Section 3.

The first function is the continuative, as illustrated in (7).

In (7), the suffixed verbal form zīnuškizzi ‘is making cross’ indicates an ongoing event as seen in its development. The continuative function roughly corresponds to the classic imperfective function according to advocates of the aspectual hypothesis. From a discourse perspective, it is often the case that -ške/a- forms “provide a background to the following action” (Josephson 2008: 137). Such a backgrounding function also fits well with the occurrence of several suffixed verbs in a row in narrative texts, as well as in relative clauses (cf. Daues 2009) and in subordinate clauses introduced by kuitman ‘while’, as in (8) (cf. Daues 2010; on the semantics of kuitman see Inglese 2016: 83), and is more generally compatible with a well-known backgrounding value of imperfective verbal forms (cf. Comrie 1976: 3; see further Caudal 2012 and Carruthers 2012 for extensive references).

From the perspective of aspectual construal, the event denoted by continuative -ške/a- verbs is construed as an activity, i.e., as a “durative, unbounded process” (cf. Croft 2012: 60). Depending on the lexical semantics of the predicate, one can further distinguish between those activities that are incremental, i.e., directed activities, such as warš- ‘harvest’ > warškanzi ‘they are harvesting (the crops)’, and those that do not depict an incremental process, i.e., undirected activities, such as šanḫ- ‘search’ > šanḫiškit ‘he was looking (for your death)’.

The second function is the frequentative one. In this case, the suffixed verb indicates an event that takes place on several occasions on a long time frame, as opposed to the simple verbal form that indicates a single event. This use of the suffix is exemplified in (9), in which a frequentative reading is strongly supported by the occurrence of the distributive adverbial expression MU-ti MU-ti ‘year after year’:

The frequentative use of -ške/a- also includes those cases in which the verbal form is paired with the inhibitive negation lē, and the whole construction refers to the interruption of an otherwise iterated event ‘stop … -ing’ (Hoffner and Melchert 2008: 319−320). This use is exemplified in (10):

Iteratives are similar to frequentatives, but they differ in that the former refers to events that take place multiple times over a short time frame. This usage is exemplified in (11), where the form ḫukkiškizzi ‘invokes’ depicts a situation in which the ‘physician’ repeats the invocations several times in a single occasion.

Another function that is connected with the notion of repetition of the same event is habituality. What characterizes habituals as opposed to both iteratives and frequentatives is that with habituals the repetition is construed as taking place in a time frame that is typical for the event. As an example of the habitual reading of -ške/a-, one may consider the form pišker ‘they used to give’ in (12). Notably, in (12) the habitual reading is fully compatible with the occurrence of karū ‘formerly’, which indicates that the event typically took place in the past as opposed to the present time, as signaled by kinuna ‘now’ (cf. Bertinetto and Cambi 2006).

So far, we have discussed those functions of -ške/a- that involve a repetition of the event denoted by the base verb. We now turn to a different function, in which the plurality of the participants involved to the event, rather than the iteration of the event itself, is focused upon. Examples of -ške/a- forms occurring in contexts that involve a plurality of participants are given in (13).

As the examples in (13) show, the participants whose plurality is highlighted by -ške/a- can occur in different syntactic positions. Mattiola (2017: 124) remarks that “the argument that is pluralized […] tends to be the transitive object and the intransitive subject”. This tendency is fully confirmed by the Hittite data, where most of the instances of plurality of participants concern the direct object, as in (13a). Plurality of intransitive Patient subjects is also well attested, as shown in (13c). By contrast, in our corpus we found only one instance of intransitive subject with an Agent semantic role, as shown in (13b). In line with Mattiola’s prediction, there are no cases of pluralized participants in transitive subject position.

In absence of native speakers’ intuition, it is admittedly not always easy to single out the motivations for the use of -ške/a- as a marker of plurality of participants. As a matter of fact, -ške/a- forms only occur with a subset of verbs that take plural intransitive subjects/transitive objects, and the latter are also perfectly compatible with simplex verbal forms. It is thus clear that plural number of the S and O participants cannot be the only triggering factor for the occurrence of -ške/a-. As a tentative explanation, we would like to suggest that the difference between simple and -ške/a- forms is one of construal of the structural schematization type, akin to the mass vs. count distinction in the nominal domain (Croft and Cruse 2004: 45, 63−64). Specifically, it might be the case that base verbs construe the relevant participants as a homogeneous group/set to which the event denoted by the verb applies uniformly, whereas -ške/a- forms construe the participants as constituting a set of distinguishable individual entities, to which the verbal event may not apply uniformly (in place and/or time). Indeed, this is precisely the function that PMs perform in several North-American languages (e.g., Yurok [Algic], Nisga’a [Tsimshian], Central Pomo [Pomoan], among others; see Mithun 1988). Consider the following example from Central Pomo:

In (14a), we have a single event (manáač’ ‘pay.sem’) involving a single O participant (múuṱu ‘him’). With a plural O (múuṱuyal ‘them’), two options are available. On the one hand, the verb can occur in the singular, as in (14b), thereby construing the situation as a single event (‘The crew work received a single check jointly’). Alternatively, the verb can occur with the PM -ṱa-, which triggers a pluractional reading of the verb, whereby the situation is conceived as composed of several individual actions involving each O participant individually (‘Each worker was paid individually.’).

We would like to suggest that Hittite -ške/a- may in fact behave similarly to the Central Pomo PM -t̯a- as described in Mithun (1988) . Let us consider the behavior of the verb āk- ‘die’ in (13c) and (15).

The simple form akir in (15) refers to a single event, in which all participants are construed as dying in the same place and time as a whole group, as a consequence of the defeat of the city of Zalpa. Conversely, in (13c) the form akkiškittari ‘people die (lit. it dies)’ is used to indicate the devastating effect of the plague on the land of Hatti, where people die one after the other in different locations. Clearly, the fact that the participants do not uniformly undergo the event of dying, but rather do so at different times, easily paves the way for a frequentative interpretation of the suffix in contexts such as (13a), whereby repetition of the event instead of plurality of the participants is carried into focus. Similarly, in (13a) and (13b), what is highlighted is the individuality of each participant, which also leads to a sequential interpretation of the verbal event: in (13a) the queen takes of her sons one by one and puts them in the basket one after the other, and similarly, in (13b), the men are involved in a competition, so that each one is shooting to the target on his own, and they are not acting as a group. Based on these reasonings, we suggest that the use of -ške/a- might be sensitive to a higher degree of individualization of the event’s participants.

A function that is particularly difficult to individuate is the intensive one. Already Dressler (1968: 188) noted that intensification is difficult to assess in a corpus language. A possible example for the intensive function is the verb ēšši(š)ker in (16). This form is based on the verb iya- ‘make do’ and receives double marking with both the suffixes -šša- and -ške/a- (cf. Daues 2012; Pisaniello 2016: 345−346). Based on the context, it can hardly indicate but a single perfective and telic event affecting a single participant. The contribution of the suffix thus lies elsewhere. Whereas the meaning of the simple verb is basically ‘do, make’, the suffixed verb in (16) seemingly display an additional intensive nuance, which highlights the thoroughness with which the event is carried out (see Dardano 1997: 44−45).

Other two PM functions of -ške/a- that are documented are the spatial distributive and the generic imperfective functions (even though these functions are attested in our corpus, we prefer to illustrate them with passages taken from other sources, in which the proposed interpretation is more straightforward).^[14]

Spatial distributive is defined as the use of the suffix in contexts in which plurality affects not the participants or the temporality of the event but rather its place, so that what is highlighted is the displacement of the event across different locations. An example of this use is given in (17), in which the form papparaškizzi ‘he sprinkles’ possibly indicates that the event of sparkling water is distributed across different locations, as also suggested by the occurrence of the distributive spatial adverb duwan duwann = a ‘here and there (lit. here and here)’.

With the term generic imperfectivity, Mattiola (2017, 2019 refers to statements that indicate general truths or properties of an entity, and that as such always hold true. A case in point is the use of the verb ḫuišnuškizzi ‘keeps alive’ in (18). In this case, the suffixed verbal form indicates the property of grains, which by encyclopedic knowledge are known to serve the purpose of feeding, i.e., keeping alive, human beings and animals.

Finally, Hittite -ške/a- attests to a function that does not feature in the conceptual space of pluractional constructions discussed in Section 3, i.e., the inceptive function. This usage of -ške/a- is shown in (19), in which the predicate šeš- ‘sleep’, which normally indicates a state, receives an inceptive interpretation ‘start to, go to sleep’ when suffixed with -ške/a-. Notably, the inceptive reading of finite -ške/a- forms is only available to atelic predicates (cf. Hoffner and Melchert 2002: 384−385, cf. also iya- ‘march’ vs. iy-anna- ‘begin to move’; see Pisaniello 2016: 132−166 for discussion).

The inceptive function is admittedly scarcely documented with finite forms of the verb, but it is robustly attested in the supine construction. This construction, which is typical of NH historical and mythological texts (cf. Daues 2007), is a type of auxiliary verb construction and is made up by the supine in -ške/a- combined with finite forms of the verbs dai- ‘put’ and tiye/a- ‘step’, which function as auxiliaries (see discussion in Luraghi 1998 and García Ramón 2007). The resulting periphrastic construction roughly means ‘begin to X’ (Hoffner and Melchert 2008: 322, 338), i.e., it profiles the ingressive phase of an event (cf. Cotticelli Kurras 2015). Notably, whereas the inceptive interpretation of finite forms with -ške/a- is limited to atelic predicates, this is not the case of the supine periphrastic construction, which also may involve telic predicates (see García Ramón 2007: 287–288). Consider in this respect the use of telic pai- ‘give’ in (20).

4.2 The semantic map of Hittite -ške/a-

In the previous section, we have shown that -ške/a- can be rightfully described as a PM, as it covers a range of functions typical of PMs. Moreover, not only does -ške/a- displays functions which are typically associated to PMs, but its polyfunctionality also complies with the conceptual space of pluractional constructions set up by Mattiola (2017), as the functions associated with the suffix occupy a contiguous region on the conceptual space (Figure 2).

Figure 2:

The semantic map of Hittite -ške/a-.

Admittedly, the map in Figure 2 presents two problems. In the first place, the map apparently features a gap, since the event internal plurality node is not covered. In this respect, the map would violate Croft’s (2001) semantic map connectivity hypothesis. This is however not a major issue. On the one hand, the absence of this function might simply be accidental, and reflect the limited nature of the corpus, on the other hand, evidence for postulating an intensive node is rather scanty at best, as discussed for example (16).

A more serious problem is placement of the inceptive function. As we have discussed for examples (19) and (20), when used with atelic predicates and in the supine construction, the suffix -ške/a- may indicate entry in a state/beginning of an action. This function is however absent from Mattiola’s conceptual space, so that it is not immediately clear where one should place it in the semantic map of Hittite -ške/a-. Note that this is partly a by-product of the methodology behind Mattiola’s map, since in building up the conceptual space in Figure 1, the author has only taken into account functions that are robustly associated with PMs and has excluded other functions that individual language-specific markers may occasionally encode (Mattiola 2017: 128). Therefore, the lack of the inceptive function in the conceptual space simply means that this function is only rarely performed by PMs, but the map in principle does not rule out the possibility that individual PMs may additionally perform this function (Mattiola 2019: 40−42). Indeed, a closer look at Mattiola’s sample shows that the inceptive function is sporadically attested for other PMs (cf. Zúñiga and Díaz-Fernández 2014 on the functions of reduplication in Mapudungun – Araucanian).

In absence of relevant cross-linguistic data, we cannot securely place the inceptive function within the conceptual space of PMs following the standard methodology for first-generation semantic maps. However, we would like to propose a possible semantic connection of this function with iterativity. As a matter of fact, iteration implies boundedness, in that in order to be repeated events must have a beginning and an endpoint. When applied to events that are usually construed as atelic, a marker that conveys iterativity might simply impose a boundary on the beginning of the atelic situation, hence yielding an inceptive reading.

The semantic map in Figure 2 is a useful tool to neatly visualize the pattern of polyfunctionality of Hittite -ške/a-. This geometrical representation can be further enriched with quantitative data concerning the frequency of individual nodes, in order to show what meanings on the maps are more robustly associated with the suffix in its actual usage and what are only marginally so. Data concerning the frequency of the individual functions of -ške/a- is given in Table 4.

As the data in Table 4 show, the frequentative function is the most frequently attested both in terms of types and tokens, whereas the continuative, iterative, plurality of participants, and habitual functions all display a lower and comparable distribution. Still, one should be aware that the frequencies in Table 3 should be handled with due care, since they partly reflect a bias in the selection of the corpus, and as such they might not accurately represent the actual distribution of the suffix in the language. For example, the high frequency of the frequentative function partly reflects the high incidence of the verbal form ḫatrae-ške/a-, which is often found as a greeting formula ‘keep writing me letters/greetings’ in MH letters. Note also that the high incidence of the inceptive function is due to the frequency of the supine constructions in NH texts (out of 31 tokens with inceptive meaning, 28 are non-finite supine forms).

In discussing the different functions of the suffix, it should be kept in mind that suffixed forms of the same verb can have different interpretations. Consider as an example the verb eku- ‘drink’, whose -ške/a- forms can either have iterative (21a), habitual (21b), or plurality of participants (21c) meaning. Disambiguation between the functions is entirely dependent on the context, and can be facilitated by the occurrence of temporal adverbs, as in the case of 2-iš ‘twice’ supporting the iterative reading of (21a).

The question is thus whether one can detect constraints on the distribution of the individual functions. A close investigation of our corpus data shows that the only factor that seemingly plays a role in constraining the interpretation of -ške/a- verbs is the verbs’ lexical aspect.^[15]

Indeed, our corpus data largely confirms the widespread observation (cf. Bechtel 1936) that the suffix is unavailable to stative verbs.^[16] Interestingly, this is a common behavior of PMs in the languages of the world (cf. Mattiola 2019: 144). The distribution with the other actional classes shows an interesting picture. As the data in Table 5 neatly show, most functions predominantly occur with telic predicates, with a preference for achievements. However, the frequentative and the continuative functions are also available for atelic activity predicates, and this is also the only class of predicates that license the inceptive reading with finite verbal forms. This distribution complies with Mattiola’s (2017: 135) observation that functions on the right side of the semantic map are unconstrained with respect to the lexical aspect of the verb they apply to, whereas “the functions on the left side express a semantics that sometimes can be incompatible with some of type of verbs.”

4.3 From PIE ∗-sḱe/o- to Hittite -ške/a-: a diachronic scenario

In this section, we discuss aspects of the diachrony of Hittite -ške/a- and tackle the issues of how the different functions of -ške/a- diachronically relate to one another. Specifically, we try and pinpoint an original function, if any, and explore possible pathways whereby the other functions arose out of it. In doing so, we combine comparative evidence from cognate suffixes in other IE languages with typological considerations on the processes that give rise to PMs in the world’s languages.

As is well known, Hittite -ške/a- continues the PIE thematic suffix ∗-sḱe/o- (the latter perhaps from earlier ∗-s-ḱe/o-; cf. Jasanoff 2003: 134–136, Oettinger 2017, Willi 2018: 480), with cognates in several ancient IE languages, e.g., Skt. -ccha-, Av. -sa-, Gr. -ske/o-, Lat. -sce/o-, OIr. -c-, OHG -sc- (Zerdin 2000: 21–37, Jasanoff 2003: 133–134, Kloekhorst 2008 s. -ške/a-, Lundquist and Yates 2018: 2164; see also Adams 2014 for a systematic comparison of the suffixes in individual languages and discussion of possible Baltic, Armenian, and Albanian comparanda, which we do not discuss further here). Let us briefly review the main features of outcomes of PIE ∗-sḱe/o- in some ancient branches of the IE family.

In Anatolian, outcomes of PIE ∗-sḱe/o- are also (scarcely) attested in languages other than Hittite, viz. Luwian and Lycian. The Luwian suffixes -zza- (CLuw.) and -za- (HLuw.) express continuative, e.g., ta-za-tu ‘let last/endure’, and inceptive functions, e.g., kappilazzata ‘became hostile’ (Melchert 2003: 205; plurality of participants, iterative, and continuative are also associated with the Luwian suffix -š(š)a-, cognate with Hittite -šš(a)-, cf. Pisaniello 2016). The Lycian s-verbs also originally feature the ∗-sḱe/o- suffix, but by the time of their earliest attestation, they fail to show any semantic difference with the corresponding base verbs (cf. Serangeli 2018).

Greek and Indo-Iranian provide crucial evidence for the reconstruction of PIE ∗-sḱe/o-. In these branches, outcomes of ∗-sḱe/o- are predominantly connected with imperfectivity, as they are used to form (imperfective) present stems as opposed to (perfective) aorist stems (on Greek see e.g., Rix 1992: 213–214; Willi 2018: 479–488, and Zerdin 2000; on Sanskrit see Burrow 1973: 329–330; on the aspectual systems of Ancient Greek and Vedic see Napoli 2006 and Dahl 2010 respectively).^[17]

Tocharian proves evidence for yet another function of the suffix. In Tocharian, the suffix -äsk- (TochB.) is employed in the formation of causatives and intensives. The development of the causative meaning likely reflects a Tocharian innovation (Adams 2014), and thus should not be projected back into the semantics of PIE ∗-sḱe/o-. Note that, whatever the details of the historical relationship between the two, a connection between pluractionality and causativity is not entirely surprising. Within PIE, the suffix ∗-éye/o- is commonly reconstructed as causative/iterative (cf. e.g., LIV² : 22–23). Compare the Latin causative monēre ‘warn’ < ∗mon-éye/o- ‘call to mind’ < ∗men- ‘think’ with iterative tondēre ‘shave’ < ∗tond-éye/o- ‘cut repeatedly’ < ∗tend- ‘cut’ (cf. Kölligan 2007). In addition, even in PIE verbal reduplication is commonly associated with both an iterative and a causative meaning (cf. Dempsey 2015 for discussion of the Hittite data). Outside IE, a typological parallel is offered by Khwe (Khoe-Kwadi, Khoe), where verbal reduplication encodes both causativity and pluractionality (cf. Kilian-Hatz 2008: 147, 161).

Latin offers a rather complex picture when it comes to outcomes of ∗-sḱe/o-, as it features several formations of different origin (Keller 1992; Weiss 2009: 407). Firstly, Latin shows traces of inherited present stem formations that have parallels in other IE languages, e.g., (g)nō-sc-ō ‘know’ < ∗ ǵneh ₃ -sḱe/o-, cf. Gr. (gi)gnṓ-sc-ō ‘know’. Traces of -sc- forms with habitual function are rather scanty. An often-cited piece of evidence is provided by the forms escit/escunt from esse ‘be’ with future habitual meaning (cf. Sihler 1995: 550). Finally, Latin displays a wealth of innovative -sc- formations with inceptive-intransitive meaning, of the type caleō ‘I am hot’ > calē-sc-ō ‘I become hot’ (cf. Haverling 2000). Overall, as Haverling (2000) discusses, the core meaning of the Latin suffix (with unprefixed verbs) is connected to the indication of durative and dynamic (atelic) events, including inceptives when based on stative verbs (cf. Haverling 2000).

To sum up, IE languages offer clear comparative evidence for the formal reconstruction of present stem verbs with zero grade roots and accent on the ∗-sḱé/ó- suffix: e.g., ∗g ^w em- ‘go’ > ∗g ^w m̥-sḱé/ó > Ved. gácchati, Gr. báske (cf. LIV² s.v.). Unfortunately, the functional reconstruction is far less immediate, and, similarly to the difficulties in assessing the function of Hittite -ške/a-, scholars have offered different reconstructions of what they consider to be the original or core meaning of PIE ∗-sḱe/o- (see Willi 2018: fn. 181 with further references). Clearly, the reconstruction of the original semantics of the PIE suffix ∗-sḱe/o- lies beyond the much more limited scope of this paper, so we do not pursue further here. In the complex issue of the reconstruction of the suffix, what is striking is that Anatolian shows a remarkably wider range of usages of -ške/a- as compared to other IE languages, in which the functions seem to be associated more with the right end of the semantic map, i.e., with imperfectivity and other more abstract functions. The question is thus whether Anatolian represents the original state of affairs, and the other IE languages innovated by narrowing down the functional range of the suffix. This scenario is for instance advocated by Adams, who observes that “there seems no doubt, however, that Hittite preserves the original meaning and that meanings other than imperfectivity [i.e., pluractionality] are innovations.” (Adams 2014: 24−25).

Useful insights concerning the prehistory of Hittite -ške/a- can be gained from the general trends in the development of PMs. Unfortunately, to date there is no comprehensive diachronic typology on the origin of PMs in the world’s languages, but recent studies have shown that at least for some areas of the conceptual space of pluractional constructions, directional diachronic links can be established between some of the nodes. For example, there is evidence that plurality of participants is the likely source of the reciprocal function rather than the other way around (cf. Lichtenberk 2000; reciprocals may eventually evolve into antipassives as well, see Sansò 2017). Cross-linguistic research on the development of aspectual markers has also brought to light the existence of two distinct clines of directional semantic change that lead from iterativity to imperfectivity ( Bybee et al. 1994: 172; see also Kouteva et al. 2019), as shown in (22):

The developments shown in (22) display some of the typical features of grammaticalization processes. In the first place, markers that undergo the changes in (22) develop more grammatical (abstract) meanings (cf. Lehmann 2015 [1995]). Whereas iterativity affects the lexical semantics of the event, in that it conveys the repetition of the same event vs. a single instantiation of such event, imperfectivity only entails a certain viewpoint on the event, but does not affect the lexical representation of the event itself. In other words, the more a function is placed towards the grammatical pole the lesser impact it has on the verbs’ lexical meaning. In the second place, in the development from iterativity to imperfectivity one also observes class-host expansion, i.e., the expansion of an item to contexts previously incompatible with its source meaning due to the process of semantic bleaching (Himmelmann 2004). In the case of PMs, this means the extension to previously unavailable verb classes. As we have already remarked, whereas iterativity is lexically compatible with telic verbs only, be it achievements or accomplishments, imperfective markers can also apply to atelic verbs. Finally, the development of imperfectivity, which relates to the speaker’s viewpoint of events in discourse, also entails a higher degree of subjectivity, which is another hallmark of grammaticalization processes (Traugott 2010).

What remains unclear is the directionality of the diachronic link between iterativity and plurality of participants. As Mattiola (2017: 131) argues, a link exists between plurality of participants and iterativity in that the repetition of a situation can also be easily conceived as being distributed among different participants and vice versa (or places, in the case of spatial distributivity). Further research is needed to establish whether a cross-linguistic directional link can be established between these two functions. Preliminary findings suggest the existence of a two-way relationship between iterativity and plurality of participants. For example, Frajzyngier (1997) identifies different clines of grammaticalization that lead from demonstratives to nominal and verbal number affixes in Chadic languages. Among these paths, one also finds that demonstratives may develop first in markers of nominal plurality (both nominal number and participant plurality) and, then, in markers of event plurality, i.e., iteratives (cf. (23)).

In the shift from plurality of participants to iterativity, one possible bridging context is provided by the use of singular nouns that refer to groups instead of plural NPs. Consider the Hittite example in (24):

The passage in (24) refers to the fact that due to a widespread plague, the people of the Hittite kingdom keep dying. Clearly, each single event of dying being unique, here the lexical semantics of the verb āk- ‘die’ strongly favors a plurality of participants reading, as we already discussed for example (13c). However, in (24) the singular noun phrase KUR ^URU Ḫatti ‘the land of Hatti’ occurs as subject, and is used metonymically to stand for its people (on this type of metonymy see e.g., Croft 1993: 353). As a consequence, the individual entity ‘land of Hatti’ can be conceived as undergoing a single continuous event of dying as a whole, thus licensing an iterative reading, and eventually a continuative reading via structural schematization (Croft and Cruse 2004: 63−64). In other words, in contexts such as (24), the combination of metonymy and structural schematization can easily lead from an event in which the plurality of participants is expressed to one in which the continuation of the event itself is highlighted instead.

The reverse process has been proposed by Mithun (1988), who discusses some cases in which markers that originally indicate iterativity (among other things) start being used as nominal number markers in some Native North America languages. For example, this is the case for the Cayuga examples in (25) where the PM (cf. (25a)) develops into a marker of plurality of entities and starts being applied to nouns (cf. (25b)).

The diachronic information discussed so far can be integrated with the conceptual space of pluractional constructions discussed in Section 3. The result is a dynamic semantic map, in which the directionality of the historical links among individual nodes is also taken into account (cf. Luraghi 2014; Narrog and van der Auwera 2011) (see Figure 3).

Figure 3:

A dynamic conceptual space of pluractional constructions.

The combination of comparative evidence, along with the common trends in the development of PMs and in the grammaticalization of imperfective markers, allows us to propose a tentative diachronic scenario for the emergence of Hittite -ške/a-.

To begin with, it seems reasonable to reconstruct a stage in which the PIE suffix ∗-sḱe/o- functioned as marker of iterative and/or plurality of participants. As such, it could be rightfully considered as a PM, as it covered only the core PM functions. Out of the core pluractional functions, the suffix started extending its functional range towards the right end of the conceptual space, i.e., towards the expression of aspectual values, in compliance with the cross-linguistic trends of semantic change discussed in (22).^[18] In Anatolian (Hittite and possibly Luwian), the original core functions and the new ones coexisted, with a ‘layering’ of functions typical grammaticalization processes (Hopper 1991). In other words, as Josephson puts it, the suffix retained its original functions, and “it is however not obvious that there was a full-scale grammaticalization of the -ške/a- form” (Josephson 2008: 138). By contrast, in the other IE languages, as the suffix became increasingly associated with the left-side functions on the map, the original functions of the suffix were partly lost, and ∗-sḱe/o- evolved into a general maker of present stems (imperfectivity) and also underwent language-specific developments and specializations (e.g., causative in Tocharian). Again, this scenario must remain speculative, and needs to be tested against a serious and more systematic reassessment of the formal and functional reconstruction of PIE ∗-sḱe/o- and the aspectual system of PIE, which exceeds the scope of this paper.

A final word is in order concerning the relative chronology of Hittite -ške/a- as compared to other means of encoding pluractionality. As a matter of fact, Hittite attests to at least another potential PM, i.e., verbal reduplication. In Hittite, verbal reduplication seems to cover a range of functions similar to -ške/a- (see Dempsey 2015 for a thorough discussion): durative, habitual, iterative, repetitive, distributive, intensive/inceptive, and causative. Verbal reduplication is however unproductive in historical times, and many reduplicated verbs are secondarily re-characterized through the addition of -ške/a- (Dempsey 2015), e.g., ku-kkure-ške/a- ‘cut, mutilate’ (distributive according to Hoffner and Melchert 2002: 384). It seems thus that -ške/a- replaced verbal reduplication as the standard PM in Hittite. Again, typological parallels to the Hittite situation exist. Reduplication operates as a PM in some of the world’s languages (such as Pluractional derivation in Beja [Afro-asiatic, Cushitic], cf. Vanhove 2017), and the recessive status of reduplication as a PM with new PM is attested for instance in Maa (Nilotic, Eastern Nilotic; cf. Payne 2013 and Mattiola 2019: 125−142 on the andative -áa in Maa).