Secondary predicates and motion events: the rise of complex predicates in Vedic Sanskrit

1 Introduction^[1]

The following paper focuses on the origin of the Late Vedic and Classical Sanskrit periphrasis consisting of a participle (or a gerund) and a semantically bleached finite verb of motion or position (e.g., ay, car, sthā, ās). While the formal aspects of this construction have long been described in the literature (Delbrück 1888; Grieco 2023; Lowe 2015; Macdonell 1916; Speyer 1886, 1896; Whitney 1896), its semantic function appears to be less clear-cut – the various analyses center around the action as an ongoing event (‘to be continually etc. doing x’), which contains not just continuous and habitual, but also iterative and continuative readings.^[2] The latter can be illustrated by the following example with ās ‘sit’, taken from Lowe (2015: 116):

The action encoded here shows all possible readings, as it is ongoing and recurring. Furthermore, the finite verb has to be interpreted as an auxiliary because its lexical meaning is semantically incompatible with that of the participle (killing while sitting down). Similarly clear evidence for periphrasis can be found as early as the time of Vedic Prose, cp. the following examples with car and ay ‘go’:

While in (2), the actions of striking and moving might theoretically be concomitant, this is not possible for (3): here, the meaning of the participle, ‘stand firm’, argues against the interpretation of the action as a motion event, providing clear evidence for the desemanticization of yanti.

It is generally assumed that, while the origin of this construction goes back to Early Vedic, a productive and fully grammaticalized pattern has evolved only in Post-Rigvedic times (Lowe 2015: 116 with references). All examples attested in the Rigveda are ambiguous: they may either be interpreted as expressing a motion event, i.e., with the full lexical meaning of the finite verb retained, or as expressing a continuous event, i.e. with the participle used as main predicate and the finite verb as auxiliary, cp.

In this verse, both interpretations appear equally possible: the river goddess Sarasvatī either moves while roaring or she roars continually. As opposed to (1) and (3), the lexical meaning of the finite verb may here still be present. This and other Rigvedic examples are discussed in Lowe (2015: 116–118), who states that “In all these examples the sense of movement is not excluded, though it is not absolutely necessary in any” (p. 117). Note also that the word order in the RV example differs from that in (1)–(3): the participle follows the finite verb instead of preceding it, as would be the norm in the younger stages of the language, where the sentence-final position is usually occupied by the verb. In contrast, the Rigvedic participle may appear in pre- or postverbal position.^[3]

Although this construction type has already been described in early handbooks, a thorough and comprehensive analysis has not been undertaken until quite recently: in her paper, Grieco (2023) links the origins of this periphrastic construction with the fate of the Vedic intensive category. She observes on pp. 235–236 that of the 56 possible Rigvedic precursors of the periphrasis with the roots ay and car 46,4 % contain an iterative participle, belonging to the intensive stem (s. ex. [4] above) or to reduplicated class III present stems. This number is quite remarkable considering that these two formations make up only 4,6 % of all participles attested in the Rigveda. However, the Vedic intensive stem, despite its name, frequently encodes iterativity (Schaefer 1994: 75–94), cp. the following example, where context and world knowledge suggest an iterative or continuative reading:

This reading may further be enforced by the combination with temporal adverbs or accusatives of extension, as in

Formations like RV 10,95,10a dávidyot int.inj.3sg ‘keeps flashing’ (like lightning) allow for both iterative and intensive interpretation, thus enabling the transition from the former to the latter (Schaefer 1994: 93). Starting from these observations, Grieco (2023) convincingly argues that finite motion verbs combined with intensive participles have played a crucial role as “bridging contexts” (Heine 2002) in the onset stage of the grammaticalization process.^[5] This new analytical construction then competed functionally with the inherited synthetic intensive, which starts to decline after Rigvedic times and gradually disappears in later stages, having been replaced by the new periphrasis.

In this paper, I will again focus on Early Vedic and on possible precursors of the periphrasis. While the role of the intensive stem has been compellingly demonstrated by Grieco’s work, I will now take another look at the onset of the whole process in order to tie up some loose ends. These concern two features of the relevant Rigvedic data: (a) the presence of participles from stems other than intensives (or class III verbs), and (b) the general absence of directional complements.

Regarding (a), it is evident that not all of the participles used in possible precursors of the periphrasis, i.e. in combination with motion verbs, belong to reduplicated stems, cp., e.g., with a class V verb:

Since the later periphrastic construction uses various kinds of participial stems apart from the intensive (see also exx. [1] – [3] above), it seems reasonable to suppose that sentences like (7) also had a role to play in the grammaticalization process. This also fits to the observation that not all periphrastic examples have an iterative meaning, cp. (3) above. What follows is that there have to be more factors responsible for favouring an iterative, habitual etc. reading of the Rigvedic examples than the morphology of the participial stem. Closely connected to this is the following question: If, in the Rigveda, the participle was not yet part of a periphrastic construction, what exactly was its syntactic function at that stage when combined with a verb of motion? I will address this question in Section 2.

Regarding (b), since the events encoded in the periphrasis are no longer motion events, an absence of directional complements is of course to be expected (Section 3). In contrast, in the Rigveda, constructions of motion verbs without goal, path or source expression are comparatively rare and frequently refer to non-directional movement (Sections 4 and 5). I will argue that this non-directionality facillitates the interpretational switch from motion to non-motion event, thus enabling the transition from motion verb to auxiliary.

My focus in this paper is very narrow: I will restrict myself to the Rigveda and disregard the later history of the construction.^[6] Furthermore, I will only consider contexts with motion verbs and leave position verbs aside.^[7] The overall aim of this paper is to demonstrate that the Rigvedic data point towards a multi-causal origin for this periphrasis. Additionally, it will address the question of what factor exactly might cause a motion verb to change into an auxiliary. After these preliminary words, we will now turn to the syntactic function the participle has in the Rigveda, when combined with a verb of motion.

2 Secondary predicates

The participle róruvat ‘roaring’ in (4) above functions syntactically as an adjunct, more precisely, as an additional predication beside the main predicate, i.e. the finite verb. In the literature, these adjuncts have been discussed under various labels: converbs, copredicat(iv)es, predicative attributes, predicative adjuncts, Subjekts-/Objektsprädikativa, etc. Here, the term “secondary predicate” is preferred following, e.g., Himmelmann and Schultze-Berndt (2005). A secondary predicate is characterized by the following three features (Himmelmann and Schultze-Berndt 2005: 1–15):

1. It is an additional predication beside the main predicate.

2. It is participant-oriented, since it provides information about the physical or mental state of a referent.

3. It overlaps temporally with the time frame set by the main predicate.

For illustration cp. the following English examples:

(8a.) is a secondary predicate: it describes the (in this case) mental state of the referent while leaving the party (participant-oriented). (8b.) is an event-oriented adverb: it focuses on the manner of leaving without necessarily referring to the mental state of the referent. (8c.) is semantically incompatible, since tall does not refer to a changeable state. While in (8a.) and (8b.), it is to be assumed that some kind of trigger at the party caused the action of leaving, tall or clever express permanent states that continue far beyond (and, of course, precede) the temporal frame of the verb left.

A typical Vedic example for a secondary predicate would be (7), here repeated:

In the Rigveda, secondary predicates may be formed by various nominal – and more rarely – pronominal formations. However, by far the largest subgroup consists of participles. While this frequently attested syntactic function of participles has been described already by Delbrück (1878), detailed studies on secondary predicates in Vedic Sanskrit have only appeared during the last decades (Cantera 2005; Casaretto 2020; Keydana 2000; Lowe 2015; Widmer and Scarlata 2017, 2022).

Concerning the emergence of the Late Vedic periphrasis, it is well known that there exists a diachronic link between secondary predicates and complex predicates (Himmelmann and Schultze-Berndt 2005: 52 with references). Typologically, the combination of secondary predicate and main predicate shows a tendency to form fixed collocations. In such contexts, the main predicate may then gradually change into an auxiliary, while the secondary predicate becomes the main predicate. When the finite verb is semantically general, but still contains some semantic content, it may become difficult to distinguish between both constructions. As an example, the authors give the usage of Engl. lie in The man lay dead in his home for three days, which comes close to a copular construction. Transferring this onto the Vedic periphrasis, one would expect contexts in which the motion event and the expression of continuity etc. appear balanced so that either interpretation becomes possible. These ambiguous contexts are taken as the starting point for the change from a secondary predicate construction to a periphrastic construction.

Following these general observations about the relationship between secondary and complex predicates, a new survey of the Rigvedic verb-participle-combinations is called for in order to determine whether the ambiguity necessary for the reanalysis of the construction works for all combinations of motion verbs and participial secondary predicates or whether there are additional factors to consider. For that, it is now necessary to look at the clausal structure of the constituents in question, especially at their combination with arguments and adjuncts.

3 The absence of directional complements

As has been stated above, the periphrastic construction has no directional complements like source or goal arguments. They are of course not necessary, because the action expressed is no longer a motion event. Looking at ambiguous Rigvedic examples that are discussed as possible bridging contexts, it is conspicuous that they also lack goal or source expressions, cp. (4) and (7) cited above and repeated here:

Both sentences seem to express a motion event without explicit goal or source. In other words: the lack of complement turns these events into non-directional motion events, which are by their nature atelic events or activities (Vendler 1967). The expressed action continues for an unspecified amount of time without starting or end point.

Before looking at this construction type in more detail, it seems useful to first go back a step and consider how motion events are usually encoded in the Rigveda.

4 Encoding motion events in the Rigveda

In a prototypical motion event, an object (figure) changes its position with respect to another object (ground) along a trajectory (Jackendoff 1983, Talmy 1985, 2000). In the Rigveda, the concepts of ground and trajectory are expressed by case markings, sometimes combined with adverbials, while the change of position is expressed by a motion verb. In other words: motion events are typically encoded by the combination of a motion verb with a complement denoting the goal or source of the movement. Morphologically, this complement is a noun in a local case form, a local adverb, or a local particle.^[8] Like in most other Indo-European languages – with the exception of the Romance languages –, this kind of spatial marking outside the verb conforms to the satellite-framed type (Talmy 1985, 1991, 2000).^[9] The following subsections contain some examples for expressing these spatial concepts in the Rigveda, starting with goal (4.1) and source (4.2) and covering also the concepts of path and location (4.3).

5 Motion verbs without complement

In the 10 books of the Rigveda, I have counted 118 occurrences where the motion verb is not accompanied by a complement denoting goal, source, path or location. I have excluded from this list verbs of inherently directed motion like nakṣ ‘reach’ and veś ‘enter’, since these are by their lexical meaning already specified regarding direction, even without overt complement (Levin 1993, esp. p. 264). I have also naturally restricted myself to finite verbs and considered participles only in two cases where they directly support the usage of the respective finite verb (ex. [28] and [34]).

While 118 occurrences may seem like a fairly large number, one should keep in mind that, due to the function of this text, which is to invoke the gods to attend the sacrifice, motion verbs generally occur in abundance. Thus, these numerical relations may not reflect the linguistic reality outside the Rigveda. Still, the construction without complement is clearly the exception and not the rule, even in relative numbers. The data can be divided into two groups depending on whether the lack of complement is caused by argument omission or because the movement is non-directional:

1. possible cases of omission: 25 (Section 5.1)

2. possible cases of non-directional motion events: 93 (Section 5.2)

Although these numbers seem to suggest that omission occurrs only rarely, this may not be quite accurate. Rather, as discussed below, there is an ambiguous grey zone between the two groups, where no certain decision is possible.

6 Summary

In this paper, I have argued that two factors have facillitated the emergence of the periphrasis with motion verb and participle in Late Vedic: The first factor is the usage of various participial stems as secondary predicates encoding an action which overlaps temporally with the time frame set by the main predicate, thus favouring a purely continuous interpretation. However, for the reanalysis of motion event as non-motion event, which is a prerequisite for the grammaticalization process, there is an additional important factor to consider: the usage of the finite verb-participle combination for the expression of non-directional activities. These activities are either encoded by motion verbs without overt complements or by those with location complements. Only in these constructions is an interpretation as non-motion event, followed by desemanticization of the finite verb, at all possible. These ambiguous contexts can be seen as possible bridging contexts for the transition of motion verb accompanied by a participle into a complex predicate in post-Rigvedic times.

While the intensive has certainly played an important role in the development of the periphrasis by strengthening the iterative or continuative reading (Grieco 2023), I think that by considering the whole clausal structure (i.e., the syntactic function of the participle and the presence/absence of complements) additional factors have emerged that have been crucial for the development of the periphrastic construction. Apart from shedding more light on the complex origin of this periphrasis by asking the question of what makes a motion verb turn into an auxiliary, we are also now able to link the post-Rigvedic grammaticalization process with typological findings from other languages concerning the relationship between secondary and complex predicates.