Pluractional Motion Verbs in Turkish

Abstract

In this paper, we examine a small set of motion verbs in Turkish bearing the so-called ‘reciprocal’ suffix -(I)ş: kaç-ış- ‘flee in all directions’ (cf. kaç- ‘flee’), koş-uş- ‘run helter-skelter’ (cf. koş- ‘run’), uç-uş- ‘fly helter-skelter’ (cf. uç- ‘fly’). It has previously been claimed that these are collective or sociative verbs entailing a low elaboration of events and/or plural participants. We show that these -(I)ş-marked verbs, in fact, require a higher degree of individuation of events than do their unmarked counterparts. Furthermore, we show that a stipulation directly associating the suffix with a requirement for a plural subject is both unnecessary and inadequate. Instead, we propose that the pluractional under investigation manipulates the denoted events’ spatial and temporal properties in such a way that the predicate can only be felicitously used if it combines with a plural external argument.

1. Introduction

A suffix of the form -(I)ş marks hundreds of verbs in Turkish, among them dozens of verbal reciprocals (e.g., bak- ‘look’ > bak-ış- ‘look at each other’), yet it has received comparatively little attention in the generative literature. This may be because -(I)ş is a poor fit for generative models of argument structure alternations. In addition to reciprocals, -(I)ş derives a large number of inchoative verbs (e.g., bur- ‘twist’ > bur-uş- ‘crumple’), yet unlike detransitivizing morphology in other languages, it does not derive verbal reflexives or passives, which bear a different suffix in Turkish, -(I)n or -Il (Akkuş 2021; Key 2022a,b, 2024; Legate et al. 2020). The syncretism, including the passive, inchoative, reflexive, and reciprocal, is commonly referred to as u-syncretism (short for “unaccusative syncretism”—Embick 2004). This syncretism has long informed generative models of verb formation and argument structure (Alexiadou et al. 2015; Chierchia 2004; Embick 1997, 2004; Grimshaw 1982; Koontz-Garboden 2009; Reinhart 2003; Reinhart and Siloni 2005; Schäfer 2008, 2017). The identity of the reflexive and inchoative marker figures especially prominently in this line of work. In light of this, the absence of reflexives among -(I)ş is mysterious. It is obviously incompatible with approaches such as Chierchia (2004) and Koontz-Garboden (2009), which hold that marked inchoatives (anticausatives) are in fact verbal reflexives. Neither does the distribution of -(I)ş submit to Embick’s (2004) proposal that the morphology common to passives, reflexives, and anticausatives marks nonactive Voice (Kratzer 1996), which does not project a specifier and therefore precludes the merge of an external argument DP. The nonactive Voice head may semantically entail an agent, which is disjoint from the internal argument in the case of the passive and coreferential with it in the reflexive, or it may be semantically inert in the case of the anticausative. This idea is further developed in Alexiadou et al. (2015); Schäfer (2008, 2017); Spathas et al. (2015), among others. In this body of work, all differences between the u-syncretic verbal types are handled on the LF branch, which explains why PF spellout is identical. However, this does not predict and cannot explain the distribution of -(I)ş, which marks the reciprocal and anticausative but skips the reflexive and the passive. It does, however, derive collaborative (ağla- ‘cry’ > ağla-ş- ‘cry together’) and collective(-looking) motion verbs (koş- ‘run’ > koş-uş- ‘run helter-skelter’). Hence any attempt to analyze it within established approaches to valency reduction is like trying to force a square peg into a round hole. The different uses of the suffix -(I)ş are summarized in

Table 1. Pluractional syncretism in Turkish.

Reciprocal	bak-	‘look’	>	bak-ış-	‘look at each other’
	döv-	‘beat’	>	döv-üş-	‘fight each other’
Inchoative	bur-	‘twist’	>	bur-uş-	‘crumple’
	tık-	‘stuff into’	>	tık-ış-	‘get stuffed into’
Collaborative	ağla-	‘cry’	>	ağla-ş-	‘cry together’
	gül-	‘laugh’	>	gül-üş-	‘laugh together’
Collective(-looking) Motion	koş-	‘run’	>	koş-uş-	‘run helter-skelter’
	uç-	‘fly’	>	uç-uş-	‘fly helter-skelter’

.

In the most extensive treatment of -(I)ş to date, Gandon (2013) observes that reciprocal, sociative, and iterative/intensive verbs all involve a plurality of events and/or participants. She further points out that a cognate suffix in other Turkic languages is found in these and other plural contexts, such as assistive, comitative, competitive, and 3rd-person plural agreement, and proposes that the etymon of -(I)ş was an ancient marker of collective plurality.

We concur with Gandon that event plurality (in other words, pluractionality) underlies reciprocal, collaborative, and collective(-looking) motion verbs in Turkish (for a similar approach, see Atlamaz and Öztürk 2023). In this paper, we start outlining an approach that accounts for the syncretism of these categories by identifying -(I)ş as the exponent of a syntactic pluractional head. The term pluractionality refers to a large group of verbal expressions that can only be truthfully used in plural event contexts (Cusic 1981; Garrett 2001; Henderson 2012; Lasersohn 1995; Wood 2007). The overarching idea is that the syntactic pluractional head has various semantic interpretations; however, on the PF-branch there is just one lexical item, -(I)ş, that spells out the pluractional head (in the spirit of Halle and Marantz 1993). By treating -(I)ş as the exponent of pluractionality, we provide a synchronic account of all the major verbal types where they appear.

In order to start developing a comprehensive analysis of Turkish pluractional syncretism, this paper investigates one pluractional type, the collective(-looking) motion verbs. These verbs include a salient, albeit small, group of verbs, given in

Table 2. Collective(-looking) pluractional motion verbs.

Base Verb		-(I)ş-Verb
kaç-	‘escape, flee’	kaç-ış-	‘flee helter-skelter’
koş-	‘run’	koş-uş-	‘run helter-skelter’
uç-	‘fly’	uç-uş-	‘fly helter-skelter’

.

Previous work on the verbs in Table 2 considers them “collective” or “sociative" verbs (Gandon 2013; Kemmer 1993), which require the predicate to compose with two or more (agent) participants that bear identical roles and that are jointly involved in the described event, i.e., the pluractional motion verb uç-uş-1 in (1) would require that more than one bird perform a flying event. Setting the details aside (to be discussed below), such approaches would characterize pluractionality in (1) as requiring the events in the denotation of the base predicate to be plural and multiple agents to perform these events, i.e., the claim would be that uç-uş- in (1) denotes multiple flying events and each of these flying events must be performed by a different bird.

(1)	Kuş-lar uç-uş-tu.
	bird-pl fly-plrc-pst.3sg
	‘The birds flew about.’
		(Kornfilt 1997: 178 ex. (651))

However, this fails to capture the meaning of the pluractional motion verb uç-uş-: this -(I)ş-marked verbal form does not simply mean that there were many flying events such that they were performed by two or more actors. If this were the case, (1) would be felicitous in any context where it is true that The birds flew (together). This is, however, not borne out: uç-uş- is not felicitous if the birds flew in a perfect V-formation, as shown in (2a). This sharply contrasts with the felicity of (2b), containing the non-pluractional uç-. This mirrors Kornfilt’s (1997, p. 178) observation of the example in (1): she states that “[h]ere, the birds flew every which way, rather than together. However, the actions are simultaneous.” The infelicity of (2a) is mysterious if we take a “collective verb” approach.

(2)	Cranes migrate every winter... (Turnalar her kış göç ederler...)
	a.	#V düzen-in-de	uç-uş-ur-lar.
		V-formation-cm-loc	fly-plrc-aor-3pl
		Intended: ‘They fly (together) in a V-formation.’
	b.	V düzen-in-de	uç-ar-lar.
		V-formation-cm-loc	fly-aor-3pl
		‘They fly in a V-formation.’

(2a) illustrates that the pluractional does not simply require the events in the denotation of the base predicate to be plural. Rather, it defines event plurality in some other way. The main question this paper asks is how the pluractional of motion verbs defines event plurality. The proposal, in a nutshell, is that the pluractional manipulates the denoted events’ spatial and temporal properties such that the movement paths must intersect while the events are performed simultaneously.

The paper is structured as follows: Section 2 offers an overview of the syntactic diagnostics that distinguish pluractional motion verbs from other -(I)ş verbs. Section 3 turns to a detailed discussion of the pluractional motion verbs with a plural subject, paying particular attention to the contexts in which these verbs are felicitous and proposing an analysis in Section 3.3.2. Section 4 extends the proposal to pluractional motion verbs with singular subjects. Section 5 concludes.

2. Pluractional Motion Verbs vis-à-vis Other -(I)ş Verbs

This paper investigates -(I)ş-marked motion verbs in Table 2.2 These verbs are not merely distinguished from other pluractional verbs based on their base verbs’ semantics. This section offers a short overview of the syntactic criteria that set pluractional motion verbs and other -(I)ş verbs apart. In particular, we focus on the argument structural differences between the base verb and the derived pluractional verb.

The pluractional motion verbs under investigation here are all marked with the suffix -(I)ş, which is perhaps best known for deriving verbal reciprocals denoting symmetric predicates. Reciprocal verbs exhibit two notable argument structure-related characteristics: (i) they seemingly reduce the base verb’s valency, and (ii) they introduce an instrumental argument.

Reciprocals are often associated with valency reduction3 (see Siloni 2012, and references therein) because of examples such as (3), where the base verb bak- ‘look’ takes a dative argument (see Semih’e ‘Semih-dat’ in (3a)), but the reciprocal verb in -(I)ş does not take a dative argument, illustrated by the ill-formed in (3b). (3d) shows that the argument of a base verb cannot be replaced with the anaphor birbiri- ‘each other’, despite this strategy being available for the non-pluractional-marked verb in (3c).

(3)	a.	Figen Semih’-e  bak-tı.
		Figen Semih-dat  look-pst.3
		‘Figen looked at Semih.’
	b.	*Figen Semih’-e  bak-ış-tı.
		Figen Semih-dat  look-plrc-pst.3
		Intended: ‘Figen and Semih looked at each other.’
	c.	Figen  ve  Semih   birbirin-e   bak-tı.
		Figen and Semih each.other-dat look-pst.3
		‘Figen and Semih looked at each other.’
	d.	Figen  ve  Semih   (*birbirin-e)    bak-ış-tı.
		Figen and Semih  (*each.other-dat) look-plrc-pst.3
		‘Figen and Semih looked at each other.’

Second, the -(I)ş-marked reciprocal verb does introduce an instrumental argument, which is demonstrated by the acceptability of an instrumental-marked anaphor, birbiri-yle ‘each other-INSTR’. The anaphor birbiriyle cannot occur in an adjunct position, which we attribute to the anaphor’s licensing requirement. For instance, the anaphor, unlike the genuine adverb beraber ‘together’, is ungrammatical in the adjunct position in (4a). In contrast, birbiriyle is acceptable in the reciprocal construction in (4b), suggesting that the reciprocal introduces an instrumental-marked argument, which can be filled by the anaphor.

(4)	a.	Figen  ve    Semih   (beraber    /   *birbiri-yle)         yemek  tarif-in-e
		Figen  and  Semih   (together   /   *each.other-ins)  food     recipe-cm-dat
		bak-tı.
		look-pst.3
		‘Figen and Semih looked at the recipe (together / *with each other).’
	b.	Figen  ve    Semih  (birbiri-yle)         bak-ış-tı.
		Figen  and  Semih  (each.other-ins)  look-plrc-pst.3
		‘Figen and Semih looked at each other.’

The pluractional motion verbs’ base verbs are intransitive, and the pluractional does not reduce their valency. More importantly, neither the base verbs uç- and kaç- (in (5a) and (6a)) nor the pluractional motion verbs derived from these in (5b) and (6b) are well-formed with the anaphor birbiriyle. This indicates that the pluractional motion verbs pattern unlike reciprocal verbs based on their argument structure: (i) they do not co-occur with (apparent) valency reduction, and (ii) they do not introduce an instrumental argument.

(5)	a.	*Kuş-lar  birbiri-yle     uç-tu.
		bird-pl  each.other-ins  fly-pst.3
		Intended: ‘The birds flew with each other.’
	b.	*Kuş-lar  birbiri-yle     uç-uş-tu.
		bird-pl  each.other-ins  fly-plrc-pst.3
		Intended: ‘The birds flew helter-skelter with each other.’
(6)	a.	*Çocuk-lar  birbiri-yle   kaç-tı.
		child-pl   each.other-ins  flee-pst
		Intended: ‘The children fled with each other.’
	b.	*Çocuk-lar  birbiri-yle   kaç-ış-tı.
		child-pl   each.other-ins  flee-plrc-pst
		Intended: ‘The children fled helter-skelter with each other.’

At first glance, the -(I)ş-marked motion verbs appear to be a subtype of collaborative verbs such as ağla-ş- ‘cry together’, and gül-üş- ‘laugh together’. However, upon closer inspection, anaphor-licensing patterns suggest otherwise. First, consider the simplex verb ağla- ‘cry’, which cannot co-occur with the instrumental anaphor birbiriyle in (7a). This is consistent with the observations we made above about the distribution of the anaphor (see (4a), (5a) and (6a)). However, the -(I)ş-marked collaborative verb ağla-ş- ‘cry together’ does accept the instrumental anaphor in (7b). The striking contrast between (7a) and (7b) suggests that pluractional collaborative verbs introduce an instrumental argument.

(7)	a.	Figen ve   Semih  beraber     /  *birbiri-yle   ağla-dı.
		Figen and Semih  together   /  *each.other-ins  cry-pst.3sg
		‘Figen and Semih cried together/ *with each other.’
	b.	Figen ve   Semih  birbiri-yle    ağla-ş-tı.
		Figen and Semih  each.other-ins  cry-plrc-pst.3
		‘Figen and Semih cried with each other.’

Importantly, collaborative verbs such as ağla-ş- ‘cry together’ contrast with pluractional motion verbs such as uç-uş- ‘fly helter-skelter’ and kaç-ış- ‘flee helter-skelter’ in that collaborative verbs can co-occur with instrumental anaphors but not -(I)ş-marked motion verbs.

Finally, -(I)ş inchoatives might also involve an apparent valency reduction process such as decausativization (Reinhart 2003) or anticausativization (Koontz-Garboden 2009). On the face of it, these derive intransitives from transitive verbs. For instance, kır-ış- ‘wrinkle (intr.)’ is derived from the transitive kır- ‘break (tr.)’ (see Gandon 2013 for such argumentation about -(I)ş inchoatives).4 Importantly for us, the -(I)ş inchoatives’ subject is never an agent but an undergoer, as is expected from a construction that derives anticausatives. This is in stark contrast with the pluractional motion verbs, whose subject is always an agent.

This section has aimed at offering syntactic evidence for treating collective(-looking) pluractional motion verbs as a distinct type of category.

Table 3. Pluractional types distinguished by syntactic criteria.

	Reciprocals	Collaboratives	Inchoatives	Pl. Motion
	bak-ış -	ağla-ş -	kır-ış -	uç-uş -
	‘look at e.o.’	‘cry together’	‘wrinkle (intr.)’	‘fly helter-skelter’
Reduces valency	?	✗	?	✗
Instrumental anaphor is OK	✓	✓	✗	✗
Subject can’t be agent	✗	?	✓	✗

offers a summary of the introduced syntactic diagnostics. The tests indicate that pluractional motion verbs pattern differently from other types of -(I)ş-marked verbs based on (i) whether it reduces the base verb’s valency, (ii) whether it introduces an instrumental argument (which, as we argued, corresponds to the acceptability of the instrumental anaphor), and (iii) whether its subject can be an agent. Based on these syntactic diagnostics, we proceed to treat the pluractional motion verbs as a distinct category and attempt to account for their semantic contribution in the following sections.

3. Events with Intersecting Paths

3.1. The Problem with Collective Verbs

As mentioned in the introduction, existing accounts consider the pluractional motion verbs’ defining characteristic to be their “collective” property, i.e., that they (obligatorily) involve more than one (agent) participant. Gandon (2013) (following Lichtenberk 1985) defines “collectivity” as a situation where two or more participants with identical roles are jointly involved. An illustrative example is offered in (8): the pluractional motion verb kaç-ış- ‘flee helter-skelter’ can only compose with a plural5 subject in (8b); the singular subject in (8a) renders the sentence ungrammatical.

(8)	a.	There was an explosion in a building, and a child is trying to find a way out. She doesn’t know where the exit is, so she erratically runs in all directions.
		*Çocuk  kaç-ış-tı.
		child     flee-plrc-pst
		Intended: ‘The child fled helter-skelter.’
	b.	There was an explosion in a building, and children are trying to find a way out. They are fleeing in all directions in a haphazard way.
		Çocuk-lar  kaç-ış-tı.
		child-pl      flee-plrc-pst
		‘The children fled helter-skelter.’

The purported plural-subject requirement also led Kemmer (1993, pp. 96–99, 116–24) to categorize verbs such as uç-uş- as “naturally collective verbs”.6 This is one of more than a dozen middle-marked categories Kemmer identifies, all of which she claims share a low elaboration of events and/or participants. These verbs contrast with “distributive verbs” in the extent to which they distinguish the individual actions of group members (i.e., agents): “naturally collective verbs” express events with low individuation because the individual actions of the agents are not distinguished, as opposed to “distributive verbs”, which are characterized by higher levels of event individuation. Kemmer stipulates that it is an inherent property of “naturally collective verbs” to compose with plural subjects.

This paper lays out a very different interpretation of the pluractional -(I)ş’s contribution. In particular, we take issue with three aspects of Kemmer’s approach, which we show cannot be maintained: 1. That the events denoted by such -(I)ş verbs are associated with “low individuation”, (see empirical counterarguments in Section 3.2). 2. That pluractional motion verbs must have multiple agent participants (see Section 4 for empirical data showing that uç-uş- and koş-uş- can compose with a singular subject). 3. That there is a direct connection between the contribution of -(I)ş and plural subjects.

These points fit right in the linguistic discussions on the relationship between the predicate and external argument (Kratzer 1996, and subsequent work by many others). The overarching question is how a verb can determine the properties of its external argument, e.g., its number property. Given well-established theories of how verbs combine with their external arguments (Alexiadou et al. 2015; Harley 2013; Kratzer 1996; Legate 2014; Pylkkänen 2008, inter alia), it seems implausible that the verb could directly impose such a restriction on its external argument. Kratzer (1996) and subsequent research build on semantic work by Dowty 1991; Parsons 1990 and Krifka 1992, who argue that arguments combine with their event via secondary predicates called thematic roles. Kratzer (1996) establishes that this logical model is extendable to the syntax of verb phrases and their external arguments, which are not introduced by the base predicate itself but by a secondary predicate headed by Voice, as shown in the representation in (9). Thus, given such a theoretical model, it seems unlikely that a verb could directly impose a plurality restriction on its external argument. The answer that we are proposing to this question is that the plurality requirement is not imposed directly by the -(I)ş-verbs in Table 2, but rather that they define event plurality in such a way as can only be satisfied by plural subjects. The remainder of this section explores how -(I)ş verbs construe event plurality and, therefore, indirectly force their agent to be plural.

(9)

A further problem with Gandon and Kemmer’s approach concerns the empirical facts. As mentioned in the introduction, pluractional motion verbs such as uç-uş- in (2a) are not always felicitous when the “collectivity” approach predicts that they should be. The following Section 3.2 turns to the discussion of the empirical landscape, investigating in detail under what scenarios the pluractional motion verbs are felicitous. After the discussion of the data, Section 3.3 turns to the analysis and combines the insights from the empirical investigation with the above-discussed theoretical concerns relating to the plural agent requirement to formulate a novel analysis to model the pluractional head’s semantic and syntactic contribution.

3.2. Properties of Pluractional Motion Verbs

In this section, we look at the properties of -(I)ş motion verbs.7 The first observation to make is that these verbs are not acceptable in all multiple event contexts. (10) outlines a scenario where there are multiple flying events, performed by multiple birds, yet uç-uş- is not felicitous. This observation invokes our discussion in the introduction about pluractionality in (2a), where uç-uş- is not acceptable when the birds fly in a perfect V-formation. A further similar example is (11), where multiple children perform simultaneous running events on a running track, yet koş-uş- is not appropriate in this context.

(10)	A group of birds were lined up on a wire. A sudden sound scared them, and they flew off in such a way that they flew straight ahead, without crossing any other bird’s path, i.e., in the following way:
	#Kuş-lar  uç-uş-tu.
	bird-pl    fly-plrc-pst.3
	Intended: ‘The birds flew (together).’
(11)	Children were running on a track in their own lanes, i.e., in the following way:
	#Çocuk-lar  koş-uş-tu.
	child-pl      run-plrc-pst.3
	Intended: ‘The children ran (together).’

Examples (2a), (10) and (11) underscore two crucial points: (i) the contribution of the pluractional is not reducible to denoting multiple flying or running events; (ii) and pluractional motion verbs are subject to further restrictions beyond having to have multiple (agent) participants. These observations challenge existing accounts of -(I)ş-verbs, which are unable to deal with the nuances of the empirical data, and they indicate that we need to define how the plural events relate to each other.

The problem with (2a), (10) and (11) appears to be related to how the flying and running events are carried out. In (2a) and (10), the flight paths are parallel, whereas in (12), their paths intersect. Only (12) is an appropriate context for uç-uş-. Similarly, the running events in (13) have haphazard (intersecting) paths, which makes the pluractional koş-uş- felicitous.8

(12)	A group of birds were lined up on a wire. A sudden sound scared them, and they flew off (at the same time) in a random way, some of them to the right, some of them to the left, i.e., in the following way:
	Kuş-lar  uç-uş-tu.
	bird-pl    fly-plrc-pst.3sg
	‘The birds flew away helter-skelter.’
(13)	Children were running around in a big field in random directions, i.e., in the following way:
	Çocuk-lar  koş-uş-tu.
	child-pl     run-plrc-pst.3
	‘The children ran helter-skelter.’

Additionally, the denoted events must be performed at the same time. For instance, uç-uş- cannot be uttered in a context such as (14), where the birds, one after the other, fly up from the wire in a random way.

(14)	A group of birds were lined up on a wire. One bird after the other flew up in the air in a random way, some of them to the right, some of them to the left. (They didn’t fly away at the same time.) E.g., the following way:
	#Kuş-lar  uç-uş-tu.
	bird-pl    fly-plrc-pst.3
	Intended: ‘The birds flew away helter-skelter.’

In addition, there must be a larger number of flying events—a simple plurality of events, e.g., two birds flying in the specified way, as in (15), is not allowed.

(15)	*Kırmızı  kuş-la  mavi  kuş   uç-uş-tu.
	red   bird-ins   blue  bird   fly-plrc-pst.3
	Intended: ‘The red bird and the blue bird flew about helter-skelter.’

Thus, our first approximation is that the -(I)ş-verbs in Table 2 display three important properties: they denote a large number of events (see (15)) that are performed contemporaneously (see (14)) and whose paths must intersect (see (10)).

3.3. Analysis

3.3.1. Lasersohn’s Model of Pluractionality

The previous section showed that the pluractional does not merely require multiple events to be performed by multiple participants, as this approach does not allow us to account for the different meanings that arise with pluractionals. To be able to account for the relevant meanings that -(I)ş-predicates can construe, we need to take their temporal and spatial properties into consideration.

In order to do this, we adopt a formalism that considers verbs to denote event properties (Dowty 1991); these events compose with their arguments via secondary predicates, called thematic roles, which denote a relation between individuals and events (Carlson 1998; Kratzer 1996; Krifka 1992; Parsons 1990). Additionally, we also assume temporal and spatial trace functions ($\tau$ and $\sigma$), which are functions from events to their run time and space, respectively (Krifka 1998).

We capture the contribution of -(I)ş-verbs by appealing to Lasersohn’s (1995) analysis of pluractional verbs,9 which is embedded in a wider account of linguistic plurality.

The basic formulation of pluractionality is (16), which says that all events denoted by the pluractional predicate are such that they are in the denotation of the base predicate (V(e)) and the cardinality of the set of events is greater than n (CARD(X) ≥ n), where n refers to a lexically or pragmatically set number, e.g., 2 for some pluractionals, i.e., according to (16) if the base verb is fly, the pluractional would say that there are multiple events and all of them are flying events. However, this definition might fall short of capturing the pluractional’s contribution in some instances.

(16)	V-plrc(X) ⇔ ∀e ∈ X[V(e)] ∧ CARD(X) ≥ n
		(Lasersohn 1995, p. 251)

In order to better characterize pluractional meanings, Lasersohn introduces the overlap parameter (○), which models how the events relate to each other. This account is formulated within the above-sketched-out theoretical framework where events are mapped onto their running time, space, and participants by temporal trace functions, spatial trace functions, and thematic roles, respectively (Carlson 1998; Kratzer 1996; Krifka 1992, 1998; Parsons 1990). To represent, for instance, an iterative pluractional meaning, i.e., where the events in the denotation of the base predicate are temporally consecutive, Lasersohn proposes (17). (17) says that for all events e and e’ in the denotation of the pluractional predicate, e and e’ satisfy the base predicate (V(e)), and the temporal trace functions ($\tau$) of event e and e’ do not overlap (¬$\tau$(e) ○$\tau$(e’)), and the cardinality of the denoted set of events is greater than n (CARD(X) ≥ n), i.e., temporally consecutive events can be modeled by requiring the temporal trace functions of the denoted events to not overlap.

(17)	V-plrc(X) ⇔ ∀e,e’ ∈ X[V(e) ∧ ¬$\tau$(e) ○$\tau$(e’)] ∧ CARD(X) ≥ n
		(Lasersohn 1995, p. 251)

Additionally, he introduces the “between” predicate to capture potential downtime (or the lack thereof) and spatial gaps (or the lack thereof) between events. For instance, (18) models iterative events with downtime between them, i.e., a “separated in time” reading. The bold clause in (18) says that there is a time t between the running time of the events e and e’ such that no event e” in the denotation of V that was performed at time t.

(18)	V-plrc(X) ⇔ ∀e,e’ ∈ X[V(e) ∧ ¬$\tau$(e) ○$\tau$(e’) ∧ ∃t[between(t,$\tau$(e),$\tau$(e”)) ∧ ¬∃e”[V(e”) ∧ t = $\tau$(e”)]]] ∧ CARD(X) ≥ n
	(Lasersohn 1995, p. 254)

In the following section, we make use of this account to describe the meaning contribution of -(I)ş.

3.3.2. Proposal

Our proposal in (19) is that -(I)ş motion verbs denote events, such that they are in the denotation of the base predicate (V(e)), and that the events’ temporal traces ($\tau$) are identical ($\tau$(e) = $\tau$(e’)) and their spatial traces ($\sigma$) overlap ∧$\sigma$(e) ○$\sigma$(e’)); the cardinality of the set of events is greater than or equal to the pragmatically/lexically determined value n (CARD(X) ≥ n).

(19)	V-(I)ş(X) ⇔ ∀e,e’ ∈ X[V(e) ∧ V(e’) ∧ $\tau$(e) = $\tau$(e’) ∧ $\sigma$(e) ○$\sigma$(e’) ∧ $\sigma$(e) ≠$\sigma$(e’)] ∧ CARD(X) ≥ n

This definition can account for the infelicity of the pluractional motion verbs in contexts such as (2a), (10) and (11), where the motion paths are either parallel or in some sort of formation. The main insight in (19) is that chaotic paths, as in (12) and (13), can be modeled by requiring the spatial trace functions to overlap ($\sigma$(e) ○$\sigma$(e’)), which requires that the event paths intersect. This rules out parallel and V-formation paths because these paths do not intersect with each other.

This raises the question of whether the pluractional allows for the paths to fully overlap. This is easiest to show with the verb koş-uş-. The running paths in (20) satisfy the spatial traces overlap requirement ($\sigma$(e) ○$\sigma$(e’)) by depicting a scenario where the paths fully overlap (the children are running behind each other). However, the pluractional motion verb is not compatible with the context in (20). To exclude such scenarios, we added the “$\sigma$(e) ≠$\sigma$(e’)” clause to the definition in (19), which states that the events’ spatial traces cannot be identical.

(20)	Children were running in the same lane following each other, i.e., in the following way:
	#Çocuk-lar  koş-uş-tu.
	child-pl      run-plrc-pst.3
	Intended: ‘The children ran helter-skelter.’

(14) illustrated that requiring the motion paths to intersect is not sufficient. We also need to say something about the events’ time. (14) showed that if the flying events in random directions happen after each other, the pluractional is not felicitous. This shows that there is a requirement for the events to take place at the same time. This intuition is reflected by the $\tau$(e) = $\tau$(e’) part of the definition, which requires the events’ temporal traces to be identical.10 One might wonder whether requiring the event times to be fully identical is too strong; for instance, this would technically require each event to start and end at the same time. While this definitely seems to be too strong of a requirement, the issue might be related to the speaker’s perception of the event, e.g., during the relevant perceptual occasion, all the birds were flying. What happens before or after the speaker saw the birds flying is irrelevant.11

Importantly, as pointed out by an anonymous reviewer, simple overlap between the temporal traces (i.e., $\tau$(e) ○$\tau$(e’)) might not be strong enough to derive the empirical data. Imagine the scenario in (21). Here the running paths overlap, just as the pluractional requires. The events’ temporal traces also intersect here: the temporal end point of one event overlaps with the starting point of another event, i.e., there is temporal overlap between events, yet the pluractional koş-uş- is not possible. This leads us to think that simple temporal overlap analysis would not be tenable.

(21)	Children are sitting on a bench. One child gets up and runs towards a tree. The moment when the child gets to the tree, a second child gets up and runs towards a different tree. At the exact moment when the second child arrives, a third child gets up and starts running, and so on.
	#Çocuk-lar  koş-uş-tu.
	child-pl      run-plrc-pst.3
	Intended: ‘The children ran helter-skelter.’

Finally, the definition in (19) states that the pragmatically/lexically determined value of n (i.e., the cardinality of the event set) must be (much) greater than 2 because -(I)ş motion verbs are not preferred with just two events, shown in (15).

In sum, (19) can account for the three main properties of the -(I)ş motion verbs: they denote (i) multiple, (ii) contemporaneous, and (iii) spatially intersecting events. Crucially, the analysis in (19) makes no mention of plural subjects, but it can still derive the plurality of participants. -(I)ş predicates that satisfy (19) cannot be performed by a singular external argument because it is impossible for a singular agent to perform multiple events that are both contemporaneous and have intersecting paths. Thus, the multiple agent condition is just a consequence of the pluractional’s denotation. This is a welcome consequence of our analyses, as there is no need to design a mechanism that would allow the verb to directly determine the number of its external argument.

4. Extending the Analysis: Motion Verbs with Singular Subjects

The above discussion focused on pluractional motion verbs with plural subjects. This section turns to -(I)ş-marked motion verbs that can combine with singular subjects, illustrated by (22). The first thing to note is that we found speaker-variation in the acceptability of singular-subject -(I)ş verbs: out of four consultants, two straightforwardly rejected such sentences. These speakers report that there is no context where such sentences are appropriate.

(22)	%Kuş  uç-uş-tu.    (2 ✓, 2 ✗)
	bird   fly-plrc-pst.3sg
	‘The bird flew helter-skelter.’

Out of the investigated verbs, only koş-uş- and uç-uş- can co-occur with singular agents (at least for some speakers); kaç-ış- cannot (for any speaker we consulted), as in (23a) where the intended meaning is that one child flees from the school in a frantic, distraught way. Kaç-ış- is not compatible with the singular-subject ‘rabbit’ in (23b) under a slightly different scenario: here, there is a repeated threat that sends the rabbit fleeing multiple times (where the rabbit flees in multiple directions).12

(23)	a.	There was an explosion in a building, and a child is trying to find a way out. She doesn’t know where the exit is, so she erratically runs in all directions.
		*Çocuk  kaç-ış-tı.
		child     flee-plrc-pst.3
		Intended: ‘The child fled helter-skelter.’
	b.	The speaker is trying to capture a rabbit in their garden. The speaker keeps sneaking up on the rabbit but the rabbit is escaping every time.
		*Tavşan  kaç-ış-ıyor.
		rabbit     flee-plrc-prog.3
		Intended: ‘The rabbit is fleeing helter-skelter.’

4.1. Properties of Singular-Subject Motion Verbs

Given the discussion in the previous section, it comes as no surprise that singular-subject pluractional motion verbs are not felicitous in just any context where one bird performs multiple flying events. In (24a), the bird performs multiple flying events in a spatially linear way. Expectedly, the pluractional verb cannot be appropriately used in this context. (24b) describes a context where the bird performs flying events on a regular basis (frequentative meaning), but the pluractional is not felicitous here either.

(24)	a.	A bird flew (linearly) from one tree to another, to another, to another, to another (many times). E.g., the following way:
		#Kuş  uç-uş-tu.
		bird  fly-plrc-pst.3sg
		Intended: ‘The bird repeatedly flew.’
	b.	The speaker keeps a bird in a cage but lets it out to fly every day.
		#Kuş  uç-uş-ur.
		bird  fly-plrc-aor.3sg
		Intended: ‘The bird frequently flies.’

A context where (22) can be felicitously used is given in (25). Here, the bird performs frantic, fast-paced flying events in random directions.13 This is clearly parallel to the observation about intersecting paths in the previous section, i.e., singular-subject and multiple-subject pluractional motion verbs exhibit identical spatial properties.

(25)	The following image represents a cage with one bird. The bird frantically flies from one side of the cage to the other. The bird’s path is indicated by the arrows:
	%Kuş  uç-uş-tu.    (2 ✓, 2 ✗)
	bird   fly-plrc-pst.3sg
	‘The bird flew helter-skelter.’

If the two structures’ spatial properties are the same, one wonders how they differ in their temporal properties. Recall that with multiple subjects, -(I)ş motion verbs denote simultaneous (or temporarily largely overlapping) events. (26) shows that the singular-subject pluractional is infelicitous if the event times do not overlap. In this context, there is a one-minute long gap between the individual flying events, resulting in the infelicity of uç-uş-. Note that in the appropriate context (25), the flying events show overlap, as the end time of one flying event overlaps with the start time of the next flying event.

(26)	The following image represents a cage with one bird. The bird’s flight is indicated by the arrows. Each arrow signifies one flying event, after which the bird rests for one minute.
	#Kuş  uç-uş-tu.    (4 ✗)
	bird   fly-plrc-pst.3sg
	‘The bird flew helter-skelter.’

Curiously, the agent’s speed also seems to make a difference. (27) is a variation on (25): these contexts are identical but for one component; the bird in (27) flies not frantically but slowly. This seems to make a difference in terms of acceptability. Speakers who judge (25) as possible note that (25) is a much more suitable context for the sentence than (27).

(27)	In a large cage, a bird flies from one end of the cage to the other; the bird’s path is indicated by the arrows. The bird is flying without stopping anywhere. The bird is not flying particularly fast or frantically. It just slowly glides from one end of the cage to the other.
	#?Kuş  uç-uş-tu.    (2 ✗, 2 ??)
	bird  fly-plrc-pst.3sg
	Intended: ‘The bird flew helter-skelter.’

4.2. Proposal

Based on this discussion, it is clear that there are similar components underlying both the plural and the single-subject pluractional motion verbs’ denotations. (12) and (25) illustrate that both of these require the motion paths to intersect, which suggests that both denotations require the events’ spatial traces to overlap but not to be identical ($\sigma$(e) ○$\sigma$(e’) ∧$\sigma$(e) ≠$\sigma$(e’)).

The main difference between plural and singular-subject -(I)ş verbs seems to be related to the events’ temporal properties. Recall that the definition of plural-subject pluractionals stated that the temporal traces must be identical ($\tau$(e) = $\tau$(e’)) (or largely overlapping). This condition is problematic for single-subject motion verbs because it would require a single actor to perform multiple events with intersecting paths at the same time.

The first thing to point out is that all speakers reject scenarios where the events have no temporal overlap (¬$\tau$(e) ○$\tau$(e’)); this was shown in (26), where the bird takes little breaks between flying events. This suggests that even singular-subject -(I)ş verbs require some level of temporal overlap.

Second, we find the contrast between the felicitous (25) and the infelicitous (26) very revealing: the difference between these examples was the speed of the bird; when the bird is slowly flying from one end of the cage to the other, speakers report degraded acceptability. Naturally, there is no formal way to encode speed into a pluractional denotation. Instead, our proposal is that the singular-subject usage is related to the speaker’s perception and/or their depiction of swift actions. One could imagine that when subsequent events are performed very quickly (especially within a confined space such as a cage), they might appear to an observer as if they are happening at the same time.

This means that the temporal representation of plural and singular-subject -(I)ş verbs might not be that significantly different at all. We propose that some speakers allow a large overlap between the events’ temporal traces (|$\tau$(e) ○$\tau$(e’)|>|¬$\tau$(e) ○$\tau$(e’)|, i.e., the cardinality of the set of temporal overlap is larger than the cardinality of the set of temporally non-overlapping events). These speakers allow events that appear to the observer’s perception as (almost) simultaneous to be depicted as actually temporally overlapping.

Formulating the temporal properties this way allows us to give single and plural-subject pluractional motion verbs a nearly identical denotation. (28) formulates our proposal for the single-subject verbs: the (I)ş-marked verb denotes multiple events (CARD(X) ≥ n) in the denotation of the base predicate (V(e) ∧ V(e’)) such that these events’ spatial traces overlap but are not identical ($\sigma$(e) ○$\sigma$(e’) ∧$\sigma$(e) ≠$\sigma$(e’)), and the cardinality of the set of temporal overlap is larger than the cardinality of the set of temporally non-overlapping events (|$\tau$(e) ○$\tau$(e’)|>|¬$\tau$(e) ○$\tau$(e’)|).

(28)	Denotation of singular-subject pluractional motion verbs
	V-(I)ş(X) ⇔ ∀e,e’ ∈ X[V(e) ∧ V(e’) ∧|$\tau$(e) ○$\tau$(e’)|>|¬$\tau$(e) ○$\tau$(e’)| ∧ $\sigma$(e) ○$\sigma$(e’) ∧ $\sigma$(e) ≠$\sigma$(e’)] ∧ CARD(X) ≥ n

The last remaining puzzle concerns why kaç-ış- is not compatible with singular subjects when koş-uş- and uç-uş- are (for some speakers). We propose that this is related to the base verbs’ aspectual properties. Kaç- ‘flee’ is an achievement, as shown by the fact it can only compose with ‘in five minutes’, as in (29a), but not with the durative adverb ‘for five minutes’ in (29b). Koş- ‘run’ and uç- ‘fly’ are either activity verbs or, when they compose with a bounded path (Jackendoff 1983; Zwarts 2005, 2008), they are accomplishments.

(29)	a.	Patlama-dan   sonra çocuk-lar  okul    bina-sın-dan   beş  dakika   içinde
		explosion-abl after   child-pl    school building-cm-abl  five  minute  in
		kaç-tı-lar.
		flee-pst-3pl
		‘After the explosion, the children fled the school in five minutes.’
	b.	*Patlama-dan    sonra çocuk-lar   okul      bina-sın-dan      beş   dakika   boyunca
		explosion-abl  after   child-pl    school   building-cm-abl   five   minute   for
		kaç-tı-lar.
		flee-pst-3pl
		Intended: ‘After the explosion, the children fled from the school for five minutes.’

For the pluractional kaç-ış- to satisfy the denotation in (28), the punctual fleeing events need to take place at the very same time (to satisfy the temporal overlap criterion). When there are multiple participants, their events can easily take place simultaneously. However, this is not consistent with how a single agent can perform fleeing events in the real world: it is hard to imagine scenarios where someone could flee from (potentially) multiple things at the same time in various directions. Thus, we can derive the lack of singular-subject use of kaç-ış- from the base verb’s aspectual properties.

5. Conclusions

This paper investigated the Turkish -(I)ş syncretism, which includes reciprocal and inchoative verbs but curiously excludes passives and reflexives. The absence of -(I)ş-marked reflexives and passives is an unexpected gap. On the other hand, the suffix -(I)ş is found on verbs denoting collective(-looking) motion, as well as collaborative verbs, categories that cannot even superficially be regarded as cases of detransitivization. For all of these reasons, the -(I)ş syncretism is not an instance of u-syncretism but, as we proposed, a pluractional syncretism.

In order to start working towards a better understanding of the different -(I)ş-marked constructions, this paper has investigated the hitherto understudied pluractional motion verbs, which have been assumed in the literature to be an instance of collective verbs requiring multiple agent participants (Gandon 2013; Kemmer 1993). Focusing on the purported plural-subject requirement, this paper has zeroed in on the conflict between the empirical observation on pluractionals requiring plural subjects and the theory on the relation between predicates and external arguments. Within this theoretical framework (Kratzer 1996), predicates do not introduce their external argument, and thus, it is unclear how they could impose a plurality requirement on them. The solution we propose is that pluractionals do not directly control the properties of external argument. Instead, they define the events in such a way that the predicate can only be felicitously used if it combines with a plural external argument. Thus, the theoretical implication of this approach is that one can maintain the separation of the introduction of external argument from the predicate but still account for the plural-subject requirement.

After observing that overlapping but not parallel movement paths are compatible with pluractional movement verbs, we proposed that the pluractional requires the events’ spatial traces to intersect (but not to be identical). This way, we can model the “chaotic” motion by requiring the events’ paths to intersect but disallow full overlap. Additionally, we offered empirical evidence that the pluractional also mandates the events to be simultaneous. This is accounted for by defining the events’ temporal traces to be identical. Thus, the putative plural-subject requirement can be derived from the events’ temporal properties: if a pluractional denotes temporally simultaneous events, the verb phrase must combine with a plural subject, as it is impossible for a single agent to concurrently perform several events of the same type. Additionally, we showed that some speakers allow the relaxation of the temporal requirement, which allows koş-uş- and uç-uş- to compose with singular subjects.