Caused motion events in Modern Uyghur: a typological perspective

1 Introduction

Recent scholarship has shown constrained but considerable variation in how the cognitive domain of space is structured across languages (Bowerman and Choi 2001; Levinson 2003; Slobin 2004). An extensively researched area in this regard is the lexicalization of motion events (Talmy 1985, 2000) and the present study is part of this line of inquiry. We shall detail Talmy’s influential motion event typology presently, but it is appropriate to note that, while linguists and anthropologists have done their best in sampling a vast array of languages (e.g., Indo-European, African, Asian, Mayan, Austronesian), there are many languages and language families that remain unexplored. Similarly, recent studies on the inter- and intra-typological variations in this domain have mostly concerned European languages (see e.g., Goschler and Stefanowitsch 2013). Moreover, much available research has focused on the lexicalization of voluntary motion (VM) whereas the domain of caused motion (CM) is relatively untapped (but see Hickmann et al. 2018). The present study aims to fill these gaps by focusing on the encoding of CM in an underrepresented member of the Turkic language family, i.e., Modern Uyghur.

2 The Talmyan typology and related issues

Motion events are pervasive in our lives. Examples (1) and (2) depict VM and CM events^[1] respectively while (3) and (4) are their equivalents in Spanish. Both events involve a Figure either moving along a certain Path trajectory by themselves in a particular Manner or causing (Cause) the movement of an Object along a certain Path with reference to a Ground. According to Talmy (1991, 2000, semantic elements such as Figure, Object, Manner, Path, Cause and Ground are universal but the ways they are lexically conflated and mapped onto syntactic constructions differ across languages. At the level of event integration, VM and CM are macro-events consisting of a framing event (Path) and a co-event (Manner or Cause + Manner) and depending on where the framing event is encoded, Talmy classified the world’s languages into satellite-framed and verb-framed languages (S- and V-languages). In English, the framing event is expressed in a particle and the co-events in the main verb. In Spanish, the framing event is expressed in the main verb and the co-events in an adjunct, i.e., gerundive. English alongside other Indo-European languages minus Romance, Finno-Ugric, Chinese and Warlpiri are characterized as S-languages because the core schema/framing event is encoded in a satellite.^[2] Romance, Semitic, Japanese, Tamil, etc. are considered V-languages because the framing event is encoded in the main verb.

Subsequent research added that Talmy’s typological characterization for V-languages applied primarily when events involved a clear crossing of a spatial boundary, as in (3) and (4) (Aske 1989; Özçalışkan 2015). Otherwise, Path can be expressed in satellite-like devices in V-languages as in (5).

Studies also revealed that the typological dichotomy deeply informs how semantic components are mapped onto syntactic constructions and how frequently the components are expressed (Allen et al. 2007; Hendriks et al. 2021; Hickmann et al. 2018; Ji et al. 2011a; Slobin 2004; Tusun and Hendriks 2019; Özçalışkan 2015). S-language speakers typically use compact constructions in which the framing event is encoded in a satellite, thereby freeing up the main verb slot for the co-events. In contrast, V-language speakers typically employ more complex structures wherein the framing event is expressed in the main verb and the co-events in adjuncts (e.g., gerundive, prepositional phrases, adverbials). Consequently, unless the co-events are deemed essential, V-language speakers tend to produce semantically less rich descriptions compared to their S-language counterparts (cf. Harr 2012; Hendriks et al. 2021; Hickmann et al. 2018; Slobin 2003). Furthermore, the sustained use of language-specific lexicalization patterns renders Manner a high-codability domain for S-languages (e.g., rich Manner lexicon) but not necessarily for V-languages (Slobin 1996, 2004, 2006; Slobin et al. 2014; Özçalışkan and Slobin 2003).

While generating the insights sketched above, the potential limitations of the Talmyan typology began to emerge. For example, languages have been found to display much inter- and intra-typological variability in the codability of Manner (Lewandowski and Mateu 2020; Slobin 2004) as well as in the canonical segmentation of Path and the ease with which languages build complex Path constructions (Hijazo-Gascón and Ibarretxe-Antuñano 2013; Ibarretxe-Antuñano 2009). Importantly, languages did not strictly fall on either side of the typological divide but rather formed clines along Manner and Path salience. The observed intralinguistic variation further complicated the picture. For instance, while Chinese speakers were found to use S-framing lexicalization patterns more systematically for VM (Ji et al. 2011b), they used the V-framing pattern as frequently when encoding CM (Ji and Hohenstein 2014; Ji et al. 2011a). Similarly, French and Turkish speakers predominantly used V-framing constructions for VM (Allen et al. 2007; Hendriks and Hickmann 2010; Özçalışkan 2015) but showed greater variability in encoding CM (Furman 2012; Furman et al. 2014; Hendriks et al. 2008).

Moreover, the original typology has been found to apply better to some languages than others. For example, although Talmy classified serial verb languages like Chinese as S-framing, Slobin (2004) argued that, due to the lack of explicit morphological marking, the Manner- and Path-encoding elements in a serial verb should be considered as having equal grammatical status and that such languages be categorized as equipollent-framing (see also Zlatev & Yangklang 2004 for Thai and Ameka and Essegbey 2013 for some African languages). Recently, Lewandowski and Mateu (2020) showed that S-languages in which Path is expressed in morphologically free elements (e.g., English, German, non-prefix-framing) allow more flexibility in terms of the range of Manner verbs combinable with motion constructions than those in which Path is encoded in morphologically bound elements (e.g., Polish, prefix-framing). They thus proposed a finer distinction be made within the S-framing category. Based on an examination of how complex events are encoded in a range of languages, Croft et al. (2010) suggest that Talmy’s typology be elaborated to include additional types (e.g., serialization, coordination, compounding, double-framing).

Since typological research on motion differs in significant ways (written vs. spoken; dynamic stimuli vs. static pictures; single vs. multiple events; VM vs. CM; experimental vs. naturalistic), findings are not always comparable and may have yielded the divergent results we see. Adding a new category of languages to the typology seems untenable and the various proposals to revise the Talmyan typology largely remain to be tested. Meanwhile, it seems necessary that we also revise our conception of the typology. Looking at much of available research around the typology, one is but struck by a strong sense that it concerns entire languages whilst typologists insist that typological generalizations apply to constructions/constructional strategies that encode an equivalent state of affairs, i.e., events, within a language, not to languages as a whole (Croft 2003).

In specific relation to motion expression, Croft et al. (2010) have recently demonstrated how such insights from linguistic typology can be applied to and benefit motion event typological research. Based on how motion events are encoded in several languages, Croft and colleagues identified a formal scale of degree of morpho-syntactic integration (e.g., satellite framing < verb framing < coordination), which seems to broadly correspond to a conceptual scale (e.g., ‘run out of’ < ‘run into’ < ‘crawl to’ < ‘run across’ < ‘dance across’). Specifically, they noted that the motion events to the left of the scale tend to elicit constructions on the left of the formal scale. The argument is that ‘into’/‘out of’ is construed as conceptually more common than ‘across’ and cross-linguistically, the former is encoded in more integrated constructions (e.g., S-framing) and the latter in less integrated ones (e.g., V-framing, coordination). Under this conception of the typology, the affiliation of a language to one or the other type may be an epiphenomenon of the available morphosyntactic resources and constructional strategies in a given language to encode individual events (cf. Beavers et al. 2010; Croft et al. 2010; Lewandowski and Mateu 2020).

In this study, we aim to identify the most typical constructional strategies that Uyghur speakers use to encode CM events and how they compare with speakers of other languages encoding the same set of events. We follow the tradition of Hendriks et al. (2008) and Hickmann et al. (2018) in examining CM scenarios that contain multiple semantic components (e.g., ‘a man pushed the box up the roof’), which is distinct from some previous studies on placement events (e.g., ‘put the pencil in the box’) (Kopecka and Narasimhan 2012). Our assumption is that, when confronted with events comprising a variety of semantic components, the speaker will have a wider range of choices to make for online verbalization in terms of what semantic components to select for expression, how they are packaged lexically and syntactically. This will presumably magnify similarities and differences within and across typological groups. Our two-pronged objective is to offer a more comprehensive typological characterization of Uyghur and shed light on motion event typology with evidence from a Turkic language.

3 The Turkic language family

Turkic languages constitute a large family of some 40 languages, spoken by approximately 170 million speakers across a vast geographical region from the Balkans in the west through the Caucasus and Central Asia into eastern Siberia and western China (Pereltsvaig 2017). Major characteristics of these languages include vowel harmony (e.g., Turkish: kitap-lar ‘books’ (book-PL) versus şişe-ler (bottle-PL) as underlined),^[3] extensive agglutinative morphology (e.g., Turkish: liste-ler-im-de ‘on my lists’ (list-PL-1SG.POS-LOC); Uyghur: bali-lir-i-ɣa ‘for their children’ (child-PL-3PL.POS-DAT)) and head-final constituent order with dependents preceding their heads (e.g., Uyghur: Biz qizziq hikayi-ler-ni oqu-duq ‘we read interesting stories’ (we interesting story-PL-ACC read-PST.3PL)). Turkic languages have both finite and non-finite predicative forms. Finite items constitute independent sentences. Non-finite predicates, on the other hand, are based on action nouns, participles and converbs, of which converbs are especially relevant in spatial discourse.

Turkic converbs are adverbial forms of the verb that signal various semantic relations to the content of the superordinate clause. The relation is said to be modifying if the converb or the converbial clause provides further information that determines, characterizes, modulates, explains or comments on the base clause event (e.g., Turkish: gülerek cevap verdi; Uyghur: külüp javap berdi, ‘laughingly he answered’). In the context of motion events, it expresses the Manner and/or Cause of the event expressed in the main clause. A non-modifying relation is one in which the converbial clause, though syntactically dependent, represents an event of equal narrative value with that of the head clause (e.g., Turkish: oturup konuştular; Uyghur: Ular olturup sözleşti ‘(they) sat and talked’). Non-modifying converbs are ideal for narrative linking and serve as central text-constructing units in traditional narrative styles (Johanson 1995; also see Kornfilt 2009). With these general features of Turkic languages in mind, we now take a brief look at how motion events are encoded in them.

4 Expressing motion events in Turkic languages

As the largest member of the Turkic language family, Turkish is one of the widely studied languages regarding motion expression. Talmy (2000) originally categorized Turkish as a V-language and later studies confirmed that it indeed exhibits traits characteristic of V-languages, viz., Motion per se and Path are conflated in the main verb, and Manner (if expressed) is typically found in converbs. Two types of converbs are noted as especially prominent in Turkish spatial discourse: those marked by -(I)p functioning to link clauses together in narrative units, packaging constituents of an event into a larger event and those marked by -(y)ArAk which express such meanings as simultaneity, succession, instrumentality, reason or manner (Aksu-Koç 1994; Göksel and Kerslake 2005; Lewis 2000). Other important linguistic devices for encoding motion include case markers and in the case of CM, causative suffixes as well (Furman 2012; Furman et al. 2014; Özçalışkan 2015; Özçalışkan and Slobin 1999, 2003).

Examples (6) and (7) below respectively depict VM and CM events in Turkish. In both cases, the framing event is encoded in a Path verb and the co-event, i.e., Manner in (6) and Cause (7) in an -(y)ArAk converb. In both examples, additional Path information is provided via the dative case marker and in (8), which constitutes a placement situation, the Path verb is suffixed with a causative marker. When both subevents are simultaneously encoded, they are typically distributed over two clauses, as opposed to their English counterparts in (10) and (11) that package them in one compact clause.

Empirical research showed mixed results for Turkish. We mentioned in Section 2 that its verb-framing property is more obvious for VM than CM. In specific relation to CM, Furman (2012 identified a range of typical Turkish constructions as exemplified in (7–9). She noted that, while (9) is a satellite-framed pattern not typical of V-languages,^[4] it was the most frequently used construction type, especially in the absence of boundary-crossing (cf. Furman et al. 2014).

Looking at another Turkic language, i.e., Uzbek, Vandewalle (2016) analyzed motion expressions in a corpus and showed that Uzbek too is a typical V-language. Examples (12) and (13) respectively illustrate VM and CM and regardless of event type, Path is expressed in Path verbs and co-events in an -(I)b converbs, the Uzbek equivalent of -(y)ArAk converbs in Turkish.

In this article, we build on relevant information from these languages and examine how CM events are represented in yet another Turkic language, i.e., Uyghur. Hahn (2006) lamented that, despite its immense importance to comparative Turkic linguistics, Uyghur had been greatly neglected among Turcologists outside China and the Soviet Union. More than a decade on, the situation has not seen much change. Although our main motivation has been to establish the typological status of Uyghur in terms of motion expression and to shed fresh light on issues around the intra-/inter-typological variations and the nature of the motion event typology itself, it is also hoped that this work will constitute a modest contribution to Turkic linguistics more generally.

5 Linguistic means available for encoding motion in Uyghur

Uyghur belongs to the Eastern branch of the Turkic language family. It represents, together with Uzbek, the closest descendant of Chaghatay Turkic, which was the pre-twentieth-century transregional literary language of Central Asia (cf. Boeschoten and Vandamme 1998). Uyghur is primarily spoken in northwestern China’s Xinjiang Uyghur Autonomous Region (XUAR for short) with approximately 11 million native speakers (Memtimin 2016) and another half a million Uyghur diaspora in Kazakhstan, Kyrgyzstan, Uzbekistan, Tajikistan, Turkmenistan, Afghanistan and Mongolia (Johanson 2009). It is written in a reformed version of Arabic scripts and, as a co-official language of the region alongside Mandarin Chinese, it is used in media, publication and education to varying degrees. Within the XUAR, Uyghur functions as a lingua franca among the other ethnic minority groups such as Kazakh, Kyrgyz, Tajik, and Tatar (Hahn 1998; Ragagnin 2015).

Uyghur displays morphological, phonological, and syntactic properties characteristic of Turkic languages. In broad strokes, it is an agglutinative language and its rich suffix morphology is subject to sound harmony. As an SOV language, its constituent structure is left-branching with the modifiers preceding their heads. As far as expressing motion is concerned, case markers and converbs are among the frequently used linguistic devices.

Opinion is divided as to the exact number of case categories in Uyghur, with some suggesting six (Hahn 1998) and others ten (De Jong 2007; Tuohuti 2012) but all scholars recognize the dative and ablative cases that are directly relevant to spatial discourse. The dative case usually indicates movement towards, into and onto something (i.e., Goal) whereas the ablative case marker denotes movement away from or through something (the Source or medium of motion). Of the different types of converbs in Uyghur (see Zakir 2007 for a classification), the -(I)p converb is most relevant to expressing motion. It is constructed by adding an -(I)p suffix to the verb stem and performs modifying and non-modifying functions. In terms of the former, it connotes an action that is concurrent to the action indicated in the main verb and in motion event terminology, this type of -(I)p converb expresses the co-event component of motion, i.e., either Manner or Cause of motion. The non-modifying -(I)p converb serves as a clause-chaining device (see Section 6) where one or more events of equal narrative value with the base clause events are conjoined (cf. De Jong 2007; Johanson 1995; Xu 1997; Zhao and Zhining 1985). Simply put, the -(I)p converb serves modifying and non-modifying functions in Uyghur, as it does in Uzbek. This is different from Turkish, as we saw in Section 4, where the -(y)ArAk converb is used for modifying and the -(I)p converb largely for non-modifying purposes (Lewis 2000).

Below are examples of VM and CM events in Uyghur. Examples (14) and (15) illustrate VM in which the Agent, i.e., bowaq ‘baby’ moves along a particular Path (‘towards’) in a particular Manner (‘crawl’). In (14), Manner is expressed in the main verb and Path conveyed via a combination of a dative case marker and a postposition and in constructional terms, it is S-framing. Example (15) describes a closely related motion situation but here, Manner is encoded in a converb and Path in the main verb. It is a V-framing construction and the trigger for this seems to lie in the boundary-crossing constraint (Aske 1989; Özçalışkan 2015). Specifically, the Agent in (15) crosses an explicit spatial boundary, i.e., from being outside of the kitchen to being inside it and for such situations, a Path verb is typically required in Uyghur.

Example (16) is identical to (15) except that an additional deictic component of Path is expressed. While (15) is open to two readings in that the speaker can be either inside or outside the kitchen when reporting the baby’s entry, the speaker of (16) is unmistakably outside of the kitchen and the baby’s entering is construed as movement away from the deictic center, i.e., the speaker. The deictic component is expressed by a combination of a Path verb kir ‘enter’ and an aspectual/auxiliary verb ket ‘go’. The verb ket, originally meaning ‘go (away)’, belongs to a set of verbs in Uyghur whose lexical meanings have become bleached due to grammaticalization and have come to express various types of aspectual information (see Tömür 2003; Tuohuti 2012 for a detailed discussion). However, according to Tuohuti (2012), when ket occurs in an aspectual verb construction as in (16), it resumes its original directional meaning. We propose that it is the deictic dimension of Path that is expressed, not dissimilar to the use of deictic means in verb-serializing languages like Chinese as shown in (17) (Croft et al. 2010; Slobin 2004).

The following examples illustrate CM. In (18), co-event Cause is encoded in the main verb at ‘throw’ and Path via a dative case marker. It is an S-framing construction. Examples (19), (20) and (21) portray the same event but with some subtle differences. All three of them entail Hekim causing the ball to ascend to the rooftop. The difference is that, while Cause is expressed by a causative suffix added to the Path verb in (19), it is lexicalized in a semantically specific converb ittirip ‘push’ in (20) and (21). They are all V-framing constructions, varying in syntactic complexity and the number of semantic components packaged: (18) and (19) are mono-clausal constructions that express two (i.e., Cause, Path) and three components (i.e., Cause, Manner, Path) whereas (20) and (21) express three semantic components in bi-clausal constructions.

A further noteworthy difference concerns (20) and (21). Although they are identical in syntactic complexity and the number of semantic components expressed, they differ in what additional information the aspectual verbs add to the overall meaning of the motion construction. In (20), the Path verb čiq ‘ascend’ is followed by a causative suffix and the verb at. Originally meaning ‘throw’, at belongs to the category of aspectual verbs alluded to earlier and connotes that the object is being disposed of in a particular way. Note that the /p/ and /a/ sounds connecting čiq-ir-ip and at-ti become /w/ due to a process of liaison, and therefore, the aspectual phrase is typically spoken and written as čiq-ir-wet-ti as in (20). In (21), the aspectual verb, coupled with the Path verb čiq adds the deictic element of Path. A final difference between the two examples concerns whether the agent follows the same Path trajectory as the object. It is ambiguous if Hekim ascends to the rooftop alongside the ball in (20) but he clearly does in (21).

Finally, Example (22) below illustrates the clause-chaining function of the -(I)p converb in Uyghur. The utterance consists of four clauses where the first three are marked by the -(I)p suffix and the last verb is inflected for person, number and finiteness. This is an example of the non-modifying use of the -(I)p converb, where each clause, indicated by [c], is read as a separate information unit. The first three clauses are syntactically dependent on the main clause (‘make food’), but informationally understood as coordinated to information expressed in the latter.

We have now explored all the linguistic devices necessary for encoding motion events in Uyghur. Similar to what we saw for Turkish and Uzbek above, case markers are frequently used to encode additional dimensions of Path (e.g., Source or Goal of motion). Importantly, both V- and S-framing constructions are possible across different event types. But does this mean that Uyghur is of a mixed typology? In an empirical study, Tusun and Hendriks (2019) found that adult Uyghur speakers consistently use V-framing lexicalization patterns in such contexts, regardless of whether or not the events involved boundary-crossing (ACROSS vs. UP/DOWN).

6 The present study

This study is motivated by the general research question: what is Uyghur’s status in motion event typology? We take as our point of departure Tusun and Hendriks’ (2019) observation, based on a study of VM situations, that Uyghur is a V-language and examine whether and to what extent lexicalization patterns identified for VM obtain for CM. Specifically, we analyze CM expressions by Uyghur speakers along three parameters that are considered important for profiling the typology of a language, i.e., the linguistic devices used to encode various semantic components, the number of components simultaneously expressed and how they are packaged syntactically within a motion construction. To shed light on issues around inter- and intra-typological variation and the nature of the typology itself, we will compare our findings with those of existing studies on typologically different (English) and similar languages (Turkish, Uzbek and French).

7 Results and analyses

7.1 Information locus

7.1.1 Information in the MAIN VERB locus

We first examined the types of information expressed in the MAIN VERB locus. Figure 1 shows information expressed in this locus as a function of item type. It is clear that Uyghur speakers predominantly used the main verb locus to encode Path, followed by Path with Cause via a causative suffix, which occurred in all item types except INTO. Encoding Cause + Manner was rather rare and occurred only in relation to ACROSS events. Differences across lexicalization patterns were significant for ACROSS (χ ²(2) = 14.006, p < 0.001), UP (χ ²(2) = 15.182, p < 0.001) and DOWN (χ ²(2) = 15.944, p < 0.001) events. Pairwise comparisons within these item types showed the differences between Path and Path + Cause to be significant (p < 0.001 for each). Thus, as predicted, Uyghur speakers used the MAIN VERB locus to primarily encode Path information.

Figure 1:

Information in the MAIN VERB as a function of item type.

Two points should be noted in relation to the data. We mentioned that expressing Manner only occurred for ACROSS events, most typically A-Manner, i.e., walk. Since the main verb slot that is usually reserved for encoding Path is now occupied by A-Manner, Path information is indicated via the case markers as in (33). Note that the speaker could have encoded Path in the verb by using one more converb as in (34). In fact, one speaker did give such a response wherein all the major semantic components were expressed by stacking a series of converbs. While this pattern is grammatically acceptable, it seems to be pushing the boundaries in terms of how many modifying converbs an utterance can accommodate in Uyghur. At any rate, our analysis clearly shows that Uyghur systematically displays one of the hallmarks of a V-language, i.e., encoding Path in the main verb.

(33)

Popi	harvi-ni	söre-p	yol-niŋ
Popi	pram-ACC	pull-CONV	road-GEN
a čéti-din	ma čéti-ge	maŋ-d-i.
that side-ABL	this side-DAT	walk-PST-3SG
‘Popi walked from that side of the road to this side while pulling the pram.’

(34)

Popi	harvi-ni	söre-p	yol-niŋ
Popi	pram-ACC	pull-CONV	road-GEN
u-čét-i-din	bu-čét-i-ge	méŋ-ip	öt-t-i.
that-side-3POS-ABL	this-side-3POS-DAT	walk-CONV	cross-PST-3SG
‘Popi crossed from that side of the road to this side walking while pulling the pram.’

The second point concerns the occurrence of Cause + Path in the verb locus illustrated in (35) below. As mentioned earlier, this lexicalization pattern occurred most frequently with ACROSS events, much less frequently with UP and DOWN events but not at all with the INTO events. A hypothetical example of encoding Cause + Path in the verb locus for INTO events is given in (36) but its absence in our data seems related to the semantics of different item types.

(35)

Popi	yaɣač	čaq-ni	yol-din	öt-küz-d-i.
Popi	wooden	wheel-ACC	road-ABL	cross-CAUS-PST-3SG
‘Popi made the wooden horse cross the road.’

(36)

Popi	üstel-ni	öŋkür iči-ge
Popi	table-ACC	cave inside-DAT
ittir-ip	kir-güz-d-i.
push-CONV	enter-CAUS-PST-3SG
‘Popi made the table enter the cave by pushing it.’

As explained in Section 6.1, our stimuli involved an agent displacing an object in a particular Manner along a particular Path, and crucially, the Agent followed the same Path as the Object. In Talmy’s (2000) terminology, the stimuli represent extent-durational causation, i.e., the object continues in motion as the ongoing result of an extended force impingement without which it would stop. The English CM construction “I pushed the box across the ice” is ambiguous as to whether the Agent travels along the same Path as the object and whether the external Cause is present throughout the event or only at the beginning. Similarly, in constructions such as (35) and (36), the main (Path) verb is predicated of the object, which results in the same ambiguity that we see in English. The difference with Uyghur is that the strength of the ambiguous interpretation seems to interact with the semantics of the denoted item type. Specifically, item types such as UP, DOWN and ACROSS represent extended Paths in that movement along such Paths “invokes the idea of a continuous sequence of locations” whereas INTO evokes “a binary transition from one place into another with nothing in between” and “is a matter of all-or-nothing” (Cappelle and Declerck 2005: 899). While adding Cause to the Path verb via a causative suffix results in ambiguous readings for all item types in Uyghur, doing so with INTO events seems to more strongly bias the interpretation towards one where the Agent sets the Object in motion while he himself stays put. This is inconsistent with our experimental stimuli and perhaps explains why this lexicalization pattern was completely absent for INTO events.

7.1.2 Information expressed at the CONVERB locus

Figure 2 illustrates information expressed in the CONVERB locus as a function of item type. As we can see, Cause + C-Manner and Cause + O-Manner occurred for all item types and Cause-only occurred with ACROSS, UP and DOWN events. Examples (37) to (39) respectively represent the three lexicalization patterns. Across the four items, the most dominant lexicalization pattern is Cause conflated with Manner of Agent’s action (Cause + C-Manner, e.g., pull/push), followed by Cause with Manner of Object’s movement (Cause + O-Manner, e.g., roll/slide). Differences across lexicalization patterns were significant for all item types (INTO: χ ²(2) = 16.825, p < 0.001; ACROSS: χ ²(2) = 13.815, p = 0.001; UP: χ ²(2) = 14.728, p < 0.001; DOWN: χ ²(2) = 15.3, p < 0.001). Pairwise comparisons within each item type found the differences between Cause + C-Manner and Cause + O-Manner to be significant (p < 0.001 for each), as were the differences between Cause + O-Manner and Cause-only for ACROSS (p = 0.007), UP (p = 0.001) and DOWN (p < 0.001) events. Thus, in line with our prediction, Uyghur speakers used the CONVERB locus to express co-events.

Figure 2:

Information in CONVERB as a function of item type.

(37)

Popi	soɣat-sanduq-ni	ögzi-ge
Popi	gift box-ACC	roof-DAT
ittir-ip [Cause + A-Manner]	čiq-t-i.
push-CONV	ascend-PST-3SG
‘Popi ascended to the rooftop while pushing the gift-box.’

(38)

Popi	čaq-ni	dumulut-up [Cause + O-Manner]
Popi	tire-ACC	roll-CONV
ambar-ɣa	kir-d-i.
garage-DAT	enter-PST-3SG
‘Popi entered the garage while rolling the tire.’

(39)

Popi	un-ni	qum taɣ	üsti-ge
Popi	flour-ACC	sand dune	top-DAT
él-ip [Cause]	čiq-ti/é-čiq-t-i.
take-CONV	ascend-PST-3SG
‘Popi ascended to the top of the sand dune while taking the flour.’

(40)

al-ıp	gel.
take-CONV	come
‘to bring’
(Johanson 1995: 315)

A word should be said about Example (39). This example literally translates as ‘He ascended to the top of the sand dune while taking the flour’ and as such, the converb él-ip ‘taking’ does not necessarily encode Cause, i.e., there is no overt causal relationship between the Agent’s and Object’s movement.^[8] While this interpretation may hold for European languages like French and Spanish, it does not for Uyghur and perhaps other Turkic languages (see e.g., Vandewalle 2016 on Uzbek).

In his discussion of Turkic converb clauses, Johanson (1995) notes that linkage between the converb clause and the base clause varies within a converb construction. Specifically, he distinguishes several types of linkage (what he calls ‘levels’) of a converb construction depending on whether the base segment with which the converb segment forms a construction constitutes a full predication, a more reduced one, or a more limited predicative element. Of these, type 3, which he illustrated with a Turkish example as in (40), is highly pertinent to our discussion. Johanson argues that the converb segment and the base segment in such examples together have a single actancy pattern. He notes that insertion of elements between them is highly restricted or excluded and importantly, there is a strong semantic fusion, i.e., representation of one single event and a tendency towards lexicalization (Johanson 1995: 315). All these observations regarding Turkish converb constructions fully apply to similar instances in Uyghur (e.g., (39)), which justifies why we considered the converb segment él-ip ‘take-CONV’ as encoding Cause. We should further add that Johanson’s observation about the strong semantic fusion manifests phonetically in Uyghur in that, in colloquial speech, the converb él-ip ‘take-CONV’ is systematically reduced to é with no perceptible pause between the two clausal segments.

We now discuss the proportions of the él-ip category in Figure 2. Example (41) illustrates such a response where a semantically specific causative converb (Converb1) is used alongside the semantically general él-ip converb (Converb2) within one construction. These double-converb constructions are different from those exemplified in (37) and (38) where co-events are expressed in a single converb. It is clear from Figure 2 that, while the double-converb constructions occurred for all item types, they were consistently much less frequent than single-converb constructions (Cause + C-Manner and Cause + O-Manner combined). And the difference between the two construction types was significant for all item types: INTO (χ ²(1) = 8.542, p = 0.007), ACROSS (χ ²(1) = 10.649, p = 0.001), UP (χ ²(1) = 7.969, p = 0.004) and DOWN (χ ²(1) = 8.986, p = 0.002). Thus, it seems that Uyghur speakers sometimes employ double-converb constructions to highlight the causative nature of the denoted events.

This double-marking of Cause is reminiscent of Bohnemeyer et al. (2007) and Croft et al. (2010) proposal on the double-marking of Path. Specifically, these scholars noted that, in some V-languages, Path is marked once as a detached satellite (more on this in the next section) and once as part of the verb. Uyghur speakers’ use of ‘double-converb’ constructions demonstrates that double-marking can also appear for co-events in CM. In fact, Ji (2014) document a similar phenomenon in Chinese, as in (42).

(41)

Popi	su’üzüš-čaqi-ni	ögzi-ge
Popi	swimming-ring-ACC	roof-DAT
dumult-up (converb1)	[Cause + O-Manner]	él-ip (converb2) [Cause]
roll-CONV		take-CONV
čiq-t-i (é čiq-ti).
ascend-PST-3SG
‘Popi ascended the roof while rolling (and) taking a swimming ring.’

(42)

Xiao3ai3ren2	tui1 zhe	lun4zi	cong2 ma3lu4
little man	push DUR	wheel	from street
zuo3bian1	tui1-dao4	you4bian1.
left side	push-arrive	right side
‘The little man pushed the wheel from the left side of the street to the right side pushing it.’
(Ji 2014: 211)

7.1.3 Information expressed in the OTHER locus

The OTHER locus referred to devices other than the main verb and the converb, i.e., dative and ablative case markers. Our analysis showed that as much as 82% of the motion descriptions contained Source and/or Goal of the framing event (Vector in Talmy’s terminology) (cf. Talmy 2000: 53–57). Example (43) is a typical response:

(43)

Popi	bir sévet	mévi-ni	yol-niŋ
Popi	one basket	fruit-ACC	road-GEN
oŋ terép-i-din	sol terép-i-ge	ittir-ip	öt-t-i.
right side-3POS-ABL	left side-3POS-DAT	push-CONV	cross-PST-3SG
‘Popi crossed while pushing a basket of fruit from the right side of the road to the left side.’

These are instances of what Bohnemeyer et al. (2007) and Croft et al. (2010) call ‘double-encoding’ of Path information. Uyghur speakers’ tendency for Path elaboration could be partly attributed to the combined effect of case marking and the canonical word order of the language. Numerous studies have shown that languages with rich morphological resources facilitate the codability of Path due to the easy processability of such devices (cf. Aksu-Koç 1994; Özçalışkan 2009; Slobin 2004) and this tendency cuts across typological divides (Ibarretxe-Antuñano 2009). The easy accessibility of case markers, coupled with the verb-final sentence structure of Uyghur means that, during online production, Uyghur speakers can describe various dimensions of Path (e.g., Vector-Source and Goal) in a relatively compact fashion before reaching the main verb that encodes Path (e.g., Vector and Conformation). For speakers of verb-initial (V-)languages, the very first element encoding Path is the main verb itself and any additional Path information may seem redundant (cf. Hijazo-Gascón and Ibarretxe-Antuñano 2013).

7.2 Utterance density

We differentiated two levels of utterance density: UD2 for utterances with two semantic components and UD3+ for those with three or more semantic components.^[9] Examples (44)–(46) illustrate responses containing two, three and four semantic components. Figure 3 shows that across the four item types, UD3+ was predominant whereas UD2 utterances were much less frequent. A Density (UD2, UD3+) × Item type (INTO, ACROSS, UP, DOWN) analysis revealed a significant interaction (χ ²(3) = 19.475, p < 0.001), suggesting that utterance density varied as a function of item type. An item-type effect was identified for UD2 (χ ²(3) = 18.433, p < 0.001), such that ACROSS events occasioned more UD2 responses than INTO (p = 0.020), UP (p = 0.020), and DOWN events (p = 0.05). However, no such item-type effects were found for UD3+ responses. That is, irrespective of item type, Uyghur speakers consistently produced high-density utterances, i.e., UD3+. A qualitative inspection of the data showed that the UD3+ utterances typically expressed Cause + C-Manner + Path (78%) followed by those encoding Cause + O-Manner + Path (15%) and Cause + C-Manner + A-Manner + Path (2%). The UD2 utterances always expressed Cause and Path (4%).

Figure 3:

Utterance density as a function of item type.

(44)

Popi	sanduq-ni	taɣ-din
Popi	box-ACC	hill-ABL
él-ip [Cause]	čüš-t-i [Path].
take-CONV	descend-PST-3SG
‘Popi descended from the hill while taking the box.’

(45)

Popi	soɣat-sanduɣi-ni	ittir-ip [Cause + C-Manner]
Popi	present-box-ACC	push-CONV
qum döŋ	üsti-ge	čiq-t-i [Path].
sand dune	top-DAT	ascend-PST-3SG
‘Popi ascended to the top of the sand dune while pushing the gift box.’

(46)

Popi	bir bala harvisi-ni	yol-niŋ	oŋ-terip-i-din [Path]
Popi	one pram-ACC	road-GEN	right-side-3POS-ABL
sol-terip-i-ge [Path]	ittir-ip [Cause + C-Manner]	maŋ-d-i [A-Manner].
left-side-3POS-DAT	push-CONV	walk-PST-3SG
‘Popi walked from the right side of the road to the left side while pushing a pram.’

Our prediction about overall utterance density is thus confirmed: Uyghur speakers predominantly included at least three semantic components. Previous studies maintained that presenting more semantic information in V-languages might be more complex and involve greater online processing load resulting in V-language speakers offering semantically less dense utterances compared to S-language speakers (Hendriks and Hickmann 2010; Özçalışkan 2015; Özçalışkan and Slobin 2003; Slobin 2004). However, our findings showed that V-language speakers do not necessarily produce semantically less dense descriptions, at least for CM. We will return to this issue in Section 8 but suffice it to say here that the frequent use of converbs in Uyghur may have facilitated the production of high-density utterances (cf. Tusun and Hendriks 2019).

7.3 Syntactic packaging of semantic components

Two main strategies were established for the analysis of syntactic packaging. Tight Simple and Tight Complex (TC).^[10] Example (47) illustrates the TS strategy where semantic components are packaged in a simple clause. Examples (48) and (49) are both TC responses in which co-events are expressed in the converbial clause and the framing event in the matrix clause. Note that while Example (48) contains one converbial clause, Example (49) contains two.

(47)

Adem	mévi-ni	yol-din	öt-küz-d-i [c].
man	fruit-ACC	road-ABL	cross-CAUS-PST-3SG
‘The man made the fruit cross the road.’

(48)

Adem	orunduq-ni	sör-ep[c]
man	chair-ACC	drag-CONV
öŋkür iči-ge	kir-ip	ket-t-i [c].
cave inside-DAT	enter-CONV	ASPV/go-PST-3SG
‘The man entered the inside of the cave while dragging the chair.’

(49)

Adem	reŋlik top-ni	dügül-tüp [c]
man	colorful ball-ACC	roll-CONV
qar taɣ-din	é-p [c]	čüš-t-i [c].
snow hill	take-CONV	descend-PST-3SG
‘The man descended from the snow hill while rolling (and) taking the colorful ball.’

Figure 4 depicts syntactic packaging as a function of item type. TC was the predominant strategy across all item types whereas TS was much less frequent. A Packaging (TS, TC) × Item Type (INTO, ACROSS, UP, DOWN) analysis revealed a significant interaction (χ ²(3) = 25.6, p < 0.001), indicating that the general patterns varied by item type. Significant item-type effects were found for TS (χ ²(3) = 24.184, p < 0.001), with ACROSS items eliciting more TS constructions than other items (p = 0.020 for all). No such effects were found for TC. That is, regardless of item type, Uyghur speakers consistently used TC constructions more frequently than TS constructions.

Figure 4:

Syntactic packaging as a function of item type.

Our results confirmed the hypothesis that the dominant pattern of syntactic packaging for Uyghur is TC. The TS strategy is used across item types, and although certain items (i.e., ACROSS) occasioned more TS responses than others, it consistently remained a peripheral option. This pattern highlights the status of Uyghur as a typical V-language. Note that the measure of syntactic packaging is directly related to that of utterance density. Given that our experimental stimuli featured multiple semantic components that were equally salient and that our speakers were cognitively mature adults who had presumably developed a good understanding of the inherent complexity (e.g., number of semantic components, the temporal simultaneity of various sub-events) of the denoted events, it is expected that they would encode maximal number of semantic components in their verbalizations. As the measure of utterance density showed, this was indeed the case, and doing so necessarily entailed the use of syntactically complex constructions in this language.

8 General discussion

Our discussion will proceed in relation to the two main concerns of this study, i.e., the typological status of Uyghur with respect to motion and how cross-linguistic comparisons illuminate issues of inter-, intra-typological variation and of the typology more broadly. We will first compare our results with a previous study on VM expressions in Uyghur (Tusun and Hendriks 2019) to ascertain whether Uyghur is systematically verb-framed or whether it exhibits varying typological profiles as a function of event types. We will then compare our findings with some studies on CM expressions in English (S-language) on the one hand, and French, Turkish and Uzbek (V-languages) on the other, to see whether and to what extent Uyghur resembles other V-languages and differs from S-languages.^[11]

Our main findings, summarized in Table 1 below, can be recapitulated as follows. When expressing CM, Uyghur speakers consistently encoded Path in the main verb (99%) and the co-event components, i.e., Cause and Manner, in a converb (100%). Making full use of the relevant case markers, they also provided detailed Path information in addition to Path encoded in the main verb (82%). Uyghur speakers also predominantly expressed a minimum of three semantic components as evidenced by the preponderance of UD3+ utterances (96%). Their most typical pattern of syntactic packaging was TC (96%) while the TS pattern was rather infrequent (4%). Contrary to our hypothesis, these patterns were not affected by item type. Overall, the results clearly show that Uyghur is prototypically verb-framing in terms of the lexicalization patterns and syntactic packaging.

The clear verb-framedness we see in CM is consistent with what has been found for VM in Uyghur. Tusun and Hendriks (2019) showed that Uyghur speakers use the main verb to encode Path (83%) and the converb to encode the co-event, i.e., Manner (57%). Speakers also frequently provided extra Path information in the OTHER locus, i.e., dative and ablative case markers (84%) and when they expressed the subevents simultaneously, they typically packaged them in TC constructions (58%). Looking across event types, it is striking how consistently Uyghur speakers distribute the key semantic components across the motion construction, i.e., framing event in the MAIN VERB and co-event in the CONVERB (TC constructions). Furthermore, their tendency for supplying rich Path information outside the verb applies to both event types. What is different, however, is how frequently the co-event is mentioned (Cause + Manner 96% for CM vs. Manner 58% for VM), which is almost always indexed by the type of motion constructions (TC vs. TS) employed.

Tusun and Hendriks (2019) attributed the relatively lower occurrence of TC for VM to the influence of typological factors. Encoding the co-event (Manner) alongside Path typically entails using a subordinate construction for V-language speakers and due to the presupposed cognitive load, they tend to either establish Manner information in surrounding discourse or omit it altogether (Özçalışkan 2015; Özçalışkan and Slobin 2003; Slobin 2004; Talmy 2000). That Uyghur speakers encoded Cause as frequently as Path in a TC construction indicates that Manner and Cause as co-events do not enjoy the same level of centrality, at least for linguistic representation of the event. Recall that Uyghur offers a rather “economical” option for representing the two essential components of CM, i.e., adding a causative suffix to the Path verb (TS). However, speakers hardly used this option (4%) and instead tended to encode Cause doubly, once in the MAIN VERB and once in the CONVERB or in constructions containing two consecutive converbs. The double encoding of Cause underscores how central this component is to a complete representation of CM events. We will revisit this point shortly but suffice it to say that the differences in Uyghur speakers’ use of TC constructions across event types stem from the (potentially) differential importance the two respective co-events hold in relation to the framing event, not from the inconsistency of lexicalization patterns used (as we will see with other languages below). In other words, Uyghur is a highly systematic verb-framed language.

In a similar study, Hendriks et al. (2008) found that English speakers systematically expressed Manner/Cause in the main verb (93%) and Path in a satellite (94%). This meant that they packaged the semantic components mainly in TS constructions (84%) (Hickmann et al. 2018). While French speakers showed a general tendency to express Cause and Path in the main verb, their overall patterns of distributing information were much varied across verbs and other devices (Cause: 72 vs. 28%; Manner of Cause: 63 vs. 37%; Manner of Motion: 80 vs. 20%; Path: 62 vs. 38%). Uyghur patterns strictly with neither of these languages. Befitting its V-framed typology, it showed the opposite lexicalization pattern (Cause/Manner converb + Path verb) of English (Cause/Manner verb + Path satellite) and syntactic packaging strategies. Both putatively V-languages, Uyghur is extremely systematic in its lexicalization patterns while French shows much flexibility. This intra-typological difference has cascading effects on how many semantic components are typically selected for expression and how they are syntactically organized in these languages.

According to Hendriks et al. (2008), 75% of the French responses contained three semantic elements (UD3), in comparison to 92% of such responses by English speakers. Hickmann et al. (2018) observed that French speakers’ preferred patterns for syntactic packaging are more varied as well (TS: 17%; TC: 59%; LS: 12%: LC: 11%) compared to the relatively consistent patterns found in English (TS: 84%; TC: 4%; LS: 10%; LC: 4%). Again, Uyghur is very different to English (except for utterance density) but does not align with French along the two parameters either. In terms of utterance density, Uyghur speakers’ UD3+ responses were close to the ceiling (98%), which was much more frequent than those of French speakers (75%). In terms of syntactic packaging, Uyghur speakers systematically and overwhelmingly used TC constructions (96%), which is even more significantly different from French (59%). LS constructions were extremely rare in Uyghur and LC constructions were absent but they were not infrequent in French. Moreover, although Uyghur and French speakers used TS constructions, closer inspections revealed that they are qualitatively different: while Uyghur TS constructions typically included both the framing and the co-event (Path and Cause in the VERB locus), French TS constructions primarily encoded co-events, with the framing event either expressed in satellite devices or omitted altogether (cf. Harr 2012; Harr and Engemann 2013).

Numerous studies have shown that, despite being generally considered a V-language, French displays much variability in motion event expressions (Hendriks and Hickmann 2010; Hendriks et al. 2008; Hickmann et al. 2018). Some scholars attributed this complex typological profile to the diachronic transition of French from being an S-language to a V-language (Kopecka 2006; 2013). When it comes to the comparison between Uyghur and French, however, we propose that the observed intra-typological variation in how sub-events of motion are lexicalized and organized syntactically partly also stems from their canonical word order.

We contended earlier that Cause/Manner as co-events are more central to a complete verbal representation of CM than Manner per se to VM. However, it seems that speakers of a putative V-language are not necessarily faced with the same type of challenge when verbalizing a complex motion situation featuring multiple but equally essential components. As speakers of a verb-final V-language, Uyghurs typically encode the framing event sentence finally and before reaching the main verb slot during online production, they can and do express the co-events in a converb, which necessitates a TC syntactic frame. For French speakers who speak an SVO language, the challenge seems to be slightly different. In principle, French speakers would encode the framing event early on in the main verb, and to specify the co-events, they would have to append a subordinate structure to the main clause when the basic structure of the sentence has already been completed. While they do tend to omit the co-event in verbalizing VM (see Harr and Hickmann 2013; Hendriks and Hickmann 2010), doing so for CM would be infelicitous precisely because Cause is indispensable for a full verbal representation of this type of motion.

Given the perceptually salient relationship between the agent’s action and the displaced object in the motion events involved, French speakers seem to have overcome the typological and processing constraints of a verb-initial V-language and produced appropriate TC responses fairly frequently (cf. Özçalışkan 2009; Slobin 2004). However, the frequent use of TS, LS and LC constructions also reflects their struggle with syntactically organizing multiple key components of CM in a relatively compact fashion (e.g., TC vs. LS/LC). Indeed, a closer look at the LS and LC responses showed that French speakers did encode both framing and co-events, but they did so in atypical and much less systematic ways (also see Harr 2012; Hickmann et al. 2018). Note that this intra-typological variation between Uyghur and French was also observed in the context of VM (Tusun and Hendriks 2019) and it seems that converbs/converbial clauses as a subordinate structure are a more common phenomenon in Uyghur and perhaps other Turkic languages than gerundive clauses in French. This speculation certainly needs corroboration in future research.

If Uyghur and French display intra-typological variation due to their differences in word order, how does it compare with typologically close Turkish and Uzbek? The picture looks mixed. Uyghur, Turkish and Uzbek are very similar in terms of how Path elaborations are made outside of the main verb. Özçalışkan (2009) showed that Turkish adults and children frequently express additional Path information via case markers and postpositions. The same seems to hold for Uzbek, as can be seen in the examples in Vandewalle (2016). However, there is more variation in Turkish with regard to the most typical lexicalization patterns and the motion constructions they appear in. Furman et al.’s (2006) study on Turkish and English speakers’ CM descriptions in an experimental setting (e.g., ‘The triangle man rolls the tomato man down the hill’) showed that Turkish speakers predominantly encoded the sub-events across two clauses (97%). The authors do not report how the two clauses are syntactically related (e.g., coordination, subordination) nor their respective frequencies, but it is clear from their examples that Turkish speakers certainly used LS and LC constructions (see Furman et al. 2006: 195).

More recently, Furman (2012) examined a wider range of CM events in Turkish (e.g., ‘Ali took the ball down by rolling it’; ‘Ahmet pushed the box to the car’). Despite the fact that Turkish is generally considered a V-language, she found that Turkish speakers predominantly employed TS constructions typical of S-languages (e.g., Cause/Manner in main verb + Path in case markers/postpositions, 80%) while their descriptions substantiating the verb-framing pattern (TC) (Cause/Manner converb + Path verb) accounted for only 20% of the data. She further noted that TS constructions were more frequently used in non-boundary-crossing situations and TC constructions in boundary-crossing ones (Furman 2012: 58). This again contrasts with Uyghur speakers’ consistent use of TC constructions regardless of item types and strict lexicalization of Path in the verb of Uyghur TS constructions.

The intra-typological difference between Uyghur and Turkish may be due to various reasons. It is possible that our results are not entirely comparable to available studies on Turkish. For instance, our experimental stimuli differ from those of Furman et al. (2006) in certain respects (e.g., animacy and naturalness of the agent and object, trajectory they follow, variety of item types). Furthermore, unlike our stimuli that featured a small set of motion verbs (e.g., push, pull, roll, slide), Furman (2012) examined CM involving a much wider range of motion verbs (e.g., throw, rotate, hurl, slide, put, drag, kick, bounce). Also, in light of recent proposals that different manner of motion verbs are subject to different verb-construction combinability constraints (Lewandowski and Mateu 2020), the intra-typological difference in syntactic packaging between Uyghur and Turkish could have stemmed from the disparity in the types of motion verbs examined. It could also be that Turkish speakers’ frequent use TS constructions is related to the observation that contemporary Turkish exhibits an overall decline in the use of prepositive converbial clauses, due to sustained influence from European languages (cf. Johanson 2013). Meaningful intra-typological comparisons in this domain would ultimately require a study that controls for all these relevant factors, but insofar as available studies capture the general typological profile of Turkish, the intra-typological variation between Uyghur and Turkish is unmistakable.

What do the inter- and intra-typological variations mean for motion event typology itself? Given our findings that Uyghur is a typical V-language, one possibility is to see it as a continuum where Uyghur and English sit at the opposite ends representing prototypical V- and S-languages while Turkish and French occupy a space along the cline as they display both V- and S-framing properties. However, doing so is ultimately unhelpful and may in fact obscure some important issues. For example, it would lead us to classify entire languages while evidence continuously shows that languages rather have patterns that do not fit their overall typology. Consequently, we might miss out on insights into the cognitive and linguistic mechanisms that inform speakers’ choices for both the typical and less typical lexicalization patterns within and across languages (see also Beavers et al. 2010; Lewandowski and Mateu 2020 for a related line of reasoning). Croft et al. (2010) insightfully remind us that typological classification applies to individual complex event types within a language, not languages as a whole and the basic unit of cross-linguistic comparison and contrast is not the language as a whole but each construction that is used to express an equivalent state of affairs, which in the present instance, is an equivalent motion situation.

Our CM stimuli involved combinations of various Cause/Manner (e.g., pull, push, roll, slide), Path (e.g., UP, DOWN, INTO and ACROSS) and Ground features (e.g., sand dune, rooftop, cave, road). Uyghur speakers systematically used verb-framing constructions/strategies for encoding these events and we argue that Uyghur is a prototypical verb-framed language insofar as encoding these and perhaps similar events are concerned. Of course, this does not mean things are so relative that one cannot make typological generalizations, for it is possible to establish the basic type out of the many structural types a language uses (Croft 2003). However, this construction/strategy-based conception of the typology necessarily implies that speakers may use different constructions/strategies if confronted with other motion situations.

This observation also brings us to the limitation of this study. As mentioned earlier, the motion situations our speakers were presented with involved a limited number of motion verbs. However, recent scholarship (Lewandowski and Mateu 2020; Mateu 2012) has shown that Manner verbs differ in their inherent or associated directionality, which determines their combinability with various motion constructions. For example, although Spanish is considered a V-language, Lewandowski and Özçalışkan (2018) report that Spanish speakers produce CM constructions like Tira la piedra a un estanque – ‘s/he throws the stone into a pond’ in which Manner/Cause are expressed in the verb. According to Lewandowski and colleagues, causative verbs such as throw evoke directionality, which is why they can appear in the main verb slot in V-languages (cf. Footnote 4). Indeed, in the Uyghur equivalent of this construction, U taš-ni köl-ge at-t-i (s/he stone-ACC pond-DAT throw-PST-3SG), or a similar example, U süt-ni piyali-ge quy-d-i (she milk-ACC cup-DAT pour-PST-3SG) – ‘she poured the milk into the cup’, Manner/Cause are expressed in the main verb locus. Future studies should examine the full range of motion verbs and of motion constructions in Uyghur to fully characterize the typology of this language.

Finally, our discussion highlighted that even languages sharing dominant lexicalization patterns differ in a myriad of ways (e.g., word order, morphosyntactic resources, language contact profiles). To achieve a deeper understanding of the intricate relationship between universal aspects of event cognition and language-specific principles of event representation, we need to explore how various types of motion events are encoded in as many situation types and languages as possible. Specifically, we can develop motion scenarios that vary in principled ways (e.g., animate vs. inanimate agents and objects, direct and indirect causation, differential naturalness and centrality of co-events, diverse Path trajectories, single event vs. multiple events) and examine how they are represented within and across languages. In this regard, Croft et al.’s (2010) implicational scales potentially underlying inter- and intra-linguistic variations in encoding complex events is a promising perspective to explore.

9 Conclusions

This study set out to establish the status of Modern Uyghur in motion event typology with a focus on caused motion events. Systematic analyses of several aspects of adult Uyghur speakers’ descriptions showed that Uyghur is a typical verb-framed language. Specifically, Uyghur speakers systematically encoded the framing event in the main verb and the co-event in a converb. The systematicity facilitated the simultaneous encoding of the core elements of caused motion (Cause, Manner, and Path) and the production of semantically dense UD3+ utterances. Uyghur speakers likewise packaged the framing event in the main clause and the co-event in a subordinate clause (TC constructions) and this general pattern was not influenced by item type. Double encoding of key semantic components (Cause, Path) was very common in that speakers not only used case markers to supply rich Path information (e.g., Source, Goal) in addition to encoding Path in the verb, they also typically encoded Cause twice in their verbalizations. A comparison with existing findings on voluntary motion expressions in Uyghur confirmed that it is a prototypical verb-framed language, regardless of event types.

Our comparisons with English, French, Turkish and Uzbek showed the following. Uyghur was comparable to English in how frequently speakers produce high density utterances (UD3+) but differed in information locus and syntactic packaging (TC vs. TS). Comparisons with French and Turkish, both purportedly V-languages, revealed interesting intra-typological differences. In contrast to the systematic verb-framedness of Uyghur, French showed much flexibility in how semantic information is distributed across linguistic devices and packaged syntactically. While belonging to the same language family as Uyghur, Turkish seems to exhibit strong S-framing tendencies (e.g., dominance of TS constructions). We accounted for these intra-typological variations in terms of the impact of language-specific factors (e.g., word order, presence/productive use of case markers), the possibility of the contact-induced change, and differences in denoted situation types.

Overall, our findings confirmed the basic premises of Talmy’s (1985, 2000 motion event typology while our crosslinguistic comparisons highlighted the need for a reconceptualization of the typology as one of constructional strategies that speakers of different languages use (Croft et al. 2010; Hendriks and Hickmann 2015; Lewandowski and Mateu 2020; Tusun and Hendriks 2019). Future research will benefit by exploring more varied situation types in a wider range of languages and language families so that we develop a better understanding of how universal principles of event cognition interact with language-specific properties in event representation.