The syntactic packaging of caused motion components in a second language: English learners of Chinese

Available online at www.sciencedirect.com

ScienceDirect Lingua 140 (2014) 100--116 www.elsevier.com/locate/lingua

The syntactic packaging of caused motion components in a second language: English learners of Chinese Yinglin Ji a,*, Jill Hohenstein b a

Shenzhen University, PR China b King’s College London, UK

Received 30 June 2012; received in revised form 24 November 2013; accepted 29 November 2013 Available online 24 January 2014

Abstract Previous studies of L2 acquisition of motion expressions have tended to focus solely on Indo-European languages and on spontaneous motion. This study expands the sphere of research by investigating how Chinese L2 learners, at different proficiency levels, acquire arguably more advanced linguistic skills in the syntactic packaging of caused motion information. Our results show firstly that learners are not trapped in their source pattern when packaging a set of particularly dense semantic information (Cause, Path and varied types of Manner), even, if they have not yet entirely acquired the target system. They have arrived at an inter-language, showing considerable resemblance to the target system rather than traces of the L1 influence. Further, no developmental tendency was observed at the initial and intermediate stages of acquisition; changes occurred only when learners progressed to an advanced level, suggesting that, unlike lexicalisation of motion components, syntactic strategies of packaging information are more complex and need to be adapted to over a longer period of time. © 2013 Elsevier B.V. All rights reserved. Keywords: Spatial language; Caused motion events; Development of syntax in adult L2 learners

1. Introduction Although humans share a universal biological heritage in spatial perception and understanding, representation of spatial experience differs significantly across languages. Many studies have revealed that specific properties of a given language constrain the spatial conceptualisation available in linguistic encoding of motion events (see, for instance, Allen et al., 2007; Hohenstein et al., 2004; Berman and Slobin, 1994; Bowerman, 1999; Bowerman and Choi, 2001, 2003; Hickmann, 2006; Levinson, 2003; Naigles and Terrazas, 1998; Slobin, 1996, 2004; Mandy et al., 2010). Recent findings in the domain of space, language and cognition raise important questions, among other things, for second language (L2) acquisition: to what extent can L2 learners shake off their L1 linguistic pattern in order to acquire a new language system? And to what extent do L2 learners have to alter their way of mentally representing motion information for its expression in a non-native language? The present study addresses these questions by focusing on how English learners of Chinese at three proficiency levels (low, intermediate and advanced) acquire the expression of caused motion events, paying particular attention to syntactic representation.

* Corresponding author at: Research Centre for Language and Cognition, Shenzhen University, No. 3688, Nanhai Avenue, Nanshan District, Shenzhen City 518060, Guangdong Province, PR China. Tel.: +86 0755 26749807. E-mail address: [email protected] (Y. Ji). 0024-3841/$ -- see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.lingua.2013.11.009

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

101

According to Talmy (1985), a spontaneous (or voluntary) motion event contains a set of universal semantic elements: Motion itself, Figure (moving entity), Ground (reference frame in relation to Figure’s motion), Path (e.g. up, across, towards, along), and optional Manner (e.g. swimming, jogging, limping). In caused motion events one more element is included, that is, Cause of motion coupled with Manner of Cause (e.g. pushing, pulling, throwing, kicking). As it is grammatically impossible to represent all information as equally important, speakers must make choices about how to encode complex semantic caused motion components. Firstly, they may choose to encode a motion event within the boundary of a single clause, selecting only certain types of components for expression whilst omitting others. Secondly, they may choose to manipulate linguistic devices at their command to encode all information albeit representing with unequivalent saliency across clauses. For example, speakers may foreground certain types of information in the marked grammatical category of verbs in the main clause whilst syntactically backgrounding others in the periphery of an utterance via subordination, dependency, gerunds and/or adverbials. Seen in this way, speakers’ syntactic organisation of particularly dense information within or beyond the boundary of clause presents an interesting topic, which constitutes the focus of investigation in the present paper. Talmy (1985, 2000) classified world languages into two broad categories: Satellite-framed (S-language hereafter) and Verb-framed (V-language hereafter), depending on how the core element of motion, Path, is encoded via grammatical means. In the former, Path is typically encoded outside the verb in ‘satellites’, such as particles, whilst Manner and/or Cause are expressed in main verbs (e.g. English, German, Example 1). In contrast, Path is incorporated into the main verb in V-languages; Manner and/or Cause, if expressed, are encoded independently through use of adverbials or gerundive clauses (e.g. French and most Romance languages, Example 2). (1) (2)

Hoppy rolled the ball across the street. [S-framed] Hoppy traverse la rue en faisant rouler le ballon. [V-framed] ‘Hoppy crossed the street rolling the ball.’

Not all languages fit neatly into this dichotomy. Some have argued (e.g., Slobin, 2004) that there is at least one more type in which multiple motion components can be compactly encoded via co-verbs or verb compounds. For instance, in Chinese several aspects of motion are normally packaged into a Resultative Verb Compound (RVC). The first constituent in an RVC typically encodes Manner and/or Cause, the second constituent expresses Path of motion and the third one indicates the deixis of motion (either lai ‘come’ or qu ‘go’, see Example 3). (3)

球 qiu2 ball ‘The

滚过去 gun3-guo4-qu4 roll cross go ball rolled across

了。 le ASP1 away from the speaker/listener.’

Perhaps controversially, Slobin (2004) argued that the Path constituent in an RVC is a full verb that can stand independently as a predicate (Example 4). In contrast, Talmy (2000) held that Chinese is an S-language because the second constituent in an RVC is a closed class set, and even though still existing as main verbs, these verbs are restricted in use. (4)

球 qiu2 ball ‘The

过 了 马路。 guo4 le ma3lu4 cross ASP street ball went across the street.’

Slobin (2004) termed serial verb languages such as Chinese as ‘equipollently-framed’ (E-framed hereafter) in the sense that in some predicates different semantic aspects of a motion event (Path, Manner, Cause, Deixis) can be simultaneously expressed in grammatical constituents of equal status and with equal formal significance.2 Ji et al. (2011c) further argued that ‘being equipollently-framed’ means that languages like Chinese are neither entirely S-framed nor V-framed. They show S- and V-framing properties depending on such factors as the nature of motion event involved (e.g. spontaneous versus caused).

1 The following abbreviations are used in this paper: ASP, aspectual marker; ASSOC, associative de, genitive/adjectival/adverbial marker; AUX, auxiliary verb; CL, classifier; DEF, definite article; DUR, durative aspectual marker zhe; PAST, past tense; 1sg, first person singular. 2 According to Slobin (2004), a verb in the equipollently-framed language can assume various forms. Apart from verb compounds (i.e. Manner verb + Path verb) in serial-verb languages such as Chinese and Thai, there are bipartite verbs (i.e. [Manner + Path] verb) in languages like Hokan, and cluster verbs such as ‘Manner preverb + Path preverb + Verb’ in Jaminjungan languages.

102

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

The lack of morphology in Chinese makes categorisation of the status of its verbs difficult. However, most sinologists tend to consider the second constituent in an RVC as a verb, or at least a ‘postverb’ sandwiched between verb and particle (e.g. Li and Thompson, 1981; Zou, 1993; McDonald, 1995). Further, recent studies on the typological status of Chinese in more varied motion contexts (voluntary and caused) and modes (oral and written discourses) all suggested that it is best to treat Chinese as an E-language showing both satellite- and verb-framing properties, or a language with parallel framing systems (Chu, 2009; Gao, 2001; Chen and Guo, 2009; Ji et al., 2011a; Talmy, 2009). In this light, the current paper followed Slobin (2004) in treating the Path constituent in an RVC as an independent verb and Chinese as an E-language. 2. Motion expressions in L1 and L2 acquisition Previous studies on first language (L1) acquisition of spatial expressions revealed that typology affects language use in varied spatial domains. Some have argued that the influence of language is limited to the level of linguistic expression only (e.g. Gennari et al., 2002; Malt et al., 2003; Papafragou and Selimis, 2010), while others argued that language-specific properties can even shape spatial cognition at a deeper level (see, for instance, Hohenstein, 2005; Bowerman, 1996; Levinson, 2003; Lucy, 1992). Importantly, the language-specific influence is found to go beyond the semantic level and can be detected at the semantics--syntax interface. Allen et al. (2007) pointed out that semantic components for motion are syntactically packaged in different ways across languages. In S-languages like English, adult speakers usually encoded Manner and Path within one single clause, using tight syntactic packaging (Example 5). In contrast, in V-languages such as Turkish and Japanese, Manner and Path were typically expressed through the use of semi-tight matrix--subordinate construction (Example 6a). Alternatively, speakers of such languages sometimes used a bi-clausal structure, representing loose syntactic packaging (Example 6b). (5) (6a) (6b)

The man hops all the way up the hill. The man ascends the hill hopping all the way. The man goes up the hill, and he’s hopping all the way.

Interestingly, young English learners (3--4 years) were sensitive to cause when expressing motion events. When Manner and Path were causally related, they tended to use tight syntax (Example 7a).3 However, in events in which Manner does not cause changes of location, Manner and Path were more likely to be expressed separately (i.e. semi-tight or loose packaging, Example 7b). (7a) (7b)

The ball jumps down the stairs. The ball jumps while falling down from the stairs.

Compared to the extensive literature on L1 acquisition, few studies have systematically investigated motion expression in a non-native language. Slobin’s (1996) ‘thinking for speaking’ hypothesis, claiming that language directs its speakers to attend, while speaking, to dimensions of experience that are enshrined in grammatical categories, may be useful in addressing L2 motion event description. ‘Thinking for speaking’ involves picking those characteristics of objects and events that (a) fit some conceptualisation of the event, and (b) are readily encodable in the language (75--76). Such training is carried out from childhood and ‘‘is exceptionally resistant to restructuring in adult second language acquisition’’ (89), because learning a new language, in this sense, means adapting to a new way of ‘thinking for speaking’. The few studies testing this hypothesis produced conflicting results regarding the role of L1. For instance, Navarro and Nicoladis (2005) examined whether English-speaking learners of a V-language, Spanish, can acquire the verb conflation pattern. Their results showed that learners were able to map semantic components of motion systematically onto main verbs in a frequency comparable to that of Spanish native speakers, suggesting that they had generally grasped the lexicalisation pattern of V-languages. However, they did not use bare Path verbs as frequently as did native speakers. These findings were at least partially discordant with those of Cadierno (2004), Cadierno and Ruiz (2006) and Hohenstein et al. (2006). The former group of researchers investigated how Danish-speaking (S-language) learners of Spanish narrated stimuli that attempted to depict motion in a static scene. Their results suggested a clear influence from L1: learners used significantly fewer Path verbs to depict motion scenes whilst producing more elaborate Path descriptions via redundant Path particles compared with Spanish natives.

3 Allen et al. (2007:23) used the term ‘Manner-Inherent’ to refer to the phenomenon that manner and cause are causally related in motion. To illustrate, in example 7a, the jumping of the ball causes the change of location of the ball. In comparison, in 7b, manner does not cause a change of location: the jumping of the ball is a natural result of its falling down (i.e. Manner-incidental).

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

103

Two recent studies by Hendriks et al. (2008) and Hendriks and Hickmann (2011) also led to contradictory results. They investigated how English speakers of French selected and placed motion components across an utterance (i.e. participants with L1 English and L2 French). It was reported that L2 learners’ performance varied with motion type. In depicting complex caused motion events (CM hereafter), L2 learners resorted to their L1 pattern: they still encoded Manner in the main verb whilst omitting Path or placing it peripherally if needed, reflecting difficulty in adapting to the more complex target system of motion expressions in French. In contrast, when asked to depict spontaneous motion events, the same group of speakers more readily adopted the target pattern of conflating Path with the main verb. This occurred when the target language provided a lexicalisation pattern that is no more complex than the pattern in the native language, implying that, in communicative tasks where there was a balance to be struck between being maximally explicit and being target-like, learners tended to employ a pragmatic strategy by opting for an ‘efficient’ linguistic system, that is, a relatively uncomplicated system that can get the message across adequately. It is worth mentioning that a recent study by Zeng (2011) examined the acquisition of motion expressions by a different type of learner from ours, that is, learners with L1 Chinese and L2 English. Using both static and dynamic stimuli (frog story pictures and video clips) and expanding across spontaneous and caused motion contexts, this study systematically investigated whether and to what extent the L1 ‘thinking for speaking’ pattern influenced how Chinese learners acquired English motion expressions. Findings suggested that the L1 influence was not as strong as predicted and only visible in certain aspects. Chinese learners across proficiencies generally acquired the characteristic pattern of English motion expressions (i.e. verb + satellite), showing virtually no L1 constraint. However, they differed from English native speakers in: (a) using a very limited number of manner and cause verb types, mostly referring to very general rather than specific and elaborated motion verbs, and (b) allocating more attention to physical setting descriptions of motion events, thus demonstrating some traces of L1 features at the discourse level. As shown above, there are several reasons to pursue further examination of motion event language learning. First, most studies focused on spontaneous motion events, with few investigating caused motion in detail. The current study investigated caused motion events that incorporated maximum semantic components for motion (i.e. manner, cause, path, motion itself, Figure, Ground), thus better addressing the question under discussion, viz., the syntactic packaging of multiple pieces of information. Second, most studies focused on a pair of languages with contrasting typological properties (e.g. S-framed L1 versus V-framed L2, or vice versa); yet relatively few studies involved more varied languages. The present study chose a pair of languages with partial typological similarity (i.e. S-framed English and E-framed Chinese), and presumably partial overlapping in ‘thinking for speaking’ patterns. It thus becomes interesting to see whether such overlap can facilitate L2 learners’ acquisition of the target system such as accelerating their learning pace. Further, previous studies on motion events tended to focus on lexicalisation patterns or rhetorical styles of motion discourses; relatively few studies examined the syntactic information packaging within and beyond the clause level. This paper aims to extend these lines of research, integrating typologically more varied languages, a more complex type of motion event and arguably more advanced linguistic skills of syntactic packaging in a second language learning context.4 3. Additional information about encoding caused motion in Chinese 3.1. BA and ZHE constructions As mentioned above, motion events in Chinese are typically encoded in an RVC. Given that this verb compound is essentially intransitive (Example 8), it needs to be used in conjunction with some specific grammatical constructions in order to adequately express caused motion in Chinese, namely, BA and ZHE constructions. (8)

(9a)

*他 滚过 球。 *Ta1 gun3-guo4 qiu2 he roll cross ball 他 把 球 Ta1 ba3 qiu2 he BA ball ‘He pushed the

滚过 马路。 gun3-guo4 ma3lu4 push cross street ball across the street.’

4 It is worth mentioning that lexicalisation and syntax are often two related issues. Due to limited space, we focused, in this paper, not on the systematic relation between meaning and surface expressions, i.e. which semantic elements are characteristically expressed by which surface elements, but on the syntactic expression of such morphemes as ‘discrete entities’, i.e. how these meaning-loaded morphemes are grammatically arranged across an utterance and how the position of a particular morpheme relates to that of other morphemes in a logical way.

104

(9b)

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

他 Ta1 he ‘He,

滚 gun3 roll rolling

着 zhe DUR a ball,

球 过/走过 马路。 qiu2 guo4/zou3-guo4 ma3lu4 ball cross/walk cross street went/walked across the street.’

In the BA construction, the Figure NP (e.g. the ‘ball’ in 9a) is always positioned in front of the whole RVC and a ba marker is used to signal its status as the only object affected by the causing action. The utterance concerned therefore focuses on the ‘affectedness’ of an entity, viz., how it is disposed of. The ZHE construction involves two clauses with the first one marked with zhe being subordinated to the second one. Given that Manner and Cause of motion are normally expressed in the subordinated verb, the main verb of the ZHE construction is usually reserved for the encoding of Path (e.g. ‘cross’ in 9b), or Path plus an additional type of Manner information (e.g. ‘walk-cross’ in 9b; for more details of these two constructions in Chinese caused motion descriptions, see Ji et al., 2011b). It is important to note that BA and ZHE constructions have different syntactic properties. The former is syntactically simplex and represents a compact form of information organisation (like Allen et al.’s, 2007 tight syntax) whereas the latter is syntactically complex; it is less compact with its matrix--subordinate construct and represents a more discursive way of information packaging (akin to semi-tight packaging in Allen et al., 2007). 3.2. L2 learners’ acquisition of BA and ZHE constructions The BA construction requires the post-ba verbal phrase (VP) to satisfy a complexity constraint given the resultative meaning associated with the construction (i.e. how an entity is disposed of). That is, this post-ba VP must be an RVC or a simplex verb followed by a perfective aspectual marker le to signify the ending of an event (e.g. Li and Thompson, 1981; Zou, 1993). Acquisition studies on the BA construction revealed that learning this construction did not seem to pose particular difficulties to L2 learners. For instance, the results of Du’s (2006) production task data regarding BA acquisition among adult English learners of Chinese, in which the correlation between the production of BA and RVCs (or the perfective marker le), suggested that L2 learners had acquired a good knowledge of the complexity constraint on the postba VP. Additionally, learners’ assessment of the grammaticality of BA sentences was not significantly different from that of natives, confirming that the construction can be acquired by L2 learners without particular difficulty. Although there is an absence of literature regarding L2 acquisition of the ZHE construction, research into the following two aspects might shed some light on the topic under discussion: (a) syntactic and semantic analyses of ZHE constructions; (b) L2 acquisition of complex sentences in general. Regarding the former, previous studies suggest that the order of V1 (suffixed with zhe) and V2 is fixed. Any change of this sequence results in a semantic anomaly (e.g. Li and Thompson, 1976; Chu, 1987). The correct sequencing of the two VPs depends, to a large extent, on a good understanding of the multiple functions of zhe, which include: (a) a durative marker in temporality; (b) a focus marker in semantics: the event headed by V1 is the primary topic of the sentence; (c) a subordinating marker in syntax: V1 is syntactically secondary to V2; (d) a backgrounding device in pragmatics: the event headed by V1 provides background information for the event presented via V2. So in example 9b, the addition of the zhe marker to V1 gun3 ‘roll’ transforms an instantaneous verb into a temporally durative one. Semantically, the focus of the utterance is on the event headed by V1, ‘rolling’; whereas syntactically, V1 with the zhe marker is subordinated to V2. Note that different from English, the subordinated clause always precedes the main clause in Chinese complex sentences. Xing (2002) reported that L2 learners of Chinese with diverse L1 backgrounds (including English learners of Chinese) made fewer errors in producing complex sentences with conjunctive morphemes than with sentences connected by aspectual or punctuation markers. Among the latter type, two patterns of error were frequent: a reversed order of two verbal clauses and a lack of any linking devices between two clauses. Generally it can be hypothesised that L2 learners may face an acquisitional challenge because of the varied types of constraint on the V1--V2 sequence in the ZHE construction and the lack of lexical conjunctive means between the two clauses. 3.3. Rationale of the present study This study examines L1’s influence on the process of acquisition of the target pattern of information packaging in L2 learners. Particular attention is paid to the question of whether English-speaking learners of Chinese recognise there are two syntactic options available in the target language. One is the ‘compact’ pattern of using syntactically simplex BA constructions for information packaging and the other is the ‘semi-tight’ pattern of employing syntactically complex ZHE constructions (i.e. matrix--subordinate) to organise particularly dense information. This study therefore investigates whether learners can partially restructure their L1 ‘thinking for speaking’ pattern and use both options to organise dense information, focusing in particular on the semi-tight pattern, which occurs rarely in their L1 but very frequently in the L2 and can help them better achieve the communicative goal of the task. Specifically, we will examine:

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

(a) (b)

105

To what extent the syntactic packaging of CM information by Chinese L2 learners resembles that of monolingual native speakers. To what extent the syntactic strategy of packaging CM information develops with proficiency among L2 learners.

We propose two hypotheses regarding L2 learners’ performance. Firstly, if language-specific influence is merely superficial, meaning Chinese and English speakers conceptualise caused motion in similar ways during online verbal productions, then shaking off typological constraints of L1 and reorganising motion information in an L2 should not present a major problem for learners. This implies that English learners will resemble native speakers of Chinese in using both BA and ZHE constructions for caused motion descriptions. Considering that some specific properties of the ZHE construction (both semantic and syntactic) may pose acquisitional difficulties for learners as illustrated in section 3.2, we predict that the frequency of opting for ZHE constructions would only increase significantly when learners progress to relatively advanced levels. In comparison, learners across proficiencies would use the syntactically simplex BA structure in comparable frequencies. Secondly, if conceptualisation of motion has been influenced by specific properties of L1, then adapting to a different linguistic system may be a challenge for L2 learners. They would therefore attend to the typological ‘similarity’ between L1 and L2 (i.e. S-framing properties) and rely heavily on the compact BA pattern to encode the CM event. We predict no major developmental changes across proficiencies under this circumstance. Additionally, we predict that it is unlikely that adult L2 speakers, even of low proficiency, would use a disjointed pattern to package CM information (e.g. encoding Manner and Path in two separate clauses), mainly based on the findings that motion components occurring simultaneously tend to be packaged together rather than separately (Allen et al., 2007). Further, an array of clauses, without linking lexical/aspectual morphemes to mark the semantic relationship in between, have very low acceptability in Chinese grammar. Therefore, learners, particularly those at relatively advanced levels, would be reluctant to use coordinated clauses linked by punctuation marks only to encode a motion event. 4. Method 4.1. Participants Participants in the study were two groups of monolingual native speakers (English and Chinese respectively) and three groups of L2 learners of Chinese at three proficiency levels (low, intermediate and advanced). Each of the five groups had 12 participants: six females and six males. Monolingual Chinese native speakers were students from a Technical College in Shandong Province, China and monolingual English native speakers were all students from Stanford University (US). All L2 learners were students studying Chinese as a second language at Peking University and they came from three English-speaking countries: the US, the UK and Australia. Their proficiency levels were determined by an independent language proficiency test administered by Peking University at the beginning of each academic year. At the time of data collection, all learners had studied Chinese at Peking University for a period of six months to three years and had all studied Chinese in their own country for at least six months before coming to China (Table 1). 4.2. Materials Sixteen short video clips (five seconds each), depicting CM events in which both Manner (coupled with Cause) and Path were presented as equally salient, were used to elicit data. The design of the stimuli partially followed Hickmann et al. (2009): both involved four specific types of Manner ( pushing/pulling, rolling/sliding) whilst this study depicted a richer Table 1 Groups of participants in the study. Group ID

Age range

Mean age

L1

L2

Proficiency level

Score range

Mean score

L2_Low L2_Int L2_Ad CH_NS EN_NS

19--61 19--26 18--50 17--18 20--30

25.4 22 23.7 18 23

English English English Chinese English

Chinese Chinese Chinese N/A N/A

Low Intermediate Advanced Native Native

30--44 45--89 90--118 N/A N/A

38.25 69.5 102 N/A N/A

Note: L2_Low: second language learners of Chinese (low proficiency); L2_Int: second language learners of Chinese (intermediate proficiency); L2_Ad: second language learners of Chinese (advanced proficiency); EN_NS: monolingual English native speakers, and CH_NS: monolingual Chinese native speakers. Scores in this table refer to L2 learners’ scores in proficiency tests of Chinese as a foreign language, administered by Peking University each year.

106

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

variety of Path (eight types): verticality (up and down), boundary-crossing (across and into), deixis (towards and away from) and parallel to or encircling the ground (along and around). All clips depicted the same boy performing a specific action, which caused the displacement of an entity. For example, A5 in Appendix A depicted the boy pushing a treasure bag up a pyramid, the bag sliding up the pyramid and the boy accompanying the bag all the way to the top of the pyramid. Half of the participants viewed the stimuli in Order A and half in Order B (Order A reversed). These stimuli conformed to previous models of caused motion developed by Hickmann et al. (2009) in that both presented a specific type of caused motion in which the protagonist accompanied the movement of the object. Each clip provided a particularly rich set of motion elements, suiting the aim of the present study, i.e. to investigate the syntactic strategy of distributing dense semantic information across an utterance. Following Hickmann et al. (2009), the multiple information components presented in each clip included: Cause (C), Path (P) and three types of Manner (M), as shown below: (10a)

(10b) (10c)

Manner of the agent’s causing action (Mc): how the action of the agent caused the change of location of the object (e.g. The boy rolled [transitive use, Mc] the ball across the street). This type of manner focused on the causal relation between the agent and the object; Manner of the object’s motion (Mo): how the object moved under the influence of an external force (e.g. The ball rolled [intransitive use, Mo] across the street); Manner of the agent’s motion (Ma): how the agent moves with respect to a certain path (e.g. The boy walked [Ma] across the street). This type of manner differed from Mc in emphasizing the motor pattern of the movement of the agent.

4.3. Procedure The procedures followed those developed by Hickmann et al. (2009). Participants were invited to describe the video clips in detail to an imaginary remote listener. After each clip, the participant was asked ‘What happened?’ A warm-up item was presented before the test stimuli to direct each participant’s attention to the varied types of CM components s/he was expected to provide during the test.5 4.4. Transcription and coding The data were coded partially according to the system developed by Hickmann et al. (2009) for the English and French data, which was adapted to the Chinese data. The participants’ narration was transcribed by a native speaker of Chinese and later segmented into ‘utterances’, which were defined in our study as either syntactically simplex sentences or syntactically complex sentences with subordinated clauses. Two or more coordinated clauses, marked by conjunctions, or occasionally by a phonological pause in between (in Chinese), were treated as one utterance. Narration depicting the general setting for motion (e.g. ‘This is winter time’) was excluded from analysis. The overwhelming majority of participants used one utterance to depict a motion scene (mean frequency: 99% for EN_NS, 98% for CH_NS, and 96% for each group of L2 learners). In very rare cases where the participant used more than one utterance with all CM components expressed, we coded the first utterance produced. For the sake of convenience, we used a hypothetical English example, which assembled our coding principles, to illustrate how descriptions relevant to CM were chosen (in italics), how they were segmented into ‘utterances’ (indicated by square brackets), and how the target utterance for analysis was determined (in bold). (11)

It was in cold winter and the lake was frozen. [The boy pushed a toy car while he was walking across the ice.] [He was pushing without much effort and going through the ice steadily.] [He finally walked across the lake with his car.]

Since we were interested in how varied types of CM components are organised and distributed over an utterance, we focused only on utterances that had encoded Cause (C), Path (P) and at least one type of Manner (Mc, Ma or Mo). Such utterances were semantically richest, and can thus illustrate more clearly how a speaker may exhaust grammatical means at their disposal to arrange multiple pieces of information in a syntactic hierarchy across an utterance. We therefore excluded from our analysis those utterances that encoded only one or two motion components or expressed no specific

5 As regards the warm-up item (i.e. Benny pulled the boat out of the lake), if the participant’s response to the question ‘What happened?’ lacked explicit Manner or Path information, specific questions such as ‘How did he move the boat?’ or ‘What about the lake?’ were asked in order to elicit desired responses, therefore familiarising the participant with the types of information s/he was expected to provide in the testing session.

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

107

Table 2 Mean frequency of all utterances that encoded zero, one, two or three+ semantic components for caused motion. Group ID

Zero components

One component

Two components

Three+ components

L2_Low L2_Int L2_Ad CH_NS EN_NS

.01 .00 .00 .00 .00

.07 .03 .00 .00 .01

.38 .16 .13 .02 .02

.54 .81 .87 .98 .97

(.02) (.00) (.00) (.00) (.00)

(.08) (.05) (.00) (.00) (.04)

(.26) (.11) (.07) (.05) (.03)

(.31) (.13) (.07) (.05) (.04)

Note: Only responses encoding three+ components constituted the target utterances for analysis in the present study.

motion information (hypothetical examples are given in 13). In our coding, Path was specified as the directionality or trajectory of any translocative movement that necessarily involved a change of locations. Therefore, responses indicating a general location of motion only were not included in our analysis (e.g. pulling a toy car on the ice, rolling a basketball in front of chairs). Proportions of varied types of excluded utterances were given in Table 2 in the following section. Excluded utterances encoding zero, one or two CM components (13a) He was playing basketball in front of the chairs. (13b) He was moving the toy car on the ice. (C only) (13c) He went across the ice with his toy car. (P only) (13d) He was pushing the treasure bag. (C+M) (13e) He was walking into the pyramid. (M+P) For target utterances of C+M+P, four syntactic patterns of information packaging were distinguished, largely following the coding strategy developed by Allen et al. (2007) who detected three syntactic modes of organising motion information in the context of L1 acquisition (i.e. Tight, Semi-tight, Loose). Given that this study involved experimental stimuli of the same nature, that is, caused motion events in which Manner and Path were presented as occurring simultaneously and equally prominent, we adopted this coding framework for the present study but added a fourth type to accommodate any other possible means of syntactically organising dense information in the context of L2 acquisition. Tight: C, M and P were encoded in one utterance consisting of one matrix clause. In English, this referred to syntactically simplex clauses, and in Chinese it mainly included BA constructions. (14a) He was rolling a basketball along a row of chairs. (14b) 他 把 财宝 拖 进 金字塔。 Ta1 ba3 cai2bao3 tuo1-jin4 jin1zi4ta3 he BA treasure pull enter pyramid ‘He pulled the treasure into the pyramid.’ Semi-tight: C, M and P were encoded in one utterance composed of one matrix clause and one subordinated clause. It included adverbial clauses in English and ZHE constructions in Chinese. (15a) The boy went around a small table rolling a barrel of beer. (15b) 男孩 推 着 木柴 走向 火堆。 nan2hai2 tui1 zhe mu4chai2 zou3-xiang4 huo3dui1 boy push ZHE log walk approach fire ‘The boy, pushing logs, walked towards the fire.’ Loose: C, M and P were encoded in one utterance comprising two or more coordinated clauses. (16a) He was sliding his toy car and they went over the ice. (16b) 他 正在 拉 玩具 车, 然后 走过 了 冰 湖。 ta1 zheng4zai4 la1 wan2ju4 che1 ran2hou4 zou3-guo4 le bing1 hu2 he now pull toy car then walk cross ASP icy lake ‘He was pulling the toy car and then walked across the icy lake.’ Pattern in atypical Chinese: This referred to the syntactic structure an L2 learner has adopted in producing Chinese utterances with low acceptability, which mainly included constructions in which two verb phrases are used in concatenation without any intervening conjunction in between.

108

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

Fig. 1. Mean proportion of different syntactic strategies used by monolingual native speakers of Chinese versus English.

(17)

他 拉 车 走 过 冰 湖。 Ta1 la1 che1 zou3-guo4 bing1 hu2 he pull car walk cross icy lake ‘He pulled the car and walked across the icy lake.’

5. Results Table 2 shows the proportion of utterances encoding different numbers of CM components for each of the five groups of participants. Those at intermediate and advanced levels produced a predominating proportion of target utterances with C+M+P. More than half of the responses produced by participants of low proficiency also included multiple CM components. All three groups of learners produced fewer target utterances than did the native speakers.6 5.1. The syntactic packaging of CM components between two groups of monolingual native speakers In this section, we examined how native speakers adopted varied syntactic strategies (i.e. Tight, Semi-tight, Loose and Atypical) to organise a particularly rich set of CM information (C+M+P) across an utterance and whether their performance followed the typological patterns expected for the respective language. Fig. 1 shows the proportion of syntactic constructions employed by native speakers of English versus Chinese in depicting the CM events presented in this task. The above figure illustrates that a dramatic contrast was observed between the two languages. English, as noted, exhibited a clear pattern of systematically wrapping up multiple pieces of information in a highly ‘Tight’ way, in which C and M were encoded in a standard main verb and P in particles (98% in frequency). Chinese, however, employed both ‘Tight’ and ‘Semi-tight’ patterns in roughly equal frequencies. A repeated measure analysis of variance (ANOVA) with language (Chinese, English) as between-subjects factors and syntactic pattern (Tight, Semi-tight) as a within-subjects factor was performed on the raw data and it confirmed this observation.7 It showed an interaction effect between language and syntactic pattern (F (1, 22)=68.475, p=.000) as well as a main effect of syntactic pattern (F (1, 22)=98.995, p=.000). Two paired samples t-tests, performed with Chinese and English native speakers’ data respectively, further revealed that Chinese native speakers adopted ‘Tight’ (M=8.67) and ‘Semi-tight’ patterns (M=7.25) equally frequently (t (11)=0.875, n.s.), whereas their English counterparts predominantly preferred the ‘Tight’ pattern (M=15.67) over the ‘Semi-tight’ one (M=0.25; (t (11)=43.065, p=.000).

6 Issues such as information selection (i.e. other types of responses including only one or two semantic components) and the effect of Path and/or Manner type in CM descriptions are beyond the scope of a single paper and are currently investigated separately. 7 Please note that there wasn’t enough variability in dimensions of ‘Loose’ and ‘Atypical’ patterns to warrant more factors of comparison.

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

109

Example 18 demonstrated how English and Chinese speakers depicted the same video clip using the syntactic devices provided by their native language. (18a) (18b)

(18c)

Benny pushed the bundle of wood towards the campfire. [EN_NS_01A]8 邦尼 把 一 堆 木柴 推 到 篝火 旁。 [CH_NS_02A] bang1ni2 ba3 yi4 dui1 mu4chai2 tui1-dao4 gou1huo3 pang2 Benny BA one CL log push arrive campfire side ‘Benny pushed a pile of logs to the side of the campfire.’ 邦尼 推 着 一 捆 木柴 走 向 了 一 堆 篝火。 [CH_NS_04B] Bang1ni2 tui1 zhe yi4 kun3 mu4chai2 zou3-xiang4 le yi4 dui1 gou1huo3 Benny push DUR one CL log walk approach ASP one pile campfire ‘Benny, pushing a bundle of logs, walked towards the campfire.’

The above results reflected the typological properties of English as a typical S-language and the distinctive status of Chinese as a language with ‘parallel systems’ (i.e. both S- and V-framed) in motion event typology (Slobin, 2004; Talmy, 2009). They were also in line with previous reports that, in describing motion pictures from story books, Chinese speakers tended to package information components in Tight (e.g. The owl flew out of the hole) and Semi-tight ways (e.g. The owl exited the hole flying) with comparable frequencies (Slobin, 2004). Ji et al. (2011c) suggested, in a similar study, that this phenomenon can be explained by the information density of an utterance. In cases where a Semi-tight ZHE construction with an RVC was adopted (Example 18c), one more motion component was invariably encoded in comparison with BA constructions (zou ‘walk’ in Example 18c versus 18b). Further, where a ZHE construction with a single Path verb as its V2 was used (kaojin ‘approach’ in Example 19), three sub-events for motion were readily encoded, whereas both BA and the ZHE construction with an RVC as its V2 can express a maximum of two sub-events. In either case, the Semi-tight packaging created a higher information density of utterances, viz., a greater number of either information components or sub-events for motion. This is vitally important for the current communicative task where complex information needs to be conveyed. (19)

邦尼 推 Bang1ni2 tui1 Benny push ‘Benny, pushing

着 一 堆 木柴 靠近 篝火。 [CH_NS_09A] zhe yi4 dui3 mu4chai2 kao4jin4 gou1huo3 DUR one CL log approach campfire a pile of logs, went towards the campfire.’

Sub-events expressed in 19: (a) (b) (c)

‘Benny pushed a pile of logs.’ ‘Benny went towards the campfire.’ ‘The pile of logs went towards the campfire.’ (inferable from the temporal synchrony between a and b)

While the BA construction (Example 18b) solely emphasised the ‘affectedness’ of an entity and best suited questions, such as ‘‘What happened to the direct object?’’ rather than ‘‘What happened?’’, only sub-events of (a) and (c) (as illustrated above) can be expressed. For the ZHE construction with an RVC, given that V1 and V2 (tui ‘push’ and zou ‘walk’ in Example 18c) shared the same subject, only sub-events of (a) and (b) (as illustrated above) were encoded (see Ji et al., 2011c:1062--1066 for more details). Our findings regarding native speakers of Chinese and English revealed, as expected, significant effects of typology on language use with respect to the syntactic organisation of CM components. We proceeded to investigate, in the following section, whether and how this cross-linguistic variation influenced L2 learners’ progression towards the target. 5.2. The syntactic packaging of CM components among three learner groups Fig. 2 demonstrates the proportion of syntactic structures employed by learners of different proficiencies in packaging various CM components across an utterance. L2 learners seemed to adopt some syntactic strategies that did not entirely resemble either the source pattern or the target pattern, as compared to Fig. 1 in the previous section (see also Table 3).

8 The attribution for each example indicated the participant group (e.g. NS=Monolingual native speakers, L2=second language learners of Chinese), the proficiency level (e.g. Int=Intermediate, Ad=Advanced), the participant number (e.g. 01=participant No. 1 in a given group), and the sequence of clips in which the participant viewed (A or B).

110

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

Fig. 2. Mean proportion of different syntactic strategies used by English learners of Chinese at three proficiency levels.

An overview of our findings suggested, first of all, that the performance of L2 learners differed from that of Chinese natives in several aspects. As regards the ‘Tight’ pattern, a Kruskal--Wallis test was performed on the non-normally distributed data and revealed that there was a significant difference among L1 and L2 learners in the frequency of employing the ‘Tight’ pattern (x2 (3, 44)=9.158, p=.027). A Tukey post hoc test was used to follow up this finding and revealed that learners of low proficiency used the ‘Tight’ pattern significantly less frequently than did the native speakers (L2_LowTight>Loose>Semi-tight, with neither the ‘Tight’ nor the ‘Semitight’ type being their favoured option (cf. Table 3). Generally, our results indicated that, as compared to the acquisition of lexicalisation pattern of motion components reported in previous studies (e.g. Cadierno, 2008), syntactic strategies of distributing multiple pieces of information over an utterance had to be acquired over time; moreover, developmental shifts were only visible when learners had progressed to an advanced level and they did not achieve the statistical significance.

Table 3 Mean frequency of target utterances in which different syntactic strategies were used by L2 learners of Chinese versus monolingual native speakers of Chinese and English. Group ID

Tight

Semi-tight

Loose

Atypical

L2_Low L2_Int L2_Ad CH_NS EN_NS

.23 .31 .45 .54 .98

.06 .07 .07 .45 .01

.25 .19 .09 .01 .01

.46 .43 .39 .00 .00

(.40) (.29) (.21) (.18) (.40)

(.22) (.11) (.12) (.18) (.04)

(.29) (.17) (.11) (.02) (.02)

(.40) (.33) (.24) (.00) (.00)

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

111

5.3. The use of the ‘Loose’ pattern by learner groups First of all, we predicted that the ‘Loose’ packaging would not be the favoured packaging option among all L2 learners and would only be used occasionally. We based our prediction on the specific properties of our stimuli, which had been designed to present Manner (coupled with Cause) and Path as equally salient and simultaneously occurring. According to Allen et al. (2007), speakers may pursue cohesion between cognitive and linguistic representation of an event: elements that are perceived and conceptualised as simultaneously occurring are more likely to be encoded in linguistic structures as close to each other, for example, within the boundary of a single clause, rather than across different clauses (43). Given that our participants were all cognitively mature speakers who should have developed a good understanding of the temporal simultaneity inherent in the stimuli, we expected them to use the ‘Loose’ packaging rarely, particularly at the advanced stage. The results did not conform to our prediction. Low and intermediate learners opted for this pattern relatively frequently (20%); even advanced learners used it from time to time (9%). Learners across proficiencies seemed to produce ‘Loose’ constructions of different forms. Specifically, learners at the low proficiency typically separated Manner (and Cause) from Path between two clauses in coordination (Example 20). (20)

他 把 木柴 推 Ta1 ba3 mu4chai2 tui1 he BA log push ‘He started pushing the logs

起来, 从 qi3lai2 cong2 rise from and went from

那 边 到 火。 [L2_Low_01A] na4 bian1 dao4 huo3 that side arrive fire that side to the fire.’

In contrast, more advanced learners produced ‘Loose’ structures in which the first coordinated clause normally expressed C+M, with the second clause providing extra information regarding the protagonist’s Manner of motion (walking), apart from Path. See Example 21 for an illustration. (21)

他 拉 小车, 后来 他 从 右 Ta1 la1 xiao3che1 hou4lai2 ta1 cong2 you4 he pull car then he from right ‘He was pulling the small car, then he walked from

边 走路 bian1 zou3lu4 side walk the right side to

到 dao4 arrive the left

小湖 xiao3hu2 lake side of the

的 左 边。 de zuo3 bian1 ASSOC left side lake.’ [L2_Int_06A]

One phenomenon regarding the ‘Loose’ pattern warrants a mention: there were sporadic instances of using this pattern to encode the temporal overlapping inherent in the stimuli. Specifically, conjunctive morphemes yibian. . .yibian (‘while. . .while’) were used as a linking device between two separate clauses marking the temporal simultaneity between relevant events, i.e. the participants used lexical means to replace syntactic means (i.e. ZHE constructions typical of the ‘Semi-tight’ packaging) in order to encode the temporal relationship, as illustrated, below. (22)

他 又... 一边 推 小车, 一边 散步 过 冰 湖。 [L2_Int_02A] Ta1 you4 yi4bian1 tui1 xiao3che1 yi4bian1 san4bu4 guo4 bing1 hu2 he then while push car while saunter cross icy lake ‘He then sauntered across the icy lake while he was pushing the car.’

5.4. Formal features of the syntactic pattern under the label of ‘Atypical Chinese’ Our results revealed, unexpectedly, that learners across all proficiency levels employed some syntactic means that had been labelled in our coding as ‘Pattern in atypical Chinese’. A closer look at the data showed that such structures bear some resemblance to Chinese serial verb (SV) constructions. The mean frequency of this SV pattern remained constant at around 40% across learners’ proficiencies, and it even constituted the most frequently used syntactic strategy among learners at low and intermediate levels. Such SV patterns can be formally represented as follows: Agent NP V1 Figure NP V2 Ground NP (23a) 邦尼 推 了 一 桶 绕 了 桌子 一 圈。 Bang1ni2 tui1 le yi1 tong3 rao4 le zhuo1zi yi1 quan1 Benny push ASP one barrel surround ASP table one circle ‘Benny pushed a barrel around the table in a circle.’ [L2_Ad_02A] (23b) 他 推 一 袋 东西 去 了 楼梯 的 旁边。 Ta1 tui1 yi2 dai4 dong1xi1 qu4 le lou2ti1 de pang2bian1 he push one CL goods go ASP escalator ASSOC side ‘He pushed a bag of goods to the side of the escalator.’ [L2_Int_09A]

112

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

他 拉 游泳 圈 下 了 沙 丘。 Ta1 la1 you2yong3 quan1 xia4 le sha1 qiu1 he pull swimming ring descend ASP sand dune ‘He pulled the swimming ring down the sand dune.’ [L2_Low_03B] The above SV construction involved two verbs in concatenation. It was thus sandwiched between the ‘Tight’ and the ‘Semi-tight’ type: it was less tight than the former because it involved two verbs rather than one, but tighter than the latter because the two verbs were juxtaposed rather than distributed across clauses via subordination. Two formal features of such SV constructions are worth mentioning: (23c)

(a) (b)

Two VPs were conjoined without any overt marker indicating the syntactic as well as semantic relationship between them. The intonation pattern of the SV structure was continuous and downward, with no phonological pause in between. This suggested that it is a single grammatical unit rather than two (i.e. versus coordinates).

6. Discussion Two specific findings are of particular interest and will be discussed further. Firstly, as regards the ‘Loose’ packaging of motion information, advanced learners seemed to use this pattern to maximise their utterance density. Learners of low proficiency, however, seemed to use it for a different reason. As earlier pointed out, one of the challenges of the current task was that complex semantic information needed to be communicated. To mitigate the pressure of mapping, linearising and sequencing a set of particularly dense CM components (no fewer than five) into one syntactic unit, learners of low proficiency had chosen to express varied CM components in a relatively discursive way (see Hendriks et al., 2008 for a similar result regarding French and English L2 learners). In a similar fashion, because of the complexity of subordination associated with the syntactic devices to encode caused motion (i.e. ZHE constructions), learners opted for lexical means to achieve the communicative goal. The ‘while. . .while’ responses reflected their great flexibility when packaging semantically dense information; however, they were targetdeviant and not used by the Chinese natives in our sample. Secondly, it is observed that the SV packaging remained constant across different Path types (Examples 23a--c) as well as learner groups. Questions thus arise as to how to interpret this syntactic pattern. In principle, this phenomenon can be interpreted in a number of ways. First of all, it is not impossible that learners deliberately used the typical SV structure in the target language (Example 24) to encode the CM events. (24)

Typical SV construction in Chinese: a hypothetical example 他 开 门 出去 了。 Ta1 kai1 men2 chu1-qu4 le he open door exit go ASP ‘He opened the door and went out.’

Note, however, that the canonical relationship between the two verbs in a typical SV sentence in Chinese is temporal consecutiveness: the event headed by V1 occurred before the event headed by V2 (‘opening the door’ preceded ‘going out’ in Example 24). As adult learners, our participants should be aware of the temporal simultaneity inherent in the stimuli and should therefore avoid rendering the motion events in a temporal hierarchy. This was strongly supported by the fact that participants employed coordinated clauses with the conjunctive morpheme yibian ‘while’ to highlight the concurrence of the two events, as discussed in the previous section. Further, it might be the case that learners relied heavily on the source pattern to package the dense CM information and their performance mirrored the English ‘Manner-and-Cause Verb+Path Satellite’ combination, as illustrated in Example 25. (25a) (25b)

Benny Benny Benny Benny ‘Benny

pulled a barrel of hay up the ladder. 推 草 上 梯子。 tui1 cao3 shang4 ti1zi push hay ascend ladder pushed the hay and went up the ladder.’ [L2_Int_02A]

It needs to be pointed out that if learners entirely mirrored their source pattern in packaging, they should have used the Chinese Path morpheme shang as a particle (‘up’) rather than an individual verb (‘ascend’). However, a closer look at our data revealed that V2 in the SV construction was presented as a verb in most cases: participants tended to affix the

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

113

perfective aspectual marker le to it to signify the ending of an event, a phenomenon frequently observed across learner groups at different proficiencies, as shown in Examples (23a)--(23c) (rao le ‘to surround’, qu le ‘to go away from the speaker’, xia le ‘to descend’). It is important to note another possibility: learners actually intended to use the ZHE constructions typical of the ‘Semitight’ pattern to package information, However, facing the great challenge of expressing multiple components with limited linguistic devices at their disposal, they produced a ‘weakened’ or ‘reduced’ form of ZHE constructions in which the durative marker zhe was omitted to avoid the processing load of designating the syntactic relationship between the two verbs in this construction. In order to produce target-like ZHE constructions, learners needed to understand that the verb that was realised as subordinated should be morphologically simpler (i.e. normally monosyllabic), semantically more salient (i.e. focus of the utterance) and temporally more atelic (point-like). This seemingly presented a heavy challenge to them, as evident in the reversed sequence of V1 and V2 in the following example with low acceptability. (26)

他 Ta1 he ‘He,

走 着 拉 玩具 zou3 zhe la1 wan2ju4 walk DUR pull toy while walking, pulled the toy

汽车 过 冰 上。 qi4che1 guo4 bing1 shang4 car cross ice on car across the ice.’ [L2_Low_06A]

Learners tended to remedy this problem by avoiding the fixation of the durative marker zhe to either of the two verbs, thus leaving the syntactic as well as semantic and temporal relationship between them under the interpretation of the listener. This produced a set of barely acceptable, yet not strictly ungrammatical, utterances, which sufficiently conveyed the complex information in a communicative task though sounded target-deviant. Note, however, that it seems equally reasonable to assume that L2 learners had no intention to suffix V1 in the SV construction with the subordinating zhe marker; instead, they chose to follow their L1 feature in foregrounding Manner and Cause in the main verb rather than backgrounding such information in the periphery of an utterance through subordination, which is sometimes required by the L2 syntax (i.e. the ZHE construction). That is, V1 in the SV construction combined Manner and Cause, and thus reflected the satellite-framing properties of L1 English. Meanwhile, V2 in this construction was largely used as a full verb and encoded Path of motion, thus representing the verb-framing properties that L2 Chinese sometimes demonstrated. Seen in this way, the L2 learners did not construct an inter-language free of typological constraints imposed by L1 and/or L2. Rather, both the source and the target language had an impact on the acquisition process. 7. Conclusion This study investigated whether and how English learners of Chinese at three proficiency levels acquired the target pattern of expressing complex motion events at an arguably advanced linguistic level of syntax. We asked two specific questions: (a) whether and how learners’ preferences in motion syntax resembled or differed from that of native speakers; (b) whether and how learners’ syntactic preferences developed with proficiency. Our main findings can be summarised as follows: a. Although the ‘Tight’ pattern constituted the predominating mode of information packaging in the source language (English), learners did not overwhelmingly use this pattern but frequently opted for other syntactic means to encode the CM event. b. Learners across proficiencies did not acquire the target system of packaging complex CM information in ‘Tight’ and ‘Semi-tight’ patterns with comparable frequencies. Even advanced learners only used the latter pattern occasionally (around 7%) probably due to its structural complexity. c. The ‘Loose’ pattern of information distribution was used by all learners in a frequency higher than was expected (10--20%), mainly due to participants’ efforts to be maximally explicit without incurring a heavy processing load in syntax. d. No development occurred between low and intermediate levels. Developmental tendencies were only observed as learners progressed to an advanced level. This suggested that it took a longer time to adapt to the advanced linguistic skills of syntactic organisation as compared to the speed of adapting to the lexicalisation pattern as revealed in previous studies (e.g. Cadierno, 2008). e. Learners across all proficiency levels opted for the SV construction to organise complex CM information at a mean frequency of 40%. This construction differed from the canonical SV structure in Chinese in that it encoded two events not linked by temporal consecutiveness (but temporal simultaneity), and can be interpreted in different ways. These results were not entirely in line with either of our two hypotheses. Finding (a) suggested that, contrary to our ‘language-specificity’ prediction, learners at the initial and intermediate stages of acquisition were not completely trapped

114

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

in their L1 in the sense that they did not exactly mirror the source pattern and predominantly used the Tight way (the BA construction) for dense information packaging. Meanwhile, contradicting our ‘universal motion conceptualisation’ hypothesis, the syntactic reorganisation of motion information in an L2 seemed less straightforward than predicted in the sense that the complex target pattern (i.e. the Semi-tight ZHE construction) was not fully acquired by L2 learners even at the advanced stage, as suggested by findings (b), (c) and (e). Overall, our results seem to add more complexity to the already inconsistent picture presenting mixed results regarding the role of L1 in the L2 acquisition of motion expressions. This indicates, among other things, that language-specific influence is a complex phenomenon, which needs to be considered in conjunction with variations along a number of important dimensions, such as the type of experimental stimuli (static pictures versus dynamic video clips), the level of investigation (lexicalisation, semantics--syntax interface), the degree of similarity or difference between L1 and L2, the proficiency levels of participants (low-intermediate versus advanced) and the nature of motion events investigated (spontaneous versus caused; telic versus non-telic). Let us now return to the question of whether and to what extent the specific properties of L1 influence L2 acquisition. Particular attention will be directed to the unexpected syntactic sub-strategy of SV pattern among learners. If we temporarily take this pattern as a ‘reduced’ form of the ZHE construction, then it might be claimed that participants’ performance exhibited more resemblance to the target pattern than to traces of the L1 influence. Further, the effect of positive transfer did not arise in cases where the source pattern of information packaging bore partial resemblance to the target patterns (i.e. both English and Chinese used the ‘Tight’ pattern) and the acquisition of target constructions typical of the ‘Tight’ pattern did not constitute particular problems for learners (BA). By this it is meant that learners were capable of largely overcoming their L1 pattern in L2 acquisition. Viewed in this light, our findings provided little support for Slobin’s ‘thinking for speaking’ hypothesis: in circumstances presented by the current study, the L1 ‘thinking for speaking’ pattern seemed not as resistant to remoulding as earlier predicted. Further along this line, our findings may shed some fresh light on the perplexing question of the relation between language and thought in general. As mentioned in Introduction, an increasing number of studies on L1 acquisition of spatial systems have demonstrated that typological properties of L1 constrained the way children linguistically expressed, categorised, or even conceptualised motion events, thus at least partially reviving the Whorfian hypothesis. If we assumed that in the SV construction, L2 learners actually followed their L1 framing pattern in conflating Manner and Cause in the main verb rather than downplaying such information in subordination (which was frequently the case in L2), then it can be argued that the ‘thinking for speaking’ pattern was not indeed that flexible and the L1 influence was relatively strong in L2 acquisition of syntax. Seen in this way, our findings may provide some evidence for a moderate version of linguistic relativity. That is, typology may have an impact on online motion description and conceptualisation. In the particular context of L2 acquisition, the typological properties of L1 influenced how L2 learners weighed the salience of varied types of motion information and further organised them across an utterance via syntactic devices. Acknowledgements We would like to thank the Award of the British Academy Postdoctoral Fellowship, which was awarded to the first author in 2010, for its generous financial support of this study (pf100022). We would also like to thank the Award of Shenzhen University Distinguished Professorship, which provided support in various forms to the first author, in finalising the current study (contract no.: TP026). Appendices Description of sixteen CM stimuli (sequence A) A: Warm-up item: Benny pulled a boat out of the lake. A1. Benny pushed a swimming ring down the sand dune. A2. Benny pulled a treasure bag into the pyramid. A3. Benny pushed a bundle of logs away from the campfire. A4. Benny pulled a big gift box along the tunnel. A5. Benny pushed a treasure bag up the pyramid. A6. Benny pulled a toy car across the icy lake. A7. Benny pushed a bundle of logs towards the campfire. A8. Benny pulled a toy car around the icy lake. A9. Benny slid a heavy bag towards the escalator. A10. Benny rolled a barrel of beer around the round table. A11. Benny slid a toy car across the icy lake.

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

A12. A13. A14. A15. A16.

Benny Benny Benny Benny Benny

115

rolled a barrel of hay up the ladder. rolled a basketball along a row of chairs. slid a suitcase away from the tent. rolled a golf ball into the puddle. rolled a swimming ring down the hill.

References Allen, S., Ozyurek, A., Kita, S., Brown, A., Furman, R., Ishizuka, T., et al., 2007. Language-specific and universal influences in children’s syntactic packaging of Manner and Path: a comparison of English, Japanese, and Turkish. Cognition 102 (1), 16--48. Berman, R., Slobin, D.I., 1994. Relating events in narrative: a crosslinguistic developmental study. Lawrence Erlbaum Associates, Hillsdale, NJ. Bowerman, M., 1996. The origins of children’s spatial semantic categories: cognitive versus linguistic determinants. In: Gumperz, J., Levinson, S. (Eds.), Rethinking Linguistic Relativity. Cambridge University Press, Cambridge, pp. 145--176. Bowerman, M., 1999. Learning how to structure space for language: a crosslinguistic perspective. In: Bloom, P., Peterson, M., Nadel, L., Garrett, M. (Eds.), Language and Space. MIT Press, Cambridge, MA, pp. 385--436. Bowerman, M., Choi, S., 2001. Shaping meanings for language: universal and language-specific in the acquisition of spatial semantic categories. In: Bowerman, M., Levinson, S. (Eds.), Language Acquisition and Conceptual Development. Cambridge University Press, Cambridge, pp. 475--511. Bowerman, M., Choi, S., 2003. Space under construction: language-specific categorization in first language acquisition. In: Gentner, D., GoldinMeadow, S. (Eds.), Language in Mind: Advances in the Study of Language and Thought. MIT Press, pp. 387--427. Cadierno, T., 2004. Expressing motion events in a second language: a cognitive typological perspective. In: Achard, M., Niemeier, S. (Eds.), Cognitive Linguistics, Second Language Acquisition, and Foreign Language Teaching. Mouton de Gruyter, pp. 13--49. Cadierno, T., 2008. Learning to talk about motion in a foreign language. In: Robinson, P., Ellis, N.C. (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition. Routledge, pp. 239--275. Cadierno, T., Ruiz, L., 2006. Motion events in Spanish L2 acquisition. Annual Review of Cognitive Linguistics 4, 183--216. Chen, L., Guo, J., 2009. Motion events in Chinese novels: evidence for an equivalently-framed language. Journal of Pragmatics 41, 1749--1766. Chu, C., 1987. The semantics, syntax, and pragmatics of the verbal suffix -zhe. Journal of Chinese Language Teachers Association 22 (1), 1--41. Chu, C.Z., 2009. Path of motion: conceptual structure and representation in Chinese. In: Xing, Z. (Ed.), Studies of Chinese Linguistics: Functional Approaches. Hong Kong University Press, Hong Kong, pp. 65--73. Du, H., 2006. The Acquisition of the Chinese ba-Construction. LIMCOM EUROPA, Muenchen. Gao, H., 2001. Notions of motion and contact for physical contact verbs. In: Holmer, A., Svantesson, J., Viberg, A. (Eds.), Proceedings of the 18th Scandinavian Conference of Linguistics, vol. 2. University of Lund, Department of Linguistics, Lund, pp. 193--209. Gennari, S., Sloman, S., Malt, B., Fitch, W., 2002. Motion events in language and cognition. Cognition 83, 49--79. Hendriks, H., Hickmann, M., 2011. Space in second language acquisition. In: Cook, V., Bassetti, B. (Eds.), Language and Bilingual Cognition. Psychology Press, pp. 315--339. Hendriks, H., Hickmann, M., Demagny, A.C., 2008. How English native speakers learn to express caused motion in English and French. Acquisition et Interaction en Langue Étrangère 27, 15--41. Hickmann, M., 2006. The relativity of motion in first language acquisition. In: Hickmann, M., Robert, S. (Eds.), Space in Languages: Linguistic Systems and Cognitive Categories. John Benjamins, Amsterdam/Philadelphia, pp. 281--308. Hickmann, M., Hendriks, H., Champaud, C., 2009. Typological constraints on motion in French and English child language. In: Guo, J., et al. (Eds.), Crosslinguistic Approaches to the Psychology of Language: Research in the Tradition of Dan Isaac Slobin. Psychology Press, pp. 209--224. Hohenstein, J., 2005. Language-related motion event similarities in English- and Spanish-speaking children. Journal of Cognition and Development 6, 403--425. Hohenstein, J., Naigles, L., Eisenberg, A., 2004. Keeping verb acquisition in motion: a comparison of English and Spanish. In: Hall, G., Waxman, S. (Eds.), Weaving a Lexicon. MIT Press, pp. 569--602. Hohenstein, J., Eisenberg, A., Naigles, L., 2006. Is he floating across or crossing afloat? Cross-influence of L1 and L2 in Spanish--English bilingual adults. Bilingualism: Language and Cognition 9, 249--261. Ji, Y., Hendriks, H., Hickmann, M., 2011a. Children’s expression of voluntary motion events in English and Chinese. Journal of Foreign Languages 34 (4), 2--20.

116

Y. Ji, J. Hohenstein / Lingua 140 (2014) 100--116

Ji, Y., Hendriks, H., Hickmann, M., 2011b. How children express caused motion events in Chinese and English: universal and language-specific influences. Lingua 121, 1796--1819. Ji, Y., Hendriks, H., Hickmann, M., 2011c. The expression of caused motion events in Chinese and in English: some typological issues. Linguistics 49 (5), 1041--1076. Levinson, S.C., 2003. Space in Language and Cognition: Explorations in Cognitive Diversity. Cambridge University Press. Li, C.N., Thompson, S.A., 1976. The meaning and structure of complex sentences with -zhe in Mandarin Chinese. Journal of the American Oriental Society 96 (4), 512--519. Li, C.N., Thompson, S.A., 1981. Mandarin Chinese: A Functional Reference Grammar. University of California Press. Lucy, J., 1992. Grammatical Categories and Cognition. Cambridge University Press, Cambridge. Malt, B., Sloman, S., Gennari, S., 2003. Universality and language specificity in object naming. Journal of Memory and Language 49 (1), 20--42. Mandy, M., et al., 2010. A developmental shift from similar to language-specific strategies in verb acquisition: a comparison of English, Spanish, and Japanese. Cognition 114 (3), 299--319. McDonald, E., 1995. Completive verb compounds in modern Chinese: a new look at an old problem. Journal of Chinese Linguistics 22, 317--362. Naigles, L., Terrazas, P., 1998. Motion verb generalisations in English and Spanish: influences of language and syntax. Psychological Science 9, 363--369. Navarro, S., Nicoladis, E., 2005. Describing motion events in adult L2 Spanish narratives. In: Eddington, D. (Ed.), Selected Proceedings of the 6th Conference on the Acquisition of Spanish and Portuguese as First and Second Languages. Cascadilla Proceedings Project, pp. 102--107. Papafragou, A., Selimis, S., 2010. Event categorisation and language: a crosslinguistic study of motion. Language and Cognitive Processes 25, 224--260. Slobin, D.I., 1996. From thought and language to thinking to speaking. In: Gumperz, J., Levinson, S.C. (Eds.), Rethinking Linguistic Relativity. Cambridge University Press, pp. 70--96. Slobin, D.I., 2004. The many ways to search for a frog: linguistic typology and the expression of motion events. In: Strömqvist, S., Verhoeven, L. (Eds.), Relating Events in Narratives, Typological and Contextual Perspectives, vol. 2. Lawrence Erlbaum Associates, pp. 219--257. Talmy, L., 1985. Lexicalization patterns: semantic structure in lexical form. In: Shopen, T. (Ed.), Language Typology and Syntactic Description, vol. 3. Cambridge University Press, pp. 36--149. Talmy, L., 2000. Toward a Cognitive Semantics, Vol. 2: Typology and Process in Concept Structuring. MIT Press, Cambridge, MA. Talmy, L., 2009. Main verb properties and equipollent framing. In: Guo, J., et al. (Eds.), Crosslinguistic Approaches to the Psychology of Language: Research in the Tradition of Dan Isaac Slobin. Psychology Press, pp. 389--402. Xing, J., 2002. On the Errors of Chinese Multiple Sentence by Foreign Students. , (MA thesis)Jinan University. Zeng, Y.H., 2011. Acquisition of English Motion Event Expressions by Chinese EFL Learners. (unpublished doctoral dissertation) Hunan Normal University. Zou, K., 1993. The syntax of the Chinese BA construction. Linguistics 31, 715--736.