2.2.2 Language-specificity in children's acquisition of spatial expression
Recently a growing number of scholars have questioned the hypothesis that grammatical forms are mapped fairly directly onto cognitive spatial concepts in the acquisition of spatial expressions, and have instead emphasized that the influence from the specific language input should not be neglected.In fact, the impact of specific language on the acquisition of spatial terms is already evidenced in some of the earliest comparative studies （Johnston and Slobin 1979, Hickmann 2006）.Although children from different languages follow a similar developmental sequence in learning spatial words, the exact rate at which they acquire them seems to differ “due to various factors that are truly language-specific such as the preposition versus postposition, the morphological complexity and the lexical diversity of the spatial markers” （Hickmann 2006: 285）.This observation suggests that the mapping process from surface forms to cognitive concepts is far less direct than predicted by the universal approach and that the specific properties of a language have an obvious role to play.
The impact of language-specific factors on the acquisition of spatial language and the organization of spatial information is actually found in diverse sub-domains of space such as frameworks of spatial reference, static spatial configurations, expressions of translocational motion events and overall narrative style of a discourse.Some scholars hold that the influence of language is limited to the level of linguistic expression only （e.g.Gennari, S.P.et al.2002, Malt, B.et al.2003, Ochsenbauer et al 2007, Papafragou et al.2002, Slobin 1996a）, while others argue that language-specific properties can even shape our spatial cognition at a deeper level （see, for instance, Bowerman 1996, 1999, Bowerman and Choi 2001, Levinson 2003, Lucy 1993）.Specifically, Levinson （2003） notes how spatial thinking can co-vary with languages: there are robust correlations between frames of reference used in language （e.g.intrinsic, relative or absolute） and frames of reference used in non-linguistic memory and reasoning, suggesting a major Whorfian effect of languages on cognition.In a similar fashion, Bowerman （1996） notices how quickly and easily children can adopt a language-specific principle of semantic categorization in their input language; and this may even begin to develop before production starts.A distinction is made in English, for example, between inand onsituations （i.e.the spatial category of containment versus support） while Korean distinguishes the containment of loose-fit from that of tight-fit.In preferential-looking tasks for comprehension, it is observed that English children before the age 2 look longer at scenes showing containment regardless of whether the object is loosely or tightly fit, whereas their Korean counterparts stare longer at scenes showing tight fitting objects, regardless of whether this fit involves containment or surface contact （Bowerman 1996: 161-168）.
As regards the expression of motion events, it is revealed in a series of studies by Bowerman and Choi （Bowerman 1996, 1999, Bowerman and Choi 2001, Choi and Bowerman 1991） that there exists a pervasive interaction between non-linguistic conceptual development and the relevant semantic category of an input language.English Path markers, typically prepositions, do not reflect whether a motion is presented as spontaneous or caused, while Korean marks most Path meanings with distinct verb roots in transitive and intransitive clauses.Accordingly, English children use the prepositions upand downfor a great range of spontaneous and caused motions along the vertical axis, including climbing on and off furniture, posture changes, raising and lowering things and wanting to be picked up or put down, which suggests that they possess a core spatial concept of vertical directionality.By contrast, Korean children consistently distinguish different kinds of vertical motion and use strictly different verbs （transitive versus intransitive） for spontaneous and caused motion along a path.Choi and Bowerman （1991） furthermore find that, contrary to the view that language specificity emerges only gradually with divergence from a shared starting point （e.g.Slobin 1985）, the striking differences observed in the expression of spontaneous motion between English and Korean children seem to be present from the very beginning, as predicted by Gentner （1982）.
In Hickmann's （2003） study comparing picture-elicited narratives by children in four languages （English, French, German and Chinese）, it is revealed that there are cross-linguistic differences in how speakers represent motion from the age of 4 onwards.In French, early predicates are the least diverse and mainly encode a single piece of information, namely, either Path with changes of location （e.g.to ascend , to leave ） or Manner plus a general location for motion （e.g.to fly in the sky ）.By contrast, early predicates in the other three languages are much more varied and semantically dense, normally encoding both Manner and Path information （e.g.to fly away , to jump across,to chase away ）.These results were replicated in a study of child （ages 4 to 12） and adult speakers of English and Greek （i.e.a satellite-framed versus a verb-framed language） on the basis of their narratives of still pictures from the Frog story （Papafragou et al.2002）.
It is also found that language-specific influences go beyond the semantic level and can be detected at the semantics-syntax interface.Allen et al.（2007） point out that semantic components for motion can be syntactically packaged in different ways across languages.Specifically, in satellite-framed languages like English, adult speakers usually encode Manner and Path within one single clause, using tight syntactic packaging （e.g.The man hops all the way up the hill .）.In contrast, in verb-framed languages such as Turkish and Japanese, Manner and Path are typically expressed through the use of the less syntactically tight matrix-subordinate construction; that is, they use semi-tight syntactic packaging （e.g.The man ascends the hill hopping all the way ）.Alternatively, speakers of such languages sometimes use a bi-clausal structure, which represents loose syntactic packaging （e.g.The man goes up the hill,and he's hopping all the way ）.Allen et al.examine the syntactic packaging of motion events in the above-referenced three languages and find that “children's （3 to 4 years） acquisition of packaging semantic elements into syntactic units is clearly guided by language-specific syntax-semantic mappings” （Allen et al.ibid.: 45）.More strikingly, they find that young English children actually follow adults in choosing the appropriate syntactic construction for expressing a motion event, depending on the specific relationship between Manner and Path in the utterance.To illustrate, in English, Manner and Path are usually expressed in one clause if the former is causally related to the latter （i.e.the tight packaging for Manner-Inherent events: The ball jumps down the stairs ）, whereas in events in which Manner does not cause changes of location, Manner and Path are more likely to be expressed in two verbal clauses instead of one （i.e.the semi-tight and the loose packaging for Manner-Incidental events: The ball jumps while falling down from the stairs ）.Allen et al.find that “English children from three years on are particularly sensitive to different event types and largely follow the adult pattern in using tight syntactic constructions significantly more frequently to encode Manner-Inherent events versus Manner-Incidental events” （ibid.: 39-42）.
On the discourse level, Slobin （2004） has looked at children's expression of spatial information in 11 languages from different language families （i.e.Spanish, French, Turkish, Italian, Hebrew, Dutch, German, English, Mandarin, Thai and Russian） and has studied its acquisition.His findings suggest that differences in the encoding of Path are likely to entail differences in the encoding of other spatial elements such as Manner and Cause, Figure and Ground.He discovers that the locus of Path information does indeed have a clear influence on the overall distribution of spatial information in discourse in that information may be encoded by different linguistic means, or not encoded at all.His conclusion is that although all canbe encoded, not all issystematically encoded.It is the ‘ready availability’ of a given linguistic device （i.e.the ease of expression） that will ultimately influence what speakers of a given language tend to select for expression and how speakers acquire that language （Hendriks et al.2008）.
Slobin thus finds clear differences between verb-framed and satellite-framed languages in terms of the type of information encoded in the discourse, the linguistic means used for this encoding and the resulting overall rhetorical styleof a narrative discourse （2004: 219-240）.In satellite-framed languages, for example, since Path is systematically encoded and can even be stacked in verb particles and prepositions, the main verb of a clause is readily available to express various types of Manner and/or Cause information.Therefore, both Manner and Path information should be typically selected for motion expressions.By contrast, in verb-framed languages, since Path is systematically encoded in the main verb, adding Manner information in a clause becomes an additional cognitive cost.Therefore, it is very likely that Path information alone will be expressed, with Manner information typically omitted, or if semantically salient, expressed in the periphery of a clause via gerunds or adverbials.These assumptions are substantiated by Hickmann （2006） and Hickmann et al.（2009） who examine the description of voluntary motion events in French and English children （ages 4 and 6） as well as adults and report that typological properties （i.e.verb-framed versus satellite-framed） affect the semantic density of children's utterances.Regardless of age, English children tend to express denser information than their French counterparts resulting from the availability in English of easily accessible Verb + Satellite constructions.
Concerning the language-specific impact on the rhetorical styleof a discourse as a whole, Berman and Slobin （1994） focus on the development of the narrative style of children learning verb-framed （Spanish, Hebrew, Turkish） as opposed to satellite-framed languages （English, German） and reveal that children from the age of three show a rhetorical style that is typical of adult speakers of their own language.Discourses from satellite-framed languages are characterized by a rich array of Manner verbs used in conjunction with Path phrases （e.g.fall off the tree , jump out of the jar ）.By contrast, speakers of verb-framed languages rarely mention Manner of motion and tend to leave Path details to be inferred by only indicating the successive static locations of the protagonist.It seems that the specific language ‘filters’ the incoming information and channels children's attention to different aspects of motion events （Manner and/or Path）.As Berman and Slobin （1994; cited in Bowerman 1994: 44） put it:
We began the study with an expectation that there was a basic set of semantic notions that all children would try to express...［But］ we were repeatedly surprised to discover how closely learners stick to the set of distinctions that they have been given by their language... We are left then with a new respect for the powerful role of each individual language in shaping its own world of expression... （ibid.: 641）.
These findings are further confirmed in subsequent studies by Slobin （1996b, 2000）, who compares English （satellite-framed） with Spanish （verb-framed） in different types of tasks, including picture-elicited narratives, mental images associated with narratives, literary texts and translations.His main results are summarized as follows: English has an elaborated set of verbs making fine distinctions among various manners of motion, whereas such verbs are absent in Spanish; English speakers across ages give Manner and Path details regarding motion much more frequently than their Spanish counterparts; and Spanish translation of English motion events tends to omit Manner information or substitutes static locations for full path-ground trajectory.As a result, the narrative style of a Spanish discourse tends to be relatively ‘static’ with its emphasis on static locations of the protagonist and physical settings of the story, whereas the rhetorical style of an English discourse is more ‘dynamic’ as it specifies details about Path coupled with Manner of motion.
Though an increasing number of studies have confirmed the contribution of specific language input to the acquisition of spatial expression, very few of them have hypothesized about how this influence is specifically manifested in acquisition.Slobin's proposal of learning to ‘think for speaking’ （1996a）, which is usually taken as a weak version of the Whorfian hypothesis, sheds fresh light on this issue.According to him:
Experiences are filtered through language into verbalized events ［...which are］ constructed on-line in the process of speaking...The activity of thinking takes on a particular quality when it is employed in the activity of speaking.In the evanescent time frame of constructing utterances in discourse, one fits one's thoughts into available linguistic frames.Thinking for speaking involves picking those characteristics of objects and events that （a） fit some conceptualization of the event, and （b） are readily encodable in the language...In acquiring a native language, the child learned particular ways of thinking for speaking （ibid.: 75-76）.
In this sense,language directs its speakers to attend, while speaking, to dimensions of experience that are enshrined in grammatical categories （ibid.: 71）.For example, in English, the most important grammatical category, the verb, encodes Manner information, which becomes the target of habitual attention.As a result, Manner information becomes more readily available and gets more frequently expressed.This high codability of manner information in English in turn directs the speaker's attention to this particular piece of semantic information, leading finally to ‘Manner salience’ on the discourse level.
Note that some scholars （e.g.Allen et al.2007, Choi and Bowerman 1991） argue that it is not the case that in acquiring spatial language either the universal cognitive force or the language-specific influence takes hold exclusively; rather, both forces are most likely present and contribute in differing degree to the acquisition of expressing motion events.Based on their findings from studies of English and Korean children, Choi and Bowerman （1991） suggest that young learners of both languages may have the same pre-linguistic potentialfor identifying Path as an independent component of motion events.However, they are selectively prompted by the structure of the input to develop this potential.Specifically, they state:
Children learning English are systematically shown how to isolate a few recurring kinds of Path, and they learn how to do this quickly.Children learning Korean, in contrast, meet Path mostly conflated with notions of spontaneous or caused motion and often with specific properties of the Figure and Ground as well, so it takes them longer to realize that Path can sometimes be extracted and receives its own marking （ibid.: 17）.
As is evident from the above literature review, the investigation of children's acquisition of motion expressions exclusively focuses on languages which are treated as either clearly satellite-framed or unequivocally verb-framed.There are very few studies which have looked into the acquisition issue in the equipollently-framed languages.In addition, the existing literature is predominantly concerned with western languages （especially English）; only two Asian languages are frequently written about （i.e.Korean and Japanese）, both of which are typically verb-framed and do not possess the typological complexity found in Chinese.In this light, it is particularly meaningful to explore how children acquiring equipollently-framed languages learn to express semantic components for motion: are they mainly guided by cognitively universal spatial concepts, language-specific categories and distinctions, or both, with their respective impact made more prominent in a given type of task （i.e.voluntary motion versus caused motion） or event （i.e.UP, DOWN, ACROSS or INTO）.
[1] Though most scholars consider French as clearly verb-framed, Pourcel and Kopecka （Unp.Ms.） point out that a binary typology is far from sufficient as a reliable index of motion lexicalization in French.Apart from the frequent use of Path verbs, there is also a hybrid pattern in French whereby both Path and Manner are encoded in the main verb （e.g.grimper ‘climb up’）, and a ‘reverse verb-framing’ pattern in which the Manner is actually encoded in the verb root, with Path and Ground conflated in optional constituents.On the basis of data such as these, Pourcel and Kopecka therefore suggest that the use of a particular pattern to express motion events is not reliably predictable of a language's typological status unless semantic and pragmatic factors are integrated into the morphosyntactic criteria.
[2] Slobin is generally interested in the overall language-specific rhetorical styleof discourse, not just the lexicalization pattern of a motion event; he thus further proposes that instead of being categorical, languages should be ranked according to the amount of attention they pay to the expression of Manner.That is, they can be ordered along a ‘cline of manner salience’.The high manner-salient languages are those in which there is an accessible device for encoding Manner,such as the main verb in satellite-framed languages and the Manner verb in serial verb languages.By contrast, the low manner-salientlanguages refer to those in which Manner is subordinated to Path （2004: 26）.
[3] Talmy's definition of the ‘verb’ or ‘verb stem/verb root’ is also unclear: it does not follow either a strictly morphological or a strictly syntactic analysis.Since it includes both parts of words （affixes） and free word forms （particles）, Wälchli （Unp.Ms.: 4） redefines this grammatical category as “a verb deprived of all its derivational affixes, encoding motion but retaining other affixes”.