1.1 The expression of motion in Turkish and German

The analysis of the construal of motion events has been a fruitful field of investigation in cross-linguistic studies over the last two decades. Talmy (Reference Talmy and Shopen1985) was the first to analyze the lexicalization of motion events across languages, that is, the systematic association of meaning with particular grammatical elements. A motion event refers to a situation where an entity moves or is located with respect to another entity and it is analyzed as having four internal components: figure, ground, path and motion (see Talmy, Reference Talmy and Shopen1985, Reference Talmy2000a, Reference Talmyb). These can be seen in (1):

1. (1)

The Figure is the object that moves or is located. The Motion component refers to the actual presence of motion or locatedness. The Ground is the reference object with regard to which the Figure moves or is located and Path represents the course followed or location occupied by the Figure with regard to the Ground.

Talmy observed that languages differ systematically with respect to the lexical elements that are used to encode information about Path. He suggested that languages can be categorized into two types on the basis of where they characteristically express Path (Talmy, Reference Talmy and Shopen1985, Reference Talmy2000a, Reference Talmyb). Satellite-framed languages (S-languages), such as English and German, typically express Path in a “satellite” of the verb – a member of a grammatical category of constituents other than a noun phrase or a prepositional phrase that is in a sister relation to the verb root. English verb particles such as in, out and across are typical examples of Path satellites. In German, these satellites can be prepositions or the separable part of a separable verb such as hinein-gehen “go into”. Verb-framed languages (V-languages), such as Spanish and Turkish, on the other hand, typically encode Path in the verb root of the main verb. Examples of Turkish Path verbs are çık- “ascend”, in- “descend” and gir- “enter”.

For the purpose of this investigation we also need to distinguish between the different components of which Path consists. According to Talmy, the three main components of Path are the Vector, the Conformation and the Deictic. The Vector consists of the basic types of arrival, transversal or departure of the Figure carried out with respect to particular grounds. The Vector component of the Path followed by the Figure consists, for example, of a movement TO a goal or FROM a source, or both. To this fundamental schema, the speaker can choose to add information about the Conformation, i.e., give further details about the geometric complex of which the Ground forms a part. For example, the speaker can choose to add information that the goal is “a point which is of the inside of [an enclosure] = INTO an enclosure” (Talmy, Reference Talmy2000b, p. 53). If the speaker of a V-language wishes to do this, s/he would use a verb that expresses the concept ENTER, whereas a speaker of an S-language would generally choose a Path particle that which expresses this same concept, namely INTO. Apart from this, there is the Deictic component of Path, which relates to the perspective of the speaker. In many languages, there are different verbs for the expression of movement towards the speaker, e.g. come, and movement in a direction other than towards the speaker, such as go.

The difference between the two lexicalization patterns has consequences for the expression of other aspects of the motion event. In addition to the internal components described above, a motion event can also have an external co-event such as manner, that is the specific way the Figure moves, and languages also differ in their characteristic encoding of Manner (Talmy, Reference Talmy and Shopen1985). Since Path is expressed outside the verb in S-languages “the verb is free to add nuances of MANNER without further elaboration” (Slobin, Reference Slobin, Maeder, Fisher and Herlofsky2005, p. 312; capital letters in the original). The verb slot can easily be filled with a Manner verb (e.g., run, jump) that conflates Manner and Motion in one lexical item. This option, however, is not available to speakers of V-languages since they tend to express Path through the use of Path verbs. The expression of Manner, then, might be easier for speakers of S-languages than for speakers of V-languages. As Slobin (Reference Slobin, Gentner and Goldin-Meadow2003, p. 162) puts it, speakers of S-languages get Manner “for free” whereas speakers of V-languages need much more effort since they need additional lexical items, such as participles, to express Manner in a motion event. In the next section, we outline the differences in the encoding of Manner in Turkish and German.

1.2 Conceptual transfer and thinking-for-speaking

Before embarking on a discussion of the literature on conceptual transfer, we need to define the concept of transfer itself. As Gass (Reference Gass, Ritchie and Bhatia1996, p. 318) and Jarvis (Reference Jarvis2000, p. 246) and many others have pointed out, this is difficult, not only because definitions are hardly ever theory-neutral, but also because the notion TRANSFER and many other terms used in bilingualism and related fields are often used in different ways by different researchers (see also Treffers-Daller, Reference Treffers-Daller, Bullock and Toribio2009). In defining transfer, many researchers refer to second language learners rather than bilinguals. Such definitions are less suitable for the current study, which focuses on bilinguals who have used two languages since early childhood. For the purposes of the current study it is important that the definition of transfer makes reference to bilinguals and to processes (as opposed to products that result from these processes), and covers phenomena that can occur in either language of a bilingual (bidirectional transfer) and can occur in patterns (not just elements in isolation). Although Haugen (Reference Haugen1950, p. 230) uses the label “borrowing”Footnote ¹ as the central concept, his definition comes closest to what we are looking for: “the process that takes place when bilinguals reproduce a pattern from one language in another.” Haugen does not elaborate on the cognitive implications of this process, but he does mention the existence of entrenched habits and points out that “the more habitual and subconscious a feature of language is, the harder it will be to change” (p. 224), which is interesting in relation to Slobin's Thinking-for-Speaking Hypothesis, which will be outlined below. Two important aspects of transfer are, however, not covered in Haugen's definition. First of all, transfer can be a dynamic process. That is, it can lead to the production of spontaneous phenomena resulting from “the involuntary influence of the deactivated language” during online processing (Grosjean, Reference Grosjean2008, p. 38). In other words, the phenomena need not (yet) be incorporated into the system of the recipient language. Second, according to Mougeon, Nadasdi and Rehner (Reference Mougeon, Nadasdi and Rehner2005, p. 103), transfer may be overt or covert. OVERT INTER-SYSTEMIC TRANSFER represents a qualitative development in one of the languages in a bilingual setting, such as the relaxing of a syntactic rule, whereas COVERT INTER-SYSTEMIC TRANSFER refers to a quantitative development, namely a marked increase in the frequency of a feature at the expense of an alternative feature (or alternative features).

Linguistic transfer (also called cross-linguistic influence) affects all linguistic subsystems (Odlin, Reference Odlin, Doughty and Long2005), including semantics.Footnote ² We assume with Levinson (Reference Levinson, Juytes and Pederson1997) that semantic representations are distinct from conceptual representations (but see Cadierno (Reference Cadierno, Robinson and Ellis2008) for a different opinion). If this is indeed the case, it is important to clarify the difference between semantic transfer and conceptual transfer, prior to reviewing the available evidence with respect to the existence of transfer involving conceptualization patterns.

Semantic transfer involves the link between words and concepts but not the make-up of the concepts themselves (Jarvis, Reference Jarvis2007). The following example of semantic transfer in an utterance produced by a Finnish learner of English is helpful to illustrate the difference between semantic and conceptual transfer: “he bit himself in the language” (Ringbom, Reference Ringbom, Cenoz, Hufeisen and Jessner2001, p. 64). The speaker's intention was to say “in the tongue” but Finnish has only a single word for “language” and “tongue”, which caused this incidence of semantic transfer. In this case the speaker's L1 has exerted influence on his/her L2, but there is no indication that transfer has taken place at a conceptual level, since “conceptual knowledge involves extralinguistic mental representations” (Jarvis & Pavlenko, Reference Jarvis and Pavlenko2008, p. 73) and it is obvious that the speaker knows the difference between “language” and “tongue” at a conceptual level. When, however, Russian learners of English refer to paper cups as glasses, this is arguably a case of conceptual transfer, because English and Russian differ in the categorization of objects as “glasses” or “cups” and the learner uses an L1 concept in an L2 context (see Jarvis & Pavlenko (Reference Jarvis and Pavlenko2008) for further details). The information provided about this example does not allow us to establish, however, whether the speaker's inventory of mental concepts has been affected or whether the conceptual transfer occurred during the online processing of those concepts. As Jarvis (Reference Jarvis2007) points out, it is important to distinguish between concept transfer and conceptualization transfer. The first of these two refers to transfer that arises from cross-linguistic differences in the conceptual categories stored in L2 users’ long-term memory, whereas conceptualization transfer arises from cross-linguistic differences in the ways L2 users process conceptual knowledge and form temporary representations in their working memory.

Evidence for the existence of both kinds of conceptual transfer has led to the formulation of the Conceptual Transfer Hypothesis (Jarvis, Reference Jarvis2007), according to which a person's patterns of language use in one language can reflect the concepts and patterns of conceptualization that the person has acquired as a speaker of another language. This hypothesis is closely related to Slobin's (Reference Slobin1987) Thinking-for-Speaking Hypothesis (but see Jarvis (Reference Jarvis2007) for a discussion of similarities and differences). In Slobin's framework, language specificity is not to be found in the mental images that individuals have in their conceptual stores (which would be a case of concept transfer when it involves cross-linguistic effects), but it is found with respect to which (elements) of those images are selected for verbalization – a phenomenon whose cross-linguistic effects are referred to as conceptualization transfer in Jarvis's terminology. Thinking for speaking involves “picking those characteristics that (a) fit some conceptualisation of the event and (b) are readily encodable in the language” (Slobin, Reference Slobin1987, p. 435). As events cannot be directly encoded in language, speakers or writers need to verbalize a particular aspect of motion event, that is, “experiences are filtered through language into verbalised events” (Slobin, Reference Slobin, Gumperz and Levinson1996, p. 75; emphasis in original). Children become used to paying attention to particular aspects of motion events and to particular ways of verbalizing these from a very young age, as Aksu-Koç (Reference Aksu-Koç, Berman and Slobin1994) and Allen et al. (Reference Allen, Özyürek, Kita, Brown, Furman, Ishizuka and Fuji2007) have shown for the L1 acquisition of Turkish. From a cognitive linguistic point of view, this is important because the training children receive in the verbalization of Motion may lead to these verbalizations becoming cognitively entrenched in that they are well-rehearsed and automatic routines that are easily elicited and reliably executed (Langacker, Reference Langacker2000, p. 3). Slobin (Reference Slobin, Gumperz and Levinson1996, p. 89) claims that there are important consequences for second language learning: if the training one receives in childhood is “exceptionally resistant to restructuring in adult second-language acquisition”, it is likely that the patterns learnt in early childhood are transferred to a second language.

In their studies of the cognitive processing of motion events, von Stutterheim et al. (Reference Von Stutterheim, Nüse, Murcia-Serra, Behrens, Fabricius-Hansen and Johansson2002) and von Stutterheim and Nüse (Reference Von Stutterheim and Nüse2003) provide evidence for the hypothesis that in the construal of events, speakers do indeed access extralinguistic mental representations (and not only linguistic forms). Their work shows that there are systematic differences between speakers of different languages in the ways in which they plan the production of utterances, that is, in the information they select for verbalization, in the segmentation, structuring and linearization of that information and in the perspectives they take on the event. Crucially, all the tasks involved in the construal of an event have to be carried out in the Conceptualizer (Levelt, Reference Levelt1989), prior to the retrieval of linguistic forms. Thus, the principles or strategies (or blueprints) used by speakers to construe events are deeply rooted in the planning stages of verbal processing.

An important question that remains to be answered is then at what level of processing transfer takes place. Levelt (Reference Levelt1989) suggests that the construction of the preverbal message in the Conceptualizer involves two steps: macro-planning and micro-planning. In macro-planning, the speaker elaborates the communicative intention as a sequence of subgoals and selects the information that is to be represented in speech. Micro-planning is concerned with the further shaping of the message to bring it into the format required by the preverbal message. The product of conceptualizing is the preverbal message, which is then passed on to the Formulator for grammatical and phonological encoding. According to Levelt (Reference Levelt1989, p. 145), one aspect of micro-planning is the acknowledgement of the Formulator's language-specific requirements. In the course of language acquisition, children learn what kinds of distinctions are relevant for their mother tongue(s); for example, whether a language has aspect or not. For adults, the language-specific requirements on semantic structure are part of the Conceptualizer's procedural knowledge base (p. 105). The speaker is assumed to have a set of routine procedures that acknowledges automatically whatever the language requires (p. 157).

Some evidence regarding the planning of multi-clause utterances can be obtained from studies on fluency. Pawley and Syder (Reference Pawley, Syder and Riggenbach2000, p. 195) suggest that “competent speakers are able routinely to plan in advance the core (or obligatory) grammatical phrases of simple clauses and of certain kinds of complex clauses”. As is well known, speakers do not create all utterances from scratch. Instead, speakers rely on a stock of familiar multi-clause syntactic frames that are memorized independently of lexical content (p. 195).

The notion of routine procedures or blueprints is related to the concept of thinking-for-speaking. We assume that these blueprints are cognitively entrenched, which makes it difficult to learn a new way of thinking-for-speaking, and that L2 learners use L1 blueprints in building structures in L2. Support for this position comes from Caroll and von Stutterheim (Reference Carroll, von Stutterheim and Ramat2003, p. 398), who show that even advanced L2 learners “remain rooted in at least some of the principles of conceptual organisation as constituted in the course of L1 acquisition”.

It remains controversial, however, to what extent language influences conceptualization. Levinson (Reference Levinson, Juytes and Pederson1997, p. 22) assumes that the acquired habits of language production result in speakers of typologically different languages entertaining systematically different thoughts, for example, in the domain of spatial relations. Thus, in his view, the influence of language on concept formation goes much further than the levels of macro- or micro-planning, and extends to general cognition. This would also be the position of neo-relativists, who argue that cognition in general is influenced by language-specific categories (see Lucy (Reference Lucy1992a, Reference Lucyb) or Danziger (Reference Danziger2001) for a fuller discussion).

Von Stutterheim and Nüse (Reference Von Stutterheim and Nüse2003), on the other hand, do not think that cognition in general is language-dependent, but argue that language influences micro-planning and macro-planning. They elaborate on Levelt's model by assuming that the planning of utterances involves four stages – segmentation, selection, structuring and linearization – and present evidence that there are systematic differences between speakers of English and German in the ways they organize information under different perspectives. We hope to provide some evidence that there are systematic differences between bilingual groups in the ways in which they select and linearize units of information, and that these differences relate to their language dominance patterns. To our knowledge, the issue of linearization has not yet been taken up in studies of conceptual transfer (apart from Caroll and von Stutterheim's (Reference Carroll, von Stutterheim and Ramat2003) study on perspective taking) and therefore this contribution is one of the first in this area.

Evidence for conceptual transfer in second language learners can be found in various studies. Inagaki (Reference Inagaki2001), for example, found that Anglophone learners of Japanese accept English-like manner of Motion verbs preceded by directional PPs in Japanese, although these are rejected by Japanese control groups. Cadierno and Ruiz (Reference Cadierno and Ruiz2006) also found some evidence for conceptual transfer in L2 expressions of motion events: their Danish learners of Spanish incorporated more ground specifications in Spanish clauses than Spanish native speaker groups, and they also produced constructions that involved a violation of the boundary crossing constraint. In their study of oral narratives of British learners of French and French learners of British English, Treffers-Daller and Tidball (in prep.) found evidence for the transfer of L1 conceptualization patterns, including violations of the boundary crossing constraint even among advanced learners of French. There is therefore some evidence that it is difficult to restructure conceptualization of Motion in the process of L2 learning. According to Cadierno (Reference Cadierno, Robinson and Ellis2008, p. 265), however, advanced learners are able to retrain their thinking-for-speaking patterns to a large extent when talking about Motion in an L2 that is typologically different from their L1, although in some areas (e.g., gestures) traces of L1-based patterns are still detectable in these learners.

1.3 Conceptualization of motion events in bilinguals

Most studies in this area of research deal either with the development of conceptualization patterns in the first language or with the transfer of L1 patterns in L2 acquisition. There are only a few studies that deal with the development of conceptualization patterns in bilinguals who grow up with two languages simultaneously and are exposed to conflicting patterns in their two languages from birth. Bidirectional cross-linguistic transfer can be assumed in these cases, but we know little about the factors that influence this type of transfer. Slobin (Reference Slobin, Maeder, Fisher and Herlofsky2005, p. 318) claims that the two languages of a bilingual trigger different conceptualization patterns in that “bilinguals tested in both languages systematically report more mental imagery for Manner of motion, and less for physical surroundings, when reading in English, in comparison with Spanish”. One of the characteristics that sets conceptual transfer apart from the types of transfer that are more purely linguistic in nature (e.g., phonological, syntactic, semantic transfer) is that conceptual transfer appears to be more inherently bidirectional and convergent (see Jarvis & Pavlenko, Reference Jarvis and Pavlenko2008). That is, studies that investigate cross-linguistic influence in learners’ and bilinguals’ patterns of conceptual categorization and event construal often find that the participants show simultaneous influences from both languages in a single domain, with the result that the way they express motion events, for example, falls somewhere between the patterns found among monolingual speakers of the two languages. This finding emerges in several of the articles in this issue, as well as in several previous studies, including, but not limited to, several studies discussed in Jarvis and Pavlenko (Reference Jarvis and Pavlenko2008).

In one such study that focused on bidirectional transfer in the conceptualization, or construal, of motion events, Hohenstein et al. (Reference Hohenstein, Eisenberg and Naigles2006) compare the use of Manner and Motion verbs by Spanish–English bilingual and monolingual groups. The bilinguals in this study had all learned Spanish first and then either started learning English at the age of 5 or below (early group) or after the age of 12 (late group).Footnote ³ The authors show that Spanish monolinguals used more Path than Manner verbs to describe motion events presented in videos, whereas English monolinguals used many more Manner verbs than Path verbs for the same events. This is in line with expectations, as Spanish is classified as a V-language and English as an S-language. Spanish–English bilinguals, however, have a tendency to comply with the language-specific patterns when speaking either Spanish or English, but they take the middle position overall. Thus, their Spanish descriptions include fewer Manner verbs than their English ones, but still more compared to the Spanish descriptions of the monolinguals. On the other hand, their English descriptions contain more Manner verbs than their Spanish ones, but still fewer than the description of the monolingual English speakers. Hohenstein et al. (Reference Hohenstein, Eisenberg and Naigles2006, p. 258) report that they found bidirectional transfer specifically among late bilinguals and not among early bilinguals. This is a clear indication that bidirectional transfer takes place in principle, but that it is influenced by factors such as the age of onset of second language learning. They also indicate that language dominance may play a role in transfer, and point out there is the need to study the differential effects of language dominance.

The acquisition of language-specific conceptualization patterns seems to start from a very early age. Hickmann and Hendriks (Reference Hickmann and Hendriks2006) show that three-year-old monolingual children follow the specific patterns of their first language in their description of spatial relations, and Choi and Bowerman (Reference Choi and Bowerman1991) find evidence that bilingual children are sensitive to the routinized patterns of their two languages from as early as 17–20 months. This is important for the current study, because bilinguals who have grown up with two languages will have become sensitive to the routinized patterns of two (or more) languages from a very early age. This makes the construal of motion events by bilinguals a promising research area for research into the transfer of conceptualization patterns.

Studies on bilinguals in language contact settings can give us more insight into how bilinguals in these settings react to the preferences of their different languages. Of special interest are contact situations between S-languages and V-languages. According to Kramer (Reference Kramer, Meid and Heller1981) and Slobin (Reference Slobin, Strömqvist and Verhoeven2004), in many Romance–Germanic contact situations along the geographical border between these languages, Path particles are incorporated by bilinguals into their use of the Romance languages and used in combination with path or manner verbs. Some evidence for this was found in frog stories collected among French–Dutch bilinguals, who used a redundant Path particle en bas “downwards” in combination with the Path verb tomber “fall” (Treffers-Daller, in press b) as in (7), or a Manner verb courir “run” in combination with a prepositional phrase dans les bois “into the woods” as in (8), which expresses attainment at a final location. This is not allowed in standard French because of the boundary crossing constraint.Footnote ⁴

1. (7)

2. (8)

Not all uses of redundant directional adverbs can be explained as resulting from transfer, however. Sebastián and Slobin (Reference Sebastián, Slobin, Berman and Slobin1994, p. 264) report that redundant directional adverbs can be found among preschool Spanish children (9). They also report that children used directional adverbs that were not redundant but more elaborate than one would typically find among older speakers.

1. (9) Entonces abajo se cayó

   “And downwards (he) fell.”   [s4e, 4;6] (Sebastián & Slobin, Reference Sebastián, Slobin, Berman and Slobin1994, p. 264)

Sebastián and Slobin assume that children may feel the need to reinforce the directional meaning already expressed in the verb, and suggest there may be a U-shaped development pattern here, as the phenomenon is less common with the youngest group and with older speakers.

Cadierno and Ruiz (Reference Cadierno and Ruiz2006, p. 204) found similar uses of redundant and anomalous Path particles in Spanish as spoken by Danish learners (10–11).

1. (10) La colmena cayó abajo

   “The beehive fell down.”

2. (11) El perro salío afuera

   “The dog went outside.”

Further examples of conceptual transfer in the domain of Motion can be found in the work of Helms-Park (Reference Helms-Park2001, Reference Helms-Park2003), and Odlin (Reference Odlin, Robinson and Ellis2008) provides a detailed overview of further studies in the field. The available evidence suggests that conceptualization patterns are indeed transferable. If this is the case, this constitutes important evidence for Jarvis's (Reference Jarvis2007) Conceptual Transfer Hypothesis.

1.4 Language dominance

As the current study aims to contribute to the discussion about the ways in which language dominance affects transfer, it is important to briefly outline our view of language dominance here first (but see Daller, Yıldız, De Jong, Kan and Başbağı (in press) and Treffers-Daller (in press a) for a fuller discussion). Hamers and Blanc (Reference Hamers and Blanc2000, p. 9) consider language dominance to be one of the psychological dimensions of bilingual ability (or in their terminology “bilinguality”) in individuals. Under this view, language dominance refers to the relationship between the competencies in the two languages of the bilingual. A distinction is often made between balanced bilinguals, who have equivalent competence in both languages, and dominant bilinguals, whose competence in one language is more developed than that in the other (p. 8). Many researchers point out that completely balanced bilinguals are very rare. It is much more common for bilinguals to be dominant in one or the other language (Baker & Jones, Reference Baker and Jones1998, p. 12). To Grosjean (Reference Grosjean2008, p. 24), language dominance is a reflection of the complementarity principle, i.e., the fact that bilinguals use their two languages for different purposes (or different domains) in daily life.

Language dominance is, however, a factor not only at the micro-level of the students’ individual language proficiency profiles, but also at the macro-level of society. Hohenstein et al. (Reference Hohenstein, Eisenberg and Naigles2006, p. 259) recognize this when they note that these two different kinds of language dominance may affect transfer, and they call for research in which bilinguals using the same two languages are studied in different environments, so that the differential role of both forms of language dominance may be evaluated.

We take the view that language dominance at a societal level is reflected in the individual. We use the term “reflection” on purpose because we assume that the changes in a bilingual's environment trigger the changes in his/her language dominance (not vice versa). Therefore, we consider societal or cultural language dominance to be the key factor behind language dominance at the individual level. Although we cannot separate these two aspects of dominance in a clear-cut way, it seems important to point out that if the language environment changes, e.g., in the current study through the return to Turkey of one group of participants, it is most likely that the individual dominance patterns will also change eventually. Support for this position comes from Grosjean (Reference Grosjean1998, p. 132), who explains that “the language repertoire of bilinguals may change over time: as the environment changes and the needs for particular language skills also change, so will their competence in their language.” It is exactly this change in the environment that distinguishes the two bilingual groups in the present study.

2.1 Basic word order in Turkish

According to Göksel and Kerslake (Reference Göksel and Kerslake2005), the unmarked word order for major constituents in Turkish sentences is Subject (Object) Verb (SOV). Thus, the predicate appears at the end of a sentence. In standard Turkish, nominalized subordinate clauses (including postpositional clauses) normally precede the predicate in the main clause (Kornfilt, Reference Kornfilt1997, p. 46). This is also the case with postpositional clauses containing the postposition için “to, for”, that express the goal or the reason of an action, as exemplified in (12), where the postpositional clause (topu almak için “to fetch the ball”) precedes the main clause, which expresses the action itself (babası iniyor “his father descends”). The postpositional clause is bracketed in the example to clarify the underlying structure. Alternative word orders, such as the one in (13), where the postpositional clause is extraposed to the right side of the main verb, are possible depending on considerations of topic, focus, etc. (Kornfilt, Reference Kornfilt1997, p. 91).

1. (12) [Top-u al-mak için] baba-sı iniyor

   Ball-Acc fetch-Nom to father-Poss descend-Prog

   “To fetch the ball, his father descends.”

2. (13) Baba-sı in-iyor [top-u almak için]

   father-Poss descend-Prog ball-Acc fetch-Nom to

   “His father descends to fetch the ball.”

For the purposes of the current investigation, it is important to note that the unmarked word order is the one given in (12) (see Kornfilt, Reference Kornfilt1997; Göksel & Kerslake, Reference Göksel and Kerslake2005).

In German, on the contrary, the unmarked order of these constituents is different. The most common way to express an Action–Goal sequence is as given in (14).

1. (14) Der Vater steigt in den Gully [um den Ball zu holen].

   “The father climbs into the manhole [to fetch the ball].”

The alternative order, given in (15), is possible but considered as marked (see Schoenthal, Reference Schoenthal and Glück1993).

1. (15) [Um den Ball zu holen] steigt der Vater in den Gully.

   “[To fetch the ball] the father climbs into the manhole.”

An alternative way of looking at these constructions would be to consider them as examples of syntactic transfer. It is possible to hypothesize that VO structures have increased more generally in immigrant Turkish in Germany, and that this increase is due to the influence of German. If this were the case, we would expect to see a wide range of constituents in postverbal position, and not only Goal PPs. To the best of our knowledge, there is no evidence for this hypothesis so far, even though, according to Johanson (Reference Johanson2002), constant copying from Slavic and Romanian into the Turkic languages spoken in the Balkan area has resulted in word order changes in Gagauz and Karaim to such an extent that complements and free adverbials generally follow the predicate. Doğruöz and Backus (Reference Doğruöz and Backus2009) did not find much evidence for the increase of VO structures in Turkish as spoken in the Netherlands, although in a follow-up study, Doğruöz and Backus (Reference Doğruöz and Backus2009) discuss evidence for other patterns that native speakers of Turkish as spoken in Turkey would classify as “unconventional”, many of which could be the result of Dutch influence. Whether evidence for an increase in OV patterns can be found in Turkish as spoken in Germany is a very interesting question, but it is clearly beyond the scope of this article.

2.2 Hypotheses

The key hypotheses of the current study are that both groups of bilinguals will display bidirectional transfer in the conceptualization of motion events and that the balance of influence from each language can be predicted by the dominant linguistic environment in which the groups live. Since our data are based on two different picture stories (see section 3, “Method”) we need to have different hypotheses for the two languages, but they all focus on bidirectional transfer. Based on the theoretical assumptions we discussed so far we predict the following for the construal of motion events in turkish:

1. (i) Bilinguals in Germany will describe Path in Turkish in ways that are more typical for German, in that they will choose to

   1. (a) select the Conformation component (ENTER) of Path, which is atypical for monolingual Turkish speakers.

   2. (b) use more redundant Path satellites than returnees and monolinguals in Turkey.

2. (ii) Bilinguals in Germany will choose to linearize the different components of Event 1 as an ACTION–GOAL sequence, which is typical for the German patterns for organizing information in clauses, whereas returnees will be closer to the Turkish monolingual patterns for the linearization of motion events in that they prefer GOAL–ACTION sequences.

3. (iii) A subgroup of bilinguals who live in Germany but had had more contact with Turkish at the time of recording (due to their specific migration background) will experience less influence from German in their conceptualization of events in Turkish.

4. (iv) The differences between the bilinguals and the monolinguals cannot be explained as resulting from incomplete acquisition of the typical patterns in which events are construed in Turkish; therefore all groups will avoid Manner verbs in boundary crossings in Turkish.

For the construal of motion events in german we predict that:

1. (i) The returnees will use fewer Manner verbs than the bilinguals in Germany for the description of motion events with a boundary crossing in German, because the returnees experience transfer from routinized patterns involving boundary crossings in Turkish, which does not license the use of Manner verbs in such constructions.

2. (ii) The bilinguals in Germany will use more Manner verbs in this context than the returnees, and thus will be closer to the routinized patterns of monolingual German speakers.