Motion into and out of in English , French and Norwegian

This paper presents a contrastive study of Norwegia n predications of motion events with the compound prepositions ut av (‘out of’) and inn i (‘into’) and their translations into English and French. The motivation for choosing the se two types of predication is that French, unlike English, is said to avoid the use of manner verbs with boundary-crossing events. The paper examines all occurrences in the O slo Multilingual Corpus (OMC) of self-motion predications containing the two Norwegi an prepositions, in all of which path is coded in the prepositional phrase. The verb may also code path, it may code manner, or it may be a neutral verb of movement. We first anal yse the Norwegian originals with respect to their coding of path and manner and then urn to the two sets of translations and investigate the extent to which they retain the manner/path coding choices of the source predications and, if not, what sort of alter tions they make. If the contention that French avoids manner verbs with boundary-crossing a ctions is correct, the French translations should exhibit a much greater degree o f path or neutral motion coding in the verb than either the Norwegian originals or the Eng lish translations. The data show that this is indeed the case. There are also, however, m o e occurrences of manner verbs in French with boundary-crossing actions than one woul d expect given the language’s reputation in the literature for avoiding this cons truction.


Introduction
In this paper we compare and contrast English and French translations of Norwegian predications of motion events containing the boundarycrossing compound prepositions ut av ('out of') and inn i ('into').The point of departure is the typological distinction between path-framed languages, where the semantic path component (direction of the movement) in a motion event is characteristically expressed in the verb, and satellite-framed languages, where the path of motion is typically expressed in a satellite (narrowly defined by Talmy 2000: 17 as an adverbial particle).According to Slobin (2006: 70), "in translations […] manner salience follows patterns of the target, rather than source language".Thus a translation into a path-framed language will most likely itself be path-framed even in cases where the source text codes manner in the verb phrase.This is likely to apply to an even greater extent to cases in which the preposition codes the crossing of a boundary, given that French is said to avoid the use of manner verbs with telic actions in general (Aske 1989: 6) and actions involving boundarycrossing in particular (Cappelle 2012: 189).Cappelle (2012) disagrees with Slobin on the question of manner salience in translated texts, maintaining that translations will often retain traces of the typology of the source text.Although both Slobin and Cappelle base their conclusions on studies of translated data, Slobin on translations from English into Spanish and vice versa, and Cappelle on translations from both French and German into English, neither of them compare translations into different languages of one and the same source.Such data have the obvious advantage of allowing one to compare in detail the alterations made by two sets of translators to the same set of original texts.If Slobin is correct in his contention that manner follows the target language the French translations should exhibit a much greater degree of path coding in the verb than either the Norwegian originals or the English translations.If Cappelle is correct there should be fewer differences between the two sets of translations.In particular we might expect there to be some tokens in French in which manner rather than path is coded by the verb.
Our study is based on data from the No-En-Fr-Ge part of the Oslo Multilingual Corpus (hereafter referred to as OMC), which consists of long extracts from five Norwegian novels, together with their translations into English, French and German.We have only looked at the Norwegian originals and English and French translations.We looked at all tokens of self-motion predications containing the two Norwegian prepositions ut av and inn i, in all of which path is coded in the prepositional phrase.Note that we are using the term 'self-motion' to subsume all motion predications of the type Subject-Verb-Adverbial, irrespective of whether the subject is agentive.The verb in these predications may also code path, as in (1), it may code manner, as in (2), or it may be a neutral verb of movement, as in (3).In these and all other examples underlining indicates coding of path, italics coding of manner, and bold type coding of neutral motion.
(1) a. Han hadde falt ut av vinduet….(NF1) 1 b.He'd fallen from the window… (NF1TE) (2) a. Jeg kjente meg litt svimmel da jeg stupte inn i landhandelen… (JG3) b.I was reeling a little by the time I dived into the village store… (JG3TE) (3) a. En fjern og ennå utydelig skikkelse tar form og beveger seg inn i synsfeltet… (BHH1) b.A vague, faraway figure takes form and moves into my field of vision… (BHH1TE) In ( 1)-( 3) the English translators have retained the coding of path and manner in the Norwegian originals.Our main concern in this article is the extent to which the English and French translators retain this original coding and, in cases where they do not do so, the sorts of changes they make.
In section 2 we present some theoretical perspectives and our methodological approach.Section 3 describes the classification system employed, with a particular emphasis on the Norwegian verb gå, which poses problems for classification.In section 4 we compare the codings of path and manner in the English and French translations of predications containing the Norwegian ut av.Section 5 contains a similar analysis of the texts containing Norwegian inn i. Finally section 6 contains a summary and conclusion.

Theory and methodology
Over the last quarter of a century, Talmy's distinction between pathframed and satellite-framed languages has given rise to a plethora of studies, theoretical and empirical, monolingual and multilingual.These studies have led to progressive refinements of the distinction, with Slobin 1 The first part of the code 'NF1' refers to the text in the OMC from which the example has been taken, with 'NF' being the initials of the author.'TE' means translated text in English; 'TF' stands for translated text in French.The full titles of the original works and the translations in the OMC are listed in Johansson (2007: 349-350).(2006: 64), among others, arguing the need to allow for a third type of framing, 'equipollent-framing', to cater for the sort of motion coding found in some serial verb languages.Other scholars, such as Pourcel and Kopecka (2005), Kopecka (2006) and Hickman et al. (2009), have demonstrated that French, which was originally considered a pathframed language, actually employs a variety of constructions to code motion events.Path-framing is just the most commonly employed construction.Indeed Croft et al. (2010) have argued that it is misguided to define languages in terms of framing.They maintain that there is a variety of framing construction types, six in all, and that some languages may make more use of one of these types, such as path-framing, without this constituting grounds for us calling the language itself 'pathframing '. 2  Whether one thinks of the various types of framing in terms of constructions or as languages that (proto-)typically employ these constructions, there is no getting away from the fact that Germanic and Romance languages, for example, differ markedly in the types of construction they routinely employ to code motion events.One of Talmy's seminal examples was of a bottle floating into a cave (or entering a cave (by) floating) and it is precisely this type of predication, involving a telic motion event in which a boundary is clearly crossed in a certain manner, that is frequently said to be most resistant to encoding in a verb-framed language by a satellite framed construction.Thus Beavers et al. state, with reference to Aske (1989), that "V-framed languages disallow boundary-crossing path satellites with manner verbs, although they may allow non-boundary-crossing path satellites" (2010: 347).In other words, predications of non-telic motion events, such as those containing paths coded in English and French by prepositional phrases headed by the prepositions towards/vers are more likely to be deemed acceptable with manner verbs in path-framing languages than those 2 Beavers et al. (2010) maintain that "Many languages that allow encoding possibilities 'against' their Talmyan type may in practice disprefer them as they are more complex than other available options.However, […] other factors, especially pragmatic factors, may sometimes cause the more complex types to be favored, an outcome that is only expected if, as on our approach, such options are in principle available" (Beavers, Levin and Wei Tham 2010: 335).
containing boundary-reaching prepositions such as jusqu'à/to.3Acceptability judgements, however, are likely to vary according to the verb in question, as shown by Pourcel and Kopecka (2006).Moreover, in a study based on a corpus of French novels and travelogues, Kopecka (2009) shows that some French manner verbs, such as sauter and grimper are actually more likely to be used in predications involving boundary-crossing ('changement de lieu') than predications of motion within a set of boundaries ('changement d'emplacement').
The fact that boundary-crossing events constitute the hard-core of motion predications which are less likely to be coded by manner verbs in verb-framing languages lies behind our decision to investigate predications of motion [into] and [out of], which by definition encode the crossing of a boundary.In order to carry out a comparison between the codings of such events in English and French, one needs a set of examples from both languages that encode the same motion events.One common method used to ensure comparability is to show informants visual representations of events in picture books or video snippets and ask them to describe what is going on (some examples of the former method may be found in Strömqvist and Verhoven 2004; for just one example of the latter see Engemann et al. 2012).This sort of procedure may seem to overly rely on the ability of the participants in the experiments to see the same event in the picture.However, if they do not do so, in other words if they construe the scene differently, this very difference may actually be a reflection of the way their mother tongue characteristically codes events (see Slobin's 1996 notion of 'thinking for speaking').Be that as it may, we have chosen to ensure that the tokens we compare in our study are compatible by choosing informants who are given identical prompts.Our tertia comparationis are verbal rather than visual (see Egan 2013 for the use of original texts in multilingual corpora as tertia comparationis).The translators in our study were not at liberty to construe the motion events in the original texts as freely as observers of a picture, although they were, of course, at liberty to re-construe them. 4It is indeed the very fashion and extent of their reconstruals that provide us with the material for our discussion in sections 4 and 5.
Since all translators are necessarily functionally bilingual one must bear in mind the possibility that they choose to retain in their translation the construal coded in the source text, especially if this construal involves a form of coding also possible to encode in the target language, rather than opt for a construal involving a different, albeit more common form of coding in the target language.Given the differences between French, a path-framed language, on the one hand and Norwegian and English, two satellite-framed languages on the other, one might expect to find more evidence of this lack of reconstruals in the French translations, at least if Cappelle (2012) is correct in his assertion that translators will often retain the coding of the source text, where this is typologically possible.One might for instance expect to find more tokens of manner verbs in the French translations than one would in a comparable corpus of French original texts.In order to investigate this hypothesis one would ideally need access to a corpus containing French originals and Norwegian translations in which one could search for translations containing the two forms inn i and ut av.Unfortunately, the only such corpus of which we are aware, the French Norwegian Parallel Corpus, is too small to furnish us with sufficient tokens of these low-frequency prepositions. 5 As pointed out in the introduction our material consists of translations into English and French of all tokens of self-motion in the OMC containing adverbials in the form of prepositional phrases headed by the two complex prepositions ut av and inn i.We first downloaded all sentences containing these prepositions, before manually extracting the tokens coding motion predications.We then set aside tokens instantiating caused motion, which is conceptually more complex than self-motion, entailing as it does an extra participant, a Causer in addition to a Mover.The decision to omit these from our study was prompted solely by 4 Highly polysemous verbs in the source language may constitute exceptions with respect to the extent of coerced construal.See the discussion of the Norwegian verb gå in section 3. 5 There are approx.111,200 words from fictional French original texts in the French Norwegian Parallel Corpus (FNPC), as opposed to 439,687 words of translated French in the sub-corpus of the OMC in the present study.There are just 18 tokens of ut av in self-motion constructions in the FNPC.practical concerns.To do them justice would require a separate article.We were left with 225 tokens of self-motion that were translated into both English and French.Both authors (one with Norwegian as L1 and the other with English as L1) analysed the data, identifying the semantic dimensions of the source and target text tokens.We first analysed the Norwegian originals with respect to their coding of path and manner (path being, of course, always coded in the prepositional phrase, but possibly also in the verb, as in example (1) above).We then turned to the two sets of translations and investigated the extent to which they retain the manner/path coding choices of the source predications and, if not, what sort of alterations they exhibit.The two independent analyses were then compared and coordinated.We present the results of these analyses in sections 4 and 5. First, however, a few words must be said about how we went about classifying the tokens.

Classification system
An overview of all categories included in the classification system that were applied to both the source text and target text verbs and adverbials in the present study is provided in Table 1.The categories are illustrated with English examples where available.Although the categorization of the tokens in the data was for the most part fairly straightforward, there were cases where distinctions were less clear with respect to both semantic range as well as structural complexity.

The compound prepositions inn i and ut av
First, the source text prepositions inn i and ut av may on the surface seem completely parallel and syntactically congruent.However, only inn i is given the status of a compound preposition in standard dictionaries, while ut av is not listed as an entry in the motion sense.Even though this would not affect the results of the present study where the focus is on the translations of the predications, it indicates a possible difference: while inn i functions as a unit, ut av sometimes seems to be a combination of a particle followed by a prepositional phrase, as in the sentence Han kom ut av skapet ('he came out of the closet') which could potentially be analysed as either 'han [kom ut] [av skapet]' or 'han [kom] [ut av skapet]'.In contrast, many Norwegian dialects and the written standard of Nynorsk (and Swedish) have a distinction between the separate prepositions ut + av and the unit (ut) or (from old Norse ór) with the meaning 'about motion: out from, out of'. 6 Since all occurrences of ut av in the data can be replaced with the compound preposition (ut) or with no effect on the meaning, they have all been considered as compound prepositions, parallel to inn i.

The verbs gå and komme
Another problem is related to the categorization of the Norwegian motion verb gå.When used as a self-motion verb with a human subject its basic meaning is 'walk', which codes manner of motion.However, in a wide range of contexts gå corresponds to its English cognate go ('move').In other words the verb is underspecified with respect to manner of motion, since all instances of walking are also instances of going.Viberg (2013) relates this to "the shift of attention and the selection of elements in the situation that are coded linguistically", referred to as profiling (Viberg 2013: 32).In contexts where the focus of attention is on moving on foot, as in Hun går på tå ('She walks on her toes'), the manner of motion is profiled, and the basic English translation equivalent is walk, while in other contexts where the manner of motion is irrelevant, and therefore not profiled, as in Hun går inn igjen ('She goes back inside'), the basic English translation equivalent is go. 7he translation data were obviously not used as guidelines for the categorisation of tokens of gå in this study, as this would result in circularity.Instead, tokens of gå that co-occur with manner adverbials were categorised as verbs encoding manner, since the manner of motion is specifically profiled in those contexts, as in (4).
(HW2) b.Then he trudged barefoot across the courtyard and into the hen-house.(HW2TE) Some manner adverbials are not directly related to the motion, such as flirende ('laughing') in (5a), and were not considered as relevant to the categorisation of the verb, which was classified as encoding neutral motion here, despite the choice of walk in the English translation (5b).
(5) a. Flirende går vi ut av lokalet.(KF1) b.We walk out of the place laughing.(KF1TE) The choice of walk in (5b) detects the fact that even in cases where manner is not profiled, the verb gå is not completely neutral when the subject is human.However, tokens of gå without any manner adverbials were categorised as encoding (relatively) neutral motion, as in ( 6) and ( 7).As these examples illustrate, although none of them contains the explicit coding of manner as defined above, the actual manner of motion ('walking') can be potentially implicit -and thus open for different construals on the part of the translator -in some cases, as in ( 6) ('we walked into the hut') while not in others, such as (7) (*'I walked out of my bed').A corresponding problem occurs with the categorization of the Norwegian verb komme ('come'), which may encode either path or neutral motion, depending on whether or not the motion is in the direction of the focussed participant in the context (e.g. the speaker).In most cases the verb komme encodes neutral motion, as in (8), whereas in (9) the verb expresses a path towards the main character ('me').The data in Table 2 is reproduced figuratively in Figure 1.It may be worth noting that the adverbial takes the form of a gerund as in (13) in only two of the 12 French manner adverbial tokens.The other token is also en courant.Thus in the language of these French translations we are five times more likely to encounter manner coded by the main verb, as in (12c) than by a gerund, as in (13c).These results also accord with those of Morita (2011), who studied translations of literary texts from Japanese into French and vice versa: he writes French gerundive is syntactically the most independent element and appears least in our data" (Morita 2011: §85).Even allowing for the influence of the source texts on the form of coding in the French translations, the infrequency of the gerund in translations from both a satellite-framed and a path-framed language is very much at odds with the descriptions of standard codings of manner in Romance languages in the (early) verb-and satellite-framing literature.
English is very similar typologically to Norwegian, as is shown by the similarities in the participants' encoding of motion events in the freenaming study reported on by Vulchanova et al. (2012).If anything, Norwegian is rather more satellite-framed than English since, as a result of the Norman Conquest, English contains path verbs such as enter and descend, where Norwegian has a combination of a verb and a particle.Nevertheless, we see in Table 1 that there are five more tokens in English coding manner than there are in the Norwegian originals.Two of these are cited as ( 14) and ( 15).Any suggestion as to why the English translators have chosen to substitute a manner verb for the neutral motion verbs kommet ('came') and gikk ('went') in ( 14) and (15) can only be speculative.However, one can point to a difference between the two examples, in that the Figure (the syntactic subject) in ( 14) has already been described in the co-text as being on horseback.The English translator thus cannot be said to add any information as such.In (15) the other hand the protagonists are said to be in a crowded shop, but there is no suggestion in the original text that they experienced any particular difficulty in exiting it.This addition does not appear to be attributable to any of the four types of explicitation described by Klaudy (2008; see also Becher 2010).It is not obligatory, nor is it 'optional' in the sense that it is motivated by stylistic preferences in the two languages.'We left the shop together' would be perfectly idiomatic in English.It is certainly not pragmatic since Norwegian speakers and English speakers have similar cultures with respect to shops.This leaves us with Blum-Kulka's (1986) notion of 'translation-inherent explicitation', whereby the translator renders more explicitly something which is only implicit in the source text.However, as we have seen there is no reason to think that 'battling' as a mode of exiting is implicit in the original Norwegian text.So what we see in ( 15) is mere addition on the part of the translator, adding more colour to the predication than was expressed by the original author.
Turning from manner to path, we may first note the perhaps surprising fact that there are more tokens that omit to code path altogether in French than in English (path is of course always coded in the Norwegian originals).One such example is (16).( 16 In both the English and French versions of ( 17) the path is coded twice.More accurately one could say that two different portions of the path are coded in each case, the route by the verb and the source by the adverbial.This mode of double coding, which is referred to as double-framing in Croft et al. (2010), is more than twice as common in French as in English.
To sum up this discussion of the translations of ut av, we have seen that there is no categorical difference between the three languages insofar as they all make use of the same full range of constructions to code boundary-crossing self-motion events.The difference is rather one of proportion with English resembling Norwegian in the extent to which manner is coded explicitly and French doing so to a much lesser extent.With respect to path, although French does not code it more often than the two other languages, it tends to code it twice a good deal more often.In the next section we will investigate whether predications of [into] events resemble those of [out of] events in these respects.

Translations of Norwegian inn i
The coding of manner and path in the Norwegian tokens with inn i and the corresponding English and French translations is shown in Table 3.As was the case with ut av (Table 2), both manner and path may be encoded once or twice, or not encoded at all in the translated text.The data in Table 3 is reproduced figuratively in Figure 2. Again, Table 3 and Figure 2 show that English translations are closer to the Norwegian originals than the French translations with respect to the coding of manner and, in particular, the coding of path.The difference in the coding of manner is not quite as distinct here as in the [out of] predications, but the coding of manner is dropped in almost one third of the French translations (22 of 56), as illustrated in ( 18), where a path verb combined with a direct object encoding the Ground corresponds to a manner verb and path adverbial in the Norwegian and English texts.Here, half of the 22 tokens without coding of manner in the French translations combine the coding of path in the verb and path in a prepositional phrase, most frequently with the preposition dans, as in ( 19).The reason for this encoding is hard to determine, and the fact that ( 23) and ( 24) stem from the same novel may suggest that these are examples of a translator's idiosyncratic choices, rather than a general crosslinguistic tendency.
As in the [out of] predications, the English translations of [into] predications support the typological similarity of English and Norwegian, with the difference between a much more frequent use of path verbs with a Norman origin in English (e.g.enter, penetrate, join, merge, arrive, return, etc.) where only a few equivalents occur in the Norwegian originals (forsvinne 'disappear', komme 'arrive', falle 'fall').
If we go on to consider the encoding of path, the difference between French and English tokens in which path is not encoded is even higher here than in the [out of] predications: While all Norwegian originals encode path, 19 of the French and 7 of the English translations omit this coding altogether, as in ( 25).In ( 25), neither the English nor the French version contains path, which may be a result of the very "weak path" in the Norwegian original, where var kommet inn i stuen relates basically to their location ('in the living room') rather than the entering process.
There are 132 examples containing a predication of path in the Norwegian originals, 125 in the English translations and 113 in the French translations.Again, if we count the total number of path predications and not only the examples containing them, we get a different picture; 140 for Norwegian, 144 for English and 188 for French, which shows that the double coding of path is more common in French than in the other two languages.Both the English and French versions of ( 26) have path coded twice, and as in example ( 25), the route is coded by the verb and the goal by the adverbial.Summing up the discussion, the translations of inn i, like the translations of ut av, show that English and French use the same construction types as Norwegian to code self-motion.Again, the difference lies in the proportion of explicit coding between English and French: while the English translations tend to resemble the Norwegian originals in the coding of both manner and path, French has a lower degree of explicit coding with respect to both manner and path.However, here, too, the French translations tend to code path twice far more often than both English and Norwegian.

Summary and conclusion
In this paper we compared and contrasted English and French translations of Norwegian predications of motion events containing the boundary-crossing compound prepositions ut av ('out of') and inn i ('into').Our reason for analysing these boundary-crossing predications is that these are often said to disallow manner verbs in languages such as French.According to Beavers et al., "V-framed languages disallow boundary-crossing path satellites with manner verbs, although they may allow non-boundary-crossing path satellites" (2010: 347).These predications therefore furnish us with appropriate evidence for evaluating the hypothesis of Slobin (2006: 70) that "in translations […] manner salience follows patterns of the target, rather than source language", which has been disputed by Cappelle (2012), who maintains that translations are likely to show traces of the manner coding of the source texts.
The data for our study consists of self-motion predications extracted from the Oslo Multilingual Corpus (OMC) where path is coded in the prepositional phrases inn i and ut av, and where the verb may also code path, manner, or be a neutral verb of movement.Our main concern was the extent to which the English and French translators retain this original coding and, in cases where they do not do so, the sorts of changes they make.The results of our analysis were presented in sections 4 and 5. Figure 1 for [ut av] and Figure 2 for [inn i] showed the coding options utilised by the two sets of translators.The data in the two figures for the coding of both manner and path have been collated for purposes of comparison and are here presented in Figures 3 and 4. When it comes to the coding of manner difference between English and Norwegia French omits to code manner more often than the other two languages, with the translators choosing to drop manner coding in the original Norwegian text in 39% of instances in the case of [into] and 50% of instances in the case o translators do retain the coding of manner in at least half of all instances in the cases of both path types.manner in an adverbial rather than a verb, but there are still 31 instances of verbal manner coding number for a form of coding that is said in the literature to be strong disfavoured. 88 31 tokens correspond to a percentage total of 13.8%.This may be compared to totals of 16.1% for manner verbs in translations from Japanese comprise both boundary whereas ours all involve an element of boundary study, the predications with manner verbs occur in unambiguous boundary crossing predications in 37 study, our figure of 13.8% would approximate to about double the frequency one might expect to find in French origi ut of in English, French and Norwegian When it comes to the coding of manner, Figure 3 shows that there is little difference between English and Norwegian with either type of path.French omits to code manner more often than the other two languages, with the translators choosing to drop manner coding in the original Norwegian text in 39% of instances in the case of [into] and 50% of instances in the case of [out of].On the other hand, this means that translators do retain the coding of manner in at least half of all instances of both path types.In some instances they choose to code manner in an adverbial rather than a verb, but there are still 31 instances verbal manner coding, which must be accounted a surprisingly large number for a form of coding that is said in the literature to be strong 31 tokens correspond to a percentage total of 13.8%.This may be compared to manner verbs in French original texts and 14.8% for Fr translations from Japanese in Morita's (2011) study.His figures, however, comprise both boundary-crossing and non-boundary-crossing predications, whereas ours all involve an element of boundary-crossing.In Kopecka's (2009) edications with manner verbs occur in unambiguous boundary crossing predications in 37.7% of cases.If the same is true of those in Morita's study, our figure of 13.8% would approximate to about double the frequency one might expect to find in French original texts.

29
there is little n with either type of path.French omits to code manner more often than the other two languages, with the translators choosing to drop manner coding in the original Norwegian text in 39% of instances in the case of [into] and 50% of this means that the translators do retain the coding of manner in at least half of all instances In some instances they choose to code manner in an adverbial rather than a verb, but there are still 31 instances , which must be accounted a surprisingly large number for a form of coding that is said in the literature to be strongly 31 tokens correspond to a percentage total of 13.8%.This may be compared to 14.8% for French in Morita's (2011)   With respect to path, the resemblance between the codings of the two path forms is even more striking than in the case of manner.In both cases English has more double codings means as many as French.coding more often than they do single types, coding both route and source in [out of] predications and route and goal in [in to] predications.
To sum up, we have shown i difference between the three languages insofar as they all make use of the same range of constructions to code boundary-crossing sel proportion with English resembling Norwegian in the extent to which manner is coded explicitly and French doing so to a much lesser extent.As for the question of whether the coding of manner of motion in translations is likely to conform to the norms of the target language, as Slobin maintains, or to display the influence of the source language, as Cappelle argues, our results point to Cappelle's stance being the correct one.However, more study, both of Fre boundary-crossing self translations of French originals, is clearly necessary to further buttress this provisional conclusion.Nevertheless, the fact that there are as many as 31 tokens in the French translations in which manner is coded by the With respect to path, the resemblance between the codings of the two path forms is even more striking than in the case of manner.In both cases English has more double codings than Norwegian, though by no means as many as French.Indeed the French translators employ double coding more often than they do single-coding when it comes to both path types, coding both route and source in [out of] predications and route and n to] predications.To sum up, we have shown in this article that there is no categorical difference between the three languages with respect to types of framing insofar as they all make use of the same range of constructions to code crossing self-motion events.The difference is rather one of proportion with English resembling Norwegian in the extent to which manner is coded explicitly and French doing so to a much lesser extent.As for the question of whether the coding of manner of motion in anslations is likely to conform to the norms of the target language, as Slobin maintains, or to display the influence of the source language, as Cappelle argues, our results point to Cappelle's stance being the correct one.However, more study, both of French original predications of crossing self-motion entering and exiting, and of Norwegian translations of French originals, is clearly necessary to further buttress this provisional conclusion.Nevertheless, the fact that there are as many kens in the French translations in which manner is coded by the With respect to path, the resemblance between the codings of the two path forms is even more striking than in the case of manner.In both than Norwegian, though by no Indeed the French translators employ doublecoding when it comes to both path types, coding both route and source in [out of] predications and route and that there is no categorical with respect to types of framing insofar as they all make use of the same range of constructions to code motion events.The difference is rather one of proportion with English resembling Norwegian in the extent to which manner is coded explicitly and French doing so to a much lesser extent.As for the question of whether the coding of manner of motion in anslations is likely to conform to the norms of the target language, as Slobin maintains, or to display the influence of the source language, as Cappelle argues, our results point to Cappelle's stance being the correct nch original predications of motion entering and exiting, and of Norwegian translations of French originals, is clearly necessary to further buttress this provisional conclusion.Nevertheless, the fact that there are as many kens in the French translations in which manner is coded by the verb would appear to indicate that the translators have been influenced to some extent by the coding in the source texts.However, they would not have employed this form of coding as often as they do if the construction was anathema in the French language.Hickmann et al. (2009: 707) maintain that "although mixed, contemporary French is primarily verbframed with a reduced secondary satellite-framed subsystem".The data presented in this paper lend further support to the conclusion reached by Kopecka (2009) that this secondary satellite-framed subsystem may also be employed in coding boundary-crossing events.

Figure 1 :
Figure 1: Codings of Manner and Path in [out of] predications

(
19) a. De spaserte inn den hvite grinden og inn i hagen.(HW2) b.Then they strolled through the white gate and into the garden.(HW2TE) c.Ils ouvrirent la barrière blanche pour entrer dans le jardin.(HW2TF)In 32 tokens the French translators retain the coding of manner from the Norwegian originals.Of these tokens, 21 have encoding of manner in the verb, as in (20).In the remaining 11 tokens, manner is encoded in the adverbial, as in (21).However, only 4 of these 11 tokens have Norwegian originals where manner is encoded in the verb (which in all instances is gå 'walk'); in the rest of the tokens (most with the neutral verb komme 'come') manner is encoded in the adverbial in both the Norwegian original and the translations, as in(22).None of these French manner adverbial tokens take the form of a gerund (cf.section 4 above).(20) a. En fløy like inn i flammen foran Dina.(HW2) b.One flew into the flame in front of Dina.(HW2TE) c.L'un d'eux vola droit dans la flamme, devant Dina.(HW2TF) (21) a. Går langsomt ned hagegangen, inn i den ventende vognen.(NF1) b. ... and walk slowly down the garden path into his waiting carriage.(NF1TE) c. ... il descend lentement l'allée jusqu'à la voiture qui l'attend.(NF1TF) (22) a. ... en småjente med skoleransel på ryggen kom styrtende inn i rettslokalet.(BHH1) b. ... a little girl with a schoolbag on her back came rushing into the room.(BHH1TE) c. ... une petit fille portant un cartable sur le dos entra précipitamment dans la salle.(BHH1TF) In two of the tokens where the Norwegian original has no manner encoding, shown in (23) and (24), the French translators code manner in the verbs: (23) a. Men innen gjengen vinket farvel og forsvant inn i det forbrukerparadiset ... (BHH1) b. ... by the time the gang waved goodbye and disappeared into the consumer paradise ... (BHH1TE) c.Mais avant même que la bande n'ait eu le temps de s'engouffrer dans le grand temple de la consommation ... (BHH1TF) (24) a. [hun] krøp tettere inn til muren, som om hun ønsket å forsvinne inn i den.(BHH1) b. [she] crept closer to the wall, as if she wanted to disappear into it.(BHH1TE) c. [elle] se coller contre le mur comme pour s'y enfoncer.(BHH1TF)

Figure 3 :
Figure 3: Coding of Manner [into] and [out of] compared

Figure 3 :
Figure 3: Coding of Manner [into] and [out of] compared

Figure 4 :
Figure 4: Coding of Path [into] and [out of] compared

Figure 4 :
Figure 4: Coding of Path [into] and [out of] compared

Table 1 :
The classification system for manner and path encoding in the data

Table 2 :
Table 2 contains details of the coding of manner and path in the Norwegian original tokens containing ut av and the English and French translations of these tokens.As may be seen in the table, both manner and path may be encoded once or twice, or in the case of the translated text, not encoded at all.This latter case is obviously impossible for path in the Norwegian originals which were chosen because of the sort of path they code.Codings of Manner and Path in 93 [out of] predications Both the English and Norwegian versions of (16) contain manner and path adverbials in addition to a manner verb.The French version, on the other hand, contains neither, substituting the non-motion predicate 'se sauver' for the motion predicate in the original.There is no doubt an element of implicitation at work here insofar as the obvious way for the Figure to save herself is to flee her present location.However, (16) is not representative of French translations in this respect.There are 93 examples containing a predication of path in the Norwegian originals, 91 in the English translations and 87 in the French translations.If we count the total number of path predications, rather than the examples containing them, we get a different picture; 102 for Norwegian, 110 for English and 134 for French (cf.Table2 above).The difference is due to the extent of double coding in the three languages.

Table 3 :
Codings of Manner and Path in 132 [into] predications