Between and through revisited

Thomas Egan
Hedmark University College

Abstract

1. Introduction

The starting point for this paper is onomasiological. The concepts investigated are throughness and betweenness as they are encoded by the English prepositions through and between and their equivalents in three other languages, Norwegian, German and French. Both through and between are of common occurrence in English. According to Kennedy (1991: 95), “Between and through are among the hundred or so most frequently used words in English. Like most other structural words, they are semantically complex”. Kennedy distinguishes 12 different senses of both prepositions. He asserts that “The analysis of the semantic functions in which between and through occur in the LOB corpus […] suggests why these words may be difficult to learn or use. For example, both are associated with movement, time and a variety of other relationships” (Kennedy 1991: 109).

Kennedy’s (1991) study, which constitutes the original visit to the prepositional site revisited in the present paper, was based on data from the LOB corpus. This study is based on data from two corpora, the English-Norwegian Parallel Corpus (ENPC) and the Oslo Multilingual Corpus (OMC). These are described in section 2, which also contains a justification of the employment of data from translation corpora in the effort to cast light on the structure of polysemous networks. Section 3 examines the semantic network of the preposition/particle through as this is reflected in the Norwegian translations of the form in the ENPC. In section 4 I sketch out what I take to be the seven main senses of between and compare them to the senses distinguished by Kennedy (1991). Section 5 contains details of the translation equivalents of the various senses of between in the ENPC. Section 6 looks at the translations into English, French and German of the Norwegian preposition mellom, shown in section 5 to be overwhelmingly favoured by Norwegian translators as the equivalent of English between. Finally, section 7 contains a summary of the discussion and a comparison of the results with those of Kennedy (1991).

2. Translation corpora as valuable sources of evidence

Whereas Kennedy (1991) based his analysis of through and between on the  LOB corpus, I base mine on the translation equivalents of the forms found in the ENPC and the OMC (for both of which see Johansson 2007). The former contains a broad selection of both fictional and non-fictional English texts translated into Norwegian and Norwegian texts translated into English. Only the original English texts, some 50 in number, were accessed for this study. For translations from Norwegian into English I consulted the smaller OMC. Unlike the ENPC which contains extracts from a wide variety of texts, the OMC contains long extracts from just five novels. On the other hand all five are translated into three languages, English, French and German, allowing us to compare the interpretations of the most common Norwegian counterparts of both through and between in these languages.

Various studies have shown that translation corpora can reveal which lexemes or constructions in two or more languages are felt to correspond most closely to one another (see, for example, Dyvik 1998, 2004; Johansson 2007; Noël 2003). Such corpora provide us with access to the intuitive semantic judgments of competent speakers of both languages. Of course translations may display translation effects, the form of the item in translation being influenced by the form of the source item. However, in a study of the various senses of polysemous morphemes, in the present case prepositions, there is no a priori reason to expect more translation effects in the case of one sense than the other senses. Moreover, if one sense does show more such effects than the others, this in no way compromises the results of the investigation, based as it is on the enumeration of all sorts of differences between the translated forms of the various senses, irrespective of how these differences are motivated. Translation corpora may also provide evidence of the internal structure of the polysemous network of a single lexeme. In effect, different senses of a lexeme in language A which are usually  translated  into language B by one and the same lexeme (or construction) may be hypothesised to be more closely related within the semantic network of the lexeme in language A than those translated by different lexemes. Should translations into languages C and D display a similar patterning, this would serve to strengthen the hypothesis.  It goes without saying that any such hypothesis grounded on the testimony of translations should be evaluated in the light of other forms of (mono-linguistic) evidence, such as behavioural profiles (see, for example, Gries 2010).

3. Norwegian translations of through

There are in all 696 tokens of through in the English original texts in the ENPC. [1] All of these tokens were extracted and classified into one of eight main categories. Kennedy (1991) classifies the tokens of through which he extracted from the LOB corpus into six main classes, some of which are defined according to the means of progress along a path, such as ‘unimpeded motion’ and ‘penetration of a barrier or obstacle’, while others are defined according to the ontological domain of the state of affairs encoded (‘time’) or the semantics of the predication as a whole (‘causation’). In my analysis I group together predications encoding physical motion in the space domain, irrespective of whether they involve unimpeded motion or penetration and distinguish these from predications involving non-physical domains, which may also encode both these forms of motion. The description of through below is briefer than that of between in the sections that follow, as I have already dealt with through at some length in a previous publication (Egan 2012).

The eight categories employed in the present study are labelled Motion, Perception, Time, Channel, Means, Cause, Idiom and Other. They are exemplified in turn by in (1)–(8), starting with Motion. All tokens are from the English-Norwegian Parallel Corpus.

(1) encodes actual physical motion, with the trajector traversing the landmark (in the sense of Langacker 1990), although in this case neither entry into the landmark nor exit from it is implied. The motion type accounts for just over 40% of the tokens of through in the corpus. In these the landmark traversed by the trajector varies along various dimensions and may or may not be entered and exited. The second most common use of through is to encode perception, as in (2). In the world view of a physicist no doubt physical motion is involved in acts of perception, but as it is seems unlikely that the passage of light waves is part of the layman’s understanding of the act of viewing, it is preferable to group tokens of perception separately from those encoding motion by the trajector. In tokens like (3) the landmark is temporal, while tokens like (4) contain a landmark which is a channel or medium. The category Means (5) includes tokens in which the landmark is an instrument. Similarly, the Cause category contains permissives as well as causatives such as (6).  Idiomatic expressions containing through, like the ‘lie through one’s teeth’ construction in (7), were classified together, irrespective of whether it may have been possible to subsume them semantically under one of the other headings. The reason for classifying them together is the expectation that idioms are more likely to be language specific. The validity of this expectation will, of course, be revealed by the translations. Broad as some of them are, these seven categories cannot accommodate all the tokens in the corpus. It was therefore necessary to establish a category for residual tokens, 52 in number, which I labelled Other. Most of these residual tokens involve metaphorical extensions of some sort or another. This category might therefore have been labelled ‘various metaphorical’, instead of ‘other’. However, metaphor and/or metonymy are involved in many of the other senses, so it would be misleading to privilege this category by employing the term ‘metaphorical’ in its label.

Johansson (2007) distinguishes various degrees of convergence between original and translated forms. The main distinction he draws is between forms which are structurally similar, which he labels congruent, and forms which exhibit differences in structure, which he labels divergent. In the case of prepositions, congruent translations are those which contain a preposition, divergent translations those which employ a different form of construction. In the cases of through and between, we can further subdivide the congruent translations into those containing gjennom and mellom, the closest Norwegian correspondents of the two English prepositions, and those containing other prepositions. Figure 1 shows how often each of the eight mains senses of through is translated by gjennom, another preposition or a divergent construction.

The preposition gjennom accounts for more than half of the Norwegian translations of through. If one were to ask a Norwegian speaker of English for the Norwegian equivalent of through, they would almost certainly reply gjennom. The first definition of gjennom in Norsk Ordbok, the closest Norwegian equivalent to the OED, is: “used about a movement or something perceived as motion which takes place in (within, surrounded by) that which is encoded by the landmark from one end or side all the way to the other, containing the whole landmark from start to finish; (in) from the one side or end and (out) to the other” (Vikør 2002, my translation). This definition is very similar to the standard definitions of through in English. For instance, according to Tyler and Evans (2003: 219) “the relation described by through describes a spatial relation in which a bounded LM [landmark] is transected by virtue of an entrance point and an exit point.” In the non-technical language employed by Lindstromberg (1998: 31) the central meaning of through is said to be “in and out again”. In the same vein the OED’s first definition includes the phrase: “From one end, side, or surface to the other or opposite end, side, or surface of (a body or a space) by passing within it; usually implying into, at one end, side, etc. and out of at the other”. Given the degree of similarity between these definitions of through and the cited definition of gjennom,it may come as no surprise to find that gjennom is used to translate tokens in all eight classes of through. (9)–(16) illustrate all eight in descending order of frequency of translation by gjennom, as shown in Figure 1. In each case I add the Norwegian translation of the relevant phrase.

As shown in Figure 1, gjennom is used to translate 80% of the perception tokens, as in (9), while the cause sense is the least likely to be translated by gjennom, with just 15% of tokens represented by that form. Moreover, not a single token of causation proper is so translated, (16) being an instantiation of the enablement subtype.

There are 187 tokens in which through is translated by a preposition other than gjennom. By far the most common of these is i, the Norwegian equivalent of in, which is used for 52 tokens belonging to six of the eight classes, the exceptions being idiom and channel. No other single preposition approaches i in frequency as an alternative to gjennom. Two prepositions occur 13 times. These are mellom, the Norwegian equivalent of between, and over, the Norwegian equivalent of across (we will return to the question of overlap between between and through in section 7). There is only one other preposition which is represented among the translations by more than 10 tokens. This is ved, the Norwegian equivalent of English by, which only occurs with the cause and means senses. Other prepositions utilised by translators, with their closest English counterparts, include til (to), fra (from), bak (behind), forbi (past), med (with), på (on), via (via) and under (under, but also during). Figure 1 shows that it is the cause, means and temporal senses of through that are most likely to be translated by a preposition other than gjennom.

The third category of translations of through distinguished in the present study is by means of a divergent construction. As shown in Figure 1 the ‘idiom’, ‘time’ and ‘other’ senses are the only ones in which more than 20% of the tokens are translated in this fashion. (17)–(19) illustrate divergent translations of these three senses.

Although Norwegian does have quite a few multi-word verbs, they are not nearly as common as in English. One option, chosen by the translator of (17), is to employ a verb corresponding to the head of the English multi-word verb. Something inevitably gets lost in the process of translation. In (17) the Norwegian translation does not convey the element of completeness in the English original. If you talk something through you cover the ground in a more comprehensive manner than if you simply talk about it. In the time example, (18), the translator prefers to view the period as a static whole, ‘all evening’. Finally, the translator in (19) has opted to encode the manner of explanation in an adverbial rather than in the verb.

Figure 1 contains percentages for the three translation options of each of the eight senses of through, in descending order of preference for the maximally convergent option, gjennom. A quick glance at the figure should suffice to conclude that some senses, such as motion and channel, are translated in quite similar fashion, while others, such as perception and cause, differ greatly in this respect. However, before we go on to draw any conclusions on the basis of these differences, we must establish which of them are statistically significant. Table 1 contains details of the chi-square probability values for all 28 combinations of two of the eight senses.

Table 1 shows that there are significant differences in mode of translation between the ‘perception’ sense and the other seven senses. The ‘cause’ sense resembles the ‘perception’ sense in having no non-significant counterparts, with ‘means’ and ‘time’ being closest to it. The ‘time’ sense itself is translated in a similar fashion to both ‘other’ and ‘means’, being significantly different to the remaining five others, while the ‘motion’ sense is significantly different to all senses other than channel. Finally, the ‘channel’, ‘other’ and ‘means’ senses all show non-significant differences with three of the other senses. If we were to apply the more stringent criterion for significance of p=0.01, however, ‘perception’ would not be significantly different from ‘channel’ and ‘motion’, and ‘cause’ from ‘means’ and ‘time’. ‘Idiom’ would, however, remain significantly different from all other senses.

The statistics in Table 1 represent a network of similarities and differences between the way throughness is encoded in English and Norwegian. One may legitimately ask whether the table may be interpreted as telling us something about English in itself (or indeed Norwegian) in addition to informing us about the similarities and differences between the two languages. My hypothesis, and at the present stage of inquiry it remains very much a working hypothesis, is that, when interpreted with care, these sorts of translation data can indeed cast some light on aspects of the polysemy of an item in the source language. As stated in section 2, any conclusions reached on the basis of translation equivalents should be subject to verification using other methods. However, even if the data in Table 1 should prove unsuited to a description of the polysemy of English through as such, they may still be of value in a cross-linguistic context, not least when compared with data for other prepositions. In section 5 we will see the extent to which translations of between differ from those of through sketched in this section. Before doing so, however, we must establish the main senses of between in English.

4. The seven main senses of between

Kennedy (1991) distinguishes five main senses of between. Three of these, Location, Movement and Time are further divided into two main sub-senses and a fourth, termed Other Relationships, has five sub-senses. In the present study I distinguish seven main senses, Location, Motion, Scale, Time, Relationship, Interaction and Comparison, and one minor sense Idiom, which is only instantiated in five tokens. The correspondence between my seven main categories and those of Kennedy (1991) is shown in Table 2.

There are 486 tokens of between in the English language original texts in the ENPC, as compared to 867 in LOB. The distribution of the tokens across the various semantic types is quite similar, as may be seen in Table 3, in which the totals for Kennedy’s categories are converted to those employed in this study.

The (marginally) most common sense of between in both LOB and the ENPC is Location, illustrated in Figure 2 and exemplified by (20)–(24).

The space occupied by the disc in Figure 2 may extend from the circumference of a to that of b, as is the case in (20), or it may just occupy a small space at some remove from both, as in (21). Sometimes the exact extent of the space is underspecified, as in (22) in which there is no way of telling whether the garden in question occupies the whole area between a and b. Sometimes the space contains multiple points (or, alternatively, there may be multiple spaces) as is the case in (23). Finally, some few examples resemble (24) in encoding a metaphorical, rather than a physical, space. As there were only a handful of such examples in the corpus it was not felt necessary to establish a separate category for them, as was the case with the ‘other’ category for through.

The Motion category, which overlaps completely with Kennedy’s Movement category, is illustrated in Figure 3 and exemplified by (25)–(27).

Note the direction of the arrow in Figure 3. Although it is possible for the motion along a path encoded by between to start at a and finish at b, as is the case in example (26), most tokens in my material encode movement orthogonal to the a–b axis. (25) is thus more typical of the construction than (26). Indeed, cases encoding oscillation, i.e. motion from a to b and back again, as in (27), are actually more common than cases of single direction a to b motion.

As can be seen in Table 2, Kennedy subsumes tokens encoding a scale or cline under Location. I prefer to categorise them separately, and illustrate Scale in Figure 4, which differs in one important respect from Figure 2, illustrating Location.

Whereas a location encoded by between may be situated along the axis stretching from a to b or from b to a – Narbonne may be said to be situated between Montpellier and Perpignan or between Perpignan and Montpellier – scalar between always incorporates an element of direction. Thus (28) could not be paraphrased by “between eight and six hundred pounds” nor (29) by “between a half and one-third”. Nor could someone be described as being between twenty-five and twenty-two years of age. Scalar between shares this sort of directionality with temporal between, which is illustrated in Figure 5 and exemplified by (31)–(34).

The time span encoded in (31) cannot be understood as the 19 hours stretching from midnight to the previous 5 a.m. Nor could the periods in (32) and (33) be encoded as ‘between ten and nine that night’ or ‘between 1860 and 1840’. Similarly the period from one call to another in (34) must refer to the time between a particular call and subsequent calls, not between a call and previous ones.

Directionality is also involved in the Relationship sense of between, more specifically bidirectionality, as illustrated in Figure 6 and exemplified by (35)–(38).

In the relationship sense of between, the two participants a and b are both (usually) human. Semantically they are normally Experiencers. The arrow in Figure 6 is double-headed as the construction encodes an experience that is mutual. Thus in (35) and (36) empathy and attraction are feelings experienced by both the people in question. The negative state of estrangement, encoded by ‘an iron curtain’ in (37), is similarly mutual to the participants.

Unlike the Experiencers in the relationship sense, the participants in the Interaction sense, illustrated in Figure 7 and exemplified by (38)–(41), are Agentive.

The two arrows in Figure 7 illustrate the reciprocal nature of the actions of a and b in situations encoded by the ‘interaction’ sense of between. In (38) the groups clash one with another and in (39) both parties to the argument are active and both direct their arguments towards the other party. In both cases the form of interaction is encoded by the trajector and the participants by the landmark. Similarly in (40) and (41) the participants are reciprocally active in sex and negotiation respectively.

The seventh and last of the major senses of between is the Comparison sense, which subsumes tokens encoding choice, similarity and difference. It is illustrated in Figure 8 and exemplified by (42)–(45).

The ‘comparison’ sense differs from the other senses in that it presupposes the presence of a source of comparison, represented by y in Figure 8, who is not actually encoded in the between construction, although it may well be present in the co-text, as Aristotle is in (42) and the generic animal is in (43). In this construction the trajector is the conclusion reached by y, who may be either explicit or implicit, while the landmark encodes the objects of comparison. Thus in (42) Holland and Athens are compared by Aristotle, who reaches the conclusion that they are similar in some ways. And in (43) ‘other zebras’ are compared by the individual animal who is able, in the hypothetical space of the if clause, to detect differences between them. (44) is similar to (43) in that differences may be, or in this case not be, detected, but the distinguisher is not explicitly encoded in the co-text.  In (45), too, the comparer is implicit. Moreover in (45) the comparer not only arrives at a conclusion with respect to some qualities of a and b but is expected to act upon this conclusion.

Figures 3–9 illustrate the seven main senses of between. There remain 5 tokens which display a greater or lesser degree of idiomaticity. (46)–(48) are three of them.

It might be possible to argue that (46) should be classified with the time tokens as the sense of ‘at sporadic intervals’ encoded by few and far between belongs to the temporal semantic field. On the other hand, there is no connotation in the idiomatic expression of the element of direction (from earlier to later) which we saw to be typical of time between. One could make a better case for subsuming draw the line between in (47) under the comparison sense. The decision not to do so is motivated by the desire to be consistent in the treatment of idioms. Unlike the first two examples (48) does not lend itself to inclusion in any of our seven classes. It is completely idiomatic and may be most appositely paraphrased by the equally idiomatic what with construction.

5. Translations of between into Norwegian

Having described the seven major (and one minor) senses of between, we now turn to their translation equivalents in Norwegian. Just as in the case of through, there are both congruent and divergent translations in the ENPC, with one preposition, (i)mellom (cognate with English middle) standing out in terms of frequency. It is used to translate all seven main senses of between, as may be seen in Figure 9 and examples (49)–(55), listed in the order they appear in the figure.

We saw in section 3 that there are 187 tokens in which through is translated by a preposition other than gjennom. The corresponding figure for between is only 39. Seventeen of the tokens with an alternative preposition take the form fra … til (= from … to). The from … to equivalent occurs in translations of six of the seven senses of between, the exception being the ‘interaction’ sense, perhaps because of the dual directionality connoted by this sense. The second most common form is i (=in) with ten occurrences divided between the five senses other than ‘time’ and ‘relationship’. There are four occurrences of på (=on), one each encoding ‘comparison’, ‘motion’, ‘scale’ and ‘time’. There are two instances of av (=of), in both of which “discrimination between x and y” is rendered “discrimination of x and y”. There are four tokens where the translator focuses on just one of the two poles, a or b. One of these contains the preposition fra (= from) without til, and one til (=to), without fra. The other two contain the prepositions foran (=in front of) and ved siden av (=beside). Finally there are two instances of under (=under), in both of which the landmark is ‘sheets’.

Just as there are fewer alternative prepositions utilised by translators for between than through, there are also fewer divergent constructions. Moreover, not only are there fewer tokens of between translated by prepositions other than the most common equivalent than is the case with through, but there are also fewer preposition types utilised by the translators. In addition, all of the prepositions that are used more than twice are used for multiple senses of between, reinforcing the impression conveyed by Figure 9, that there is more agreement among translators with respect to the semantic overlap of the various senses of between than in the case of through. One might therefore expect that there would be fewer non-significant differences between the translation equivalents of the various senses than was the case with through, presented in Table 1. This is indeed the case, as may be seen in Table 4.

If we compare Table 4 to Table 1, we may note first of all that there are far fewer significant differences in the translation equivalent patterns of the various senses of between than of through. Betweenness thus seems to be a ‘tighter’ concept cross-linguistically, at least with respect to these two languages, than throughness. One may also remark that the central place in the polysemous network, in the sense of the meaning whose translations have most in common with the other meanings, is not the most common sense (location) but rather the scale sense. In the next section we will investigate how encodings of betweenness in Norwegian are reflected in translations into English, French and German.

6. Mellom in the Oslo Multilingual Corpus

We saw in the previous section that the Norwegian preposition mellom is by far the most popular translation equivalent of between. We also saw that it could be used to translate all seven main senses of between. It also occurs in all these seven senses, as well as in idiomatic expressions, in Norwegian original texts, as may be seen in Table 5, which contains details of the distribution of mellom across all eight senses of betweenness with comparable details for between.

Mellom is cognate with English middle, and is most commonly translated into English by between (73.3%), into German by zwischen (77.8%) and into French by entre (68.6%). Of the eight senses of mellom listed in Table 5, no further mention will be made of the scalar sense, as it is only evidenced by five tokens. The other seven senses will be discussed in turn, starting with the ‘location’ sense, exemplified by (56a-d).

The translations in (56b-d) are typical of the strategies adopted by translators into all three languages in that they contain in each case the standard translation equivalent of mellom. Figure 10 shows that more than 70% of the English and French translations contain between and entre and over 80% of the German translations zwischen.

It is only the English translations in which over 20% of the total number of tokens contain another preposition. In 23 of 31 instances the preposition in question is among(st), as in (57), which is typical of these tokens in that the landmark is a plural countable.

When the landmark may be more appropriately illustrated as a + b + c +….. than just a + b, as in Figures 3–9, betweenness is often expressed in English by among. Interestingly the French translator of (57) renders the locational predication as au milieu des livres. There are four other tokens of au milieu de translating tokens of mellom which are translated into English by among, but most of the latter are simply translated into French by entre.

The ‘motion’ sense of mellom may also be translated into all three languages by between/entre/zwischen, as illustrated in (58a-d), but there is greater variation both within each of the three languages and between them than is the case with the ‘location’ sense, as may be seen in Figure 11.

In all three languages a greater number of tokens are translated using another preposition than is the case with the tokens encoding location. Moreover, in all three languages a variety of alternative prepositions are employed by translators, 11 in French and 10 in both English and German. The most common preposition in English is again among, used with plural countable landmarks, and in French it is au milieu de, in the same sort of context. In German the most common preposition other than zwischen is durch, the standard translation of through. It is used to translate five tokens of mellom. The ‘motion’ sense of between is the only one of the eight senses to be so translated. Similarly, in English, through is the second most common alterative preposition, represented by nine tokens, and again it only occurs as a translation of the ‘motion’ sense. In French, à travers is the third most common alternative preposition, after au milieu de and parmi. Four of the five tokens of à travers occur with the ‘motion’ sense, the fifth occurring with the ‘location’ sense. If we combine the figures for the standard through equivalents, we see that 18 of 19 instances are found in translations of tokens encoding motion.

Why should the standard equivalents of through be used to translate the ‘motion’ sense, and only the ‘motion’ sense, of mellom? The answer lies in the direction of motion typically encoded by the construction, as illustrated in Figure 3, i.e. as perpendicular to the a-b axis. The trajector enters the space between a and b, traverses that space and exits from it, thus realising the prototypical entry-traversal-exit trajectory of throughness. In cases like this we find the sort of overlap between the notions of betweenness and throughness that prompts Kennedy’s (1991) assertion that the English prepositions may be difficult for the foreign learner to acquire. An indication of the degree of semantic overlap may be gleaned from the fact that the three throughness prepositions are in no case used to translate one and the same instance of mellom, although we do find instances of two of them being so used, as in (59 a-d) and (60 a-d).

In both (59) and (60) the translator is faced with the task of rendering a token in which betweenness overlaps with throughness. In both cases the trajector, Tomas in (59) and the light in (60), traverses a space orthogonal to the landmark(s). To spell out the path traversed by the trajector, in (59) Tomas does not climb from one curtain to another, while in (60) the light does not move from one eyelid to the other! Faced with the choice of emphasizing the parallel location of the landmarks, or the path along which the trajector moves, the translators choose different options. Both English translators choose to emphasise the throughness, as does one of their French and one of their German counterparts. One German translator opts to emphasise betweenness while one French translator opts for beneathness.

Whereas translators into all three languages employ a variety of prepositions in their translation of ‘motion’ mellom, only one alterative preposition (among) is employed by English translators of the ‘interaction’ sense, one alternative preposition (zu) by German translators and none at all by French translators. In all three languages the standard preposition, as exemplified by (61a-d), is the preferred form, as may be seen in Figure 12.

As shown by Figure 12, some 30% of the French translations contain divergent constructions. (62c) may serve as an illustration of these. The English translation in (62b) is also divergent in form, while the German employs the standard zwischen. Although all three languages contain divergent translations, there is no instance in which one and the same token is realised by divergent constructions in all three languages.

The tendency to avoid alternative prepositions to encode the ‘interaction’ sense of mellom is shared by translators of the ‘relationship’ sense, illustrated by (63a-d).

Translators into all three languages overwhelmingly prefer the default preposition to translate the ‘relationship’ sense. There is only one token of another preposition in English and none whatsoever in French or German. Again the English preposition is among. Both the French and German translators employ more divergent constructions than the English ones. (64a-d) is one of the few instances in which all three translators make use of divergent forms, in this case possessive pronouns, to encode the landmark of mellom in the original Norwegian text.

‘Comparison’, illustrated by (65a-d) is yet another sense in which translators employ very few alternative prepositions, as may be seen in Figure 14.

Two English tokens are translated using from. Both of these encode a choice between plural entities, ‘choose from’ being preferred by the translator to ‘choose between’. Neither the French nor the German texts contain a preposition other than entre/zwischen.

When it comes to translating temporal tokens, as in (66a-d), between, entre and zwischen are again the preferred forms, as shown by Figure 15.

In fact of a total of 16 tokens of ‘time’ mellom translated into all three languages, there are nine which are translated by all three default prepositions, as in (66a-d). This is in marked contrast to the 21 translations of time gjennom, not a single one of which is translated by through/à travers/durch. Indeed, of 630 possible combinations of  similar translations of gjennom into two of the three languages, the 21 Norwegian originals yield in all 9 pairs, three of through/durch, two of through/in, two of during/während and two of durant/während. The point was made above that encodings of betweenness in these four languages resemble one another to a greater extent than do encodings of throughness. This is nowhere more true than in the case of temporal predications. 

The seventh class of mellom tokens comprises idiomatic expressions. Between 40% (German) and 60% (English) of these are translated by the default forms, as in (67a-d).

It will come as no surprise that the translations of idiomatic tokens feature a greater number of divergent constructions than any of the other senses. On the other hand, one may well be surprised at the extent of overlap between languages. In all four one may fall between several stools, read between lines and experience philosophical shortcomings in relation to things between heaven and earth.

Figures 10–16 show that in all seven of the most common senses of mellom, the default translation equivalent is between/entre/zwischen. Of 412 tokens of mellom in OMC translated into all three languages, there are 186 combinations (45%) of the three standard prepositional translations. This is in marked contrast with gjennom, 295 tokens of which are translated into all three languages, yielding just 55 combinations (19%) of through/à travers/durch. Again we see that throughness is encoded in a much more varied fashion cross-linguistically (at least with respect to the four languages that furnish the data for this paper) than is betweenness. The two will be the subject of further comparison in the next and final section.

7. Summary and conclusion

In section 1 I cited Kennedy’s assertion that through and between “may be difficult to learn or use. For example, both are associated with movement, time and a variety of other relationships” (Kennedy 1991: 109). The degree of difficulty they pose to the learner depends on the degree of semantic overlap between them. However, we saw in the previous section that of the 419 tokens of mellom in the OMC, only nine are translated by through, five by à travers, and four by durch, these being the standard translation equivalents of Norwegian gjennom. Moreover, all of these tokens are spatial and all but one encode motion. If we turn to the question of how often the standard equivalents of Norwegian mellom are used to translate tokens encoded by gjennom, we find that of the 321 tokens of gjennom in the OMC, none are translated by between, none by zwischen and just one by entre.  The data in the ENPC yield similar results. Just 13 of the 690 tokens of through are translated by mellom (=between). Eight of these encode motion while four are perceptual. In all cases the landmark is a plural countable. (68) is an example of the ‘motion’ sense and (69) of the ‘perception’ sense.

(68) and (69) instantiate the ‘motion’ and ‘perception’ senses of through respectively. In each case the landmark is a plural countable. While English can profile each of these landmarks as a mass, Norwegian cannot readily do so. Note that one could employ between in English in both (68) and (69) with little change in meaning. Finally we may ask how often between in the ENPC is translated by gjennom. The answer is that none of the 486 tokens of between are so translated.

To sum up, our cross-linguistic data show that there is actually very little semantic overlap between the two concepts of betweenness and throughness.It therefore seems highly unlikely that they will pose the problems anticipated by Kennedy for the learner, at least for the foreign or second language learner who has already mastered one of the four languages in this paper and is endeavouring to learn another. This is especially true of between/entre/zwischen/mellom, given that encodings of betweenness are more similar, cross-linguistically, than are encodings of throughness. From a pedagogical point of view we may conclude that, of the two English prepositions, through is likely to present a greater challenge for ESOL teachers.

The data for this paper have been taken from translation corpora. The analysis of these data has led us to conclusions at variance with those of Kennedy, who was working with a single language corpus. To the extent that the conclusions in the present paper are deemed to be valid, this may be taken as support for the employment of translation corpora in studies of polysemy.

Notes

[1] The total of 696 tokens includes both prepositions and particles. In this paper I use ‘preposition’ as the default term and do not distinguish between the various syntactic realisations of either prepositions or particles. Thus the study includes all cases where the prepositions/particlesoccur as part of a multi word verb.

[2] ‘AH1’ identifies the text in the ENPC. For details of texts and translators see http://www.hf.uio.no/ilos/english/services/omc/enpc/.

Sources

ENPC = English-Norwegian Parallel Corpus: http://www.hf.uio.no/ilos/english/services/omc/enpc/

OMC = Oslo Multilingual Corpus: http://www.hf.uio.no/ilos/english/services/omc/

References

Dyvik, Helge. 1998. “A translational basis for semantics”. Corpora and Cross-Linguistic Research: Theory, Method, and Case Studies, ed. by S. Johansson & S. Oksefjell, 51–86. Amsterdam: Rodopi.

Dyvik, Helge. 2004. “Translations as semantic mirrors: From parallel corpus to wordnet”. Advances in Corpus Linguistics: Papers from the 23rd International Conference on English Language Research on Computerized Corpora (ICAME 23), Göteborg 22–26 May 2002, ed. by K. Aijmer & B. Altenberg, 313–326. Amsterdam: Rodopi.

Egan, Thomas. 2012. “Through seen through the looking glass of translation equivalence: A proposed method for determining closeness of word senses”. Corpus Linguistics: Looking Back - Moving Forward, ed. by S. Hoffman, P. Rayson & G.N. Leech. Amsterdam: Rodopi.

Gries, Stefan Th. 2010. “Behavioral profiles: A fine-grained and quantitative approach in corpus-based lexical semantics”. The Mental Lexicon 5(3): 323–346.

Johansson, Stig. 2007. Seeing through Multilingual Corpora: On the Use of Corpora in Contrastive Studies. Amsterdam: John Benjamins.

Kennedy, Graeme. 1991. “Between and through: The company they keep and the functions they serve”. English Corpus Linguistics: Studies in Honour of Jan Svartvik, ed. by K. Aijmer, J. Svartvik & B. Altenberg, 95–110. London: Longman.

Langacker, Ronald W. 1990. Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin: Mouton de Gruyter.

Lindstromberg, Seth. 1998. English Prepositions Explained. Amsterdam: John Benjamins.

Noël, Dirk. 2003. “Translations as evidence for semantics: An illustration”. Linguistics 41(4): 757–785.

Tyler, Andrea & Vyvyan Evans. 2003. The Semantics of English Prepositions: Spatial Scenes, Embodied Meaning and Cognition. Cambridge: Cambridge University Press.

Vikør, Lars S., ed. 2002. Norsk Ordbok, Vol. IV. Oslo: Det Norske Samlaget.